Stephanie Green ishVincent Hayw ardhaywardcimmcgillca

Center for Intelligent MachinesMcGill University3480 University Street

Montreal Qc Canada H3A 2A7

Vanessa ChialA llison Ok amuraDepartment of MechanicalEngineering

The Johns Hopkins UniversityBaltimore MD

Thomas Stef f enOrthopaedic Research LaboratoryRoyal Victoria Hospital

Montreal Quebec Canada

Presence Vol 11 No 6 December 2002 626ndash651

copy 2002 by the Massachusetts Institute of Technology

Measurement Analysis andDisp lay of Hapt ic Signals DuringSurgical Cutt ing

A bstract

The forces experienced while surgically cutting anatomical tissues from a sheep and

two rats were investigated for three scissor types Data were collected in situ using

instrumented Mayo Metzenbaum and Iris scissors immediately after death to mini-

mize postmortem effects The force-position relationship the frequency compo-

nents present in the signal the signicance of the cutting rate and other invariant

properties were investigated after segmentation of the data into distinct task phases

Measurements were found to be independent of the cutting speed for Mayo and

Metzenbaum scissors but the results for Iris scissors were inconclusive Sensitivity to

cutting tissues longitudinally or transversely depended on both the tissue and on

the scissor type Data from cutting three tissues (rat skin liver and tendon) with

Metzenbaum scissors as well as blank runs were processed and displayed as haptic

recordings through a custom-designed haptic interface Experiments demonstrated

that human subjects could identify tissues with similar accuracy when performing a

real or simulated cutting task The use of haptic recordings to generate the simula-

tions was simple and efcient but it lacked exibility because only the information

obtained during data acquisition could be displayed Future experiments should ac-

count for the user grip tissue thickness tissue moisture content hand orientation

and innate scissor dynamics A database of the collected signals has been created

on the Internet for public use at wwwcimmcgillca haptictissuedatahtml

1 Introd uct ion

Virtual surgery holds many promises It could be used to train surgeonsin virtual environments without the need for cadavers It could also allow ex-perienced surgeons to practice new techniques accommodate a ldquobrushing-uprdquofor rarely used techniques and help plan an operation Furthermore it couldenable the planning and evaluation of completely new surgical techniques andeventually serve as a prediction method for their outcome (KuppersmithJohnston Jones amp Herman 1996)

Realism in a simulation is increased with the addition of each sensorial mo-dality but at present most systems lack a good integration of tactile feedbackin part due to hardware limitations Very few instruments and operating condi-tions can be reproduced Other limitations arise from the lack of computa-tional models representing the mechanical phenomena underlying haptic inter-action Yet other limitations come from the lack of knowledge of the factorsthat contribute to realism in the simulation of haptic tasks

626 PRESENCE VOLUME 11 NUMBER 6

Current simulations in surgical procedures are mostlyif not completely visual in the cutting aspect This workis motivated by the need for haptic displays not to belimited to ldquopokingrdquo and ldquopullingrdquo but also to addresssuch basic tasks as cutting with scissors Virtual simula-tions when carefully constructed from measured dataand conveyed via adequate devices could actually bemore realistic than cadaver tissues which have signi-cantly different mechanical properties than do livingtissues These simulations will also be more practicalthan physical phantoms which cannot be reused Inaddition the need to sacrice animals for surgical train-ing or dissection is reduced

This paper investigates the forces experienced by asurgeon while cutting different tissues The data wereanalyzed to answer the following questions Is there aquantiable difference when cutting different tissues Isthe force measured invariant with the velocity Are thereany invariant properties of the tissues or of the scissorsthat can be determined Is it possible to design computer-controlled scissors that can pass a test of realism Cansubjects identify tissues by feel alone and how does thisperformance degrade when cutting virtual tissues

2 Prev ious Work

One of the earliest efforts in virtual surgery wasthe Green Telepresence Surgery System which was de-veloped at SRI International and included a surgeonrsquosconsole and a remote unit that performs operations on apatient (Green Jensen Hill amp Shah 1993) SRI has alsodeveloped the MEDFAST system a battleeld version toprovide surgical services in dangerous environments(Satava 1996) wherein a surgeon operates in a virtualspace at a console and the motions are translated by aslave robot to a remote patient Hunter Jones SagarLafontaine amp Hunter (1995) pioneered this conceptfor microsurgery Baumann Maeder Glauser andClavel (1998) described a force-reecting system capa-ble of addressing laparoscopic procedures as did Bas-dogan Ho Srinivasan Small and Dawson (1998)

Other research in the development of surgical simula-tors was the work of Delp Rosen and associates of

MusculoGraphics and MedicalMedia who developedthe anatomical rendering tissue interactions and surgi-cal instruments for a specic surgical instance a gunshotwound in the thigh Delp Loan Basdogan and Rosen(1997) presented a three-dimensional interactive com-puter model of the human thigh which was constructedfrom data provided by the Visible Human Project (Ack-erman 1994) Similarly Merrill Higgins and col-leagues of High Techsplanations developed the ana-tomic rendering the tissue interactions and surgicalinstrumentation for a shattered kidney (Satava amp Jones1997) Eisler of Mission Research Corp worked in tis-sue damage as a result of a ballistic wound for whichCuschieri from the University of Dundee supplied basictissue properties (Satava amp Jones 1997)

Several researchers built systems to display ldquointernalforcesrdquo between the thumb and nger for surgical tasksHannaford and colleagues (Hannaford et al 1998MacFarlane Rosen Hannaford Pellegrini amp Sinanan1999 Rosen Hannaford MacFarlane amp Sinanan1999) designed a force feedback endoscopic grasperthat displayed forces measured by a teleoperated grasperto the userrsquos thumb and index ngers through a masterhaptic device They demonstrated that subjects coulddiscriminate between six different tissue types duringteleoperated grasping using force feedback aloneResolved-force haptic devices were used to display externalcutting forces of a single blade in surgical procedures(Mahvash amp Hayward 2001 Biesler amp Gross 2000Hirota Tanaka amp Kaneko 1999) In addition cuttingwith scissors was modeled without the display of inter-nal cutting forces by Picinbono Delingette andAyache (2000) and Basdogan Ho and Srinivasan (1999)Other related works were concerned with the recordingof the interaction of the jaws of a laparoscopic surgicalinstrument with tissues (Bicchi Canepa De Rossi Iac-coni amp Scilingo 1996) with internal force feedbackdevice transparency (Longnion Rosen Sinanan ampHannaford 2001) or with the puncture-force responseof spleen and liver in animals (Carter Frank Davies ampCuschieri 2000) Several efforts are now directed atobtaining in vivo measurements of the mechanical prop-erties of biological materials (Vuskovic Kauer Szekely

Greenish et al 627

amp Reidy 2002 Ottensmeyer amp Salisbury 2001Brower et al 2001)

The present study of using haptic recordings ts intoa larger class of research devoted to the use of real-world measurements in haptic rendering The mostcomprehensive system for acquiring data for reality-based modeling thus far is the Active Measurement Fa-cility (ACME) built by Pai Lang Lloyd and Woodham(2000) They have obtained reectance stiffness tex-ture and sound data by probing objects t these datato analytical and empirical models and generated realisticvirtual environments (Richmond amp Pai 2000) MacLean(1996) has created realistic virtual environments usingthe concept of a ldquohaptic camerardquo that captured hapticinformation about real environments Other examplesof empirical haptic models include work on friction byRichard Cutkosky and MacLean (1999) and vibrationsby Okamura Dennerlein and Howe (1998) and Oka-mura Hage Dennerlein and Cutkosky (2000)

3 Materials and Methods

Data were acquired with three instrumented surgi-cal scissors Mayo dissection scissors Metzenbaum dis-section scissors and the Iris scissors The last authorprovided general directions and professional expertise toprepare three separate animal experiments at the ap-proved McGill University animal-care facilities The datacollected are from his hand and that of his students Alldata collection was performed on animals that had beensacriced for other research and in accordance with ani-mal protection laws Because they were planned sacri-ces it was possible to perform the experiments imme-diately after death minimizing postmortem effects

3 1 Instruments

Among the thousands of surgical instrumentsthere exist distinct classes of basic handle designs Amost common class is that of the scissor handles Forscissors many variations exist in the handles and bladesThe blades can have different lengths and breadthsThey can have teeth or not be curved or straight and

pointed or blunt Furthermore the blades can be sharpor dull tight or loose depending on the joint screw andeach of these factors inuences the ldquofeelrdquo of the scissorFor the scissor the central aspect of the haptic experi-ence results from closing them around a tissue regionand hence the scissor angle is the primary state variable

It was learned that surgeons often prefer a curved tipwhich allows more visibility of the work area The in-struments used for experimentation were a 7 in (178cm) Metzenbaum scissor with tungsten carbide insertsa 675 in (171 cm) Mayo scissor with tungsten carbideinserts and a pair of 45 in (114 cm) Iris scissors (allwith curved tips) Piling-Weckrsquos gold-handled scissorswere selected for their smooth and sturdy feel (See g-ure 1)

3 2 Tissues

The tissues compared were from a sheep and tworats Tissues were selected because of their accessibilityand relative abundance but also because of their dis-tinct properties skin (abdominal region) the abdominalmuscle wall the liver and the calcaneal (or Achilles)tendon The three layers of the abdominal muscle wall

Figure 1 The instruments used (from left to right Mayo

Metzenbaum and Iris as in gure 5)

628 PRESENCE VOLUME 11 NUMBER 6

(oblique external internal and transversal abdominal)were cut as one tissue For the sheep experimentsmuscle-rectus fascia was also cut because it was availablein abundance in the fatty abdominal area of the sheep

321 Tissue Preparation For each run a pieceof tissue was prepared and isolated from the surround-ing layers This involved scraping away excess fascia tokeep the thickness relatively constant and pulling thetissue out so that it was accessible Care was taken toremove as much of the surrounding tissues howeverthere is an inherent tradeoff between tissue preparationand time and hence postmortem effects Thereforethe thickness varied as an innate property and as a func-tion of the amount of other tissues attached to the sam-ple of interest Hence it was only feasible to use ap-proximately even samples from run to run This wouldnot be a problem if the sample sizes were very large butthat could not be the case here This brings up anotherissue which is mentioned by Duck (1990) These ex-periments were performed on one sheep and two ratswhich did not allow for an average between animal agesor sizes The effect of the experimental limitations re-quired to minimize the postmortem effects are furtherdiscussed in subsection 426

322 Cutting Conditions Skin and muscle tis-sues were tested in the longitudinal (L) and transversal(T) directions which for the muscle was dened by theorientation of the contractile bers and for the skin bythe Langer lines Cuts were made at two speeds selectedby the physician one designated as fast and the otherslow Generally slow corresponded to less than 15degsec and fast to greater than 15 degsec Moredetails are provided in subsection 45

3 3 Instrumentat ion

The scissors were instrumented with strain gauges(Entran ESU-025-1000) and quarter-turn precision po-tentiometers (Midori CP-2UTX) so both displacementand force could be measured An Analog Devices 3B

Series signal conditioning unit was used with externalbridge balancing Half-bridge congurations were used

for the Mayo and Metzenbaum scissors because onegauge was in compression and the other in extension onthe scissor arm at the point of maximum bending Aquarter-bridge conguration was used for the space-limited Iris scissors Bridge completion was providedwith two dummy resistors (MicroMeasurement moduleMR1-10C-129) The signals from the potentiometers didnot require conditioning Identication of the mechani-cal transfer function of the scissors was conducted todetermine an adequate sampling frequency It was takento be ten times the maximum bandwidth found or 1kHz Twelve-bit analog-to-digital converters were usedA personal computer running QNX handled data acquisi-tion of the force and position data A graphical user in-terface was implemented to provide for simple operationin the eld

331 Calibration The free body diagram for thescissor assuming symmetrical loading is presented ingure 2(a)

The static equilibrium conditions are

O Fx 5 Rx 1 F1 sin u 2 F2 2 Fct cos u 1 Fcb sin u 5 0 (1)

O Fy 5 Ry 1 F1 cos u 2 Fct sin u 2 Fcb cos u 5 0 (2)

O Fz 5 Rz 1 F3 2 Fcz 5 0 (3)

O Moz 5 Mf 1 F1L1 1 F2L1 sin u 1 FcbL2 5 0 (4)

O Moy 5 F3L1 FczL2 5 0 (5)

The force F2 applied by the surgeon to push against thescissors was assumed to be small compared to F1 theforce to close the scissors and hence was neglected F2

may become important in unusual cases such as whenthe object being cut is very hard (cutting bone) or verythin (cutting a sheet) Assuming small angles ( u 30deg)equation (4) reduces to

O Moz lt Mf 1 F1L1 1 FcbL2 5 0 (6)

Neglecting friction Mf in the joint equation (6) reducesto

F1L1 5 FcbL2 (7)

Greenish et al 629

This equation was used to convert the strain gaugereadouts to a cutting force FC measured in Newtonsafter calibration using the setup in gure 2 with a cali-brated load cell (RDP Electrosense type E308)

Equation (7) is re-expressed as

FM OF k FC OC (8)

where k is a constant characteristic of each instrumentThe linear correlation r SSxy SSxyxxSSxyyy was

used as a goodness-of-t measure (Mann 1998) Therrsquos were found to be 100 for the Mayo 100 for theMetzenbaum and 096 for the Iris The forces com-pared in all subsequent discussion are resolved at thecutting point

3 4 Data Segmentat ion

The data were segmented with a velocity zero-crossing method rst by selecting the data points in thevelocity vector smaller than some (typically 005)and then eliminating those that are closer than g datapoints apart (typically g 100) to ignore segments ofsteady position It was found that there are consistentlyfour distinct phases in cutting (i) opening the scissor(ii) resting (iii) closing the scissor and (iv) resting be-fore opening again (See gure 3(a) and 3(b)) Negativeforces indicate scissor opening

4 R esu lt s and Discussion

4 1 Scissor Dif ferences and Inv ariants

In situ measurements prevented us from control-ling many variables To partially compensate for thisseveral controlled runs were performed cutting paperThese runs are presented in gure 4 for the cutting oftwo and four sheets of 20 lb standard printer paperEach graph is composed of more than ten cuts over-lapped

411 Breakthrough Angle In gures 4(a) and4(b) as the Mayo scissor cuts through the paper the FC

increased until close to 5deg after which the forcedropped off This dropoff for the Metzenbaum scissorsgures 4(c) and 4(d) was closer to 8deg whereas thisdropoff did not seem to exist for the Iris scissors untilthe scissors had fully closed and the cut was nished(gures 4(e) and 4(f)) This was related to the geometryof the blades both the Mayo and Metzenbaum haveblunt tips and therefore the scissor blades slide pasteach other before the handles are closed The Iris scis-sors have pointed tips so their blades close when thehandles collide This is shown pictorially in gure 5(b)the Metzenbaum blades just closed at the 8deg openingangle The Mayo scissors in gure 5(a) must close a bitmore before the blades completely close against one

Figure 2 (a) Free body diagram of scissors (b) Setup used to calibrate scissors

630 PRESENCE VOLUME 11 NUMBER 6

another which corresponds to the 5 deg mentionedearlier

Also depicted in gure 5 are the differences in thecontact surface areas of the two blades for the differentscissor types at the two angles 20 deg and 8 degThese differences are seen in ldquoblanksrdquo that corre-sponded to opening and closing the scissors in thin airIt was found that the Mayo scissors had the tightestjoint screw which is seen in the blank runs (gure 6)The tightness of the screw the sharpness of the bladeedges and the exibility of the arms vary considerablyfrom scissors to scissors

412 Scissor Characteristic Slope One mightexpect that the force required to cut four sheets of pa-per as shown in gures 4(b) (d) and (f) would requiretwice as much force as cutting two sheets of paper as ingures 4(a) (c) and (e) As the data show this is notso The ratio was approximately 16 for the Mayo 25for the Metzenbaum and 14 for the Iris

There was however a consistent increase in the FC asthe scissor closes Moreover the slope of the u -FC curveare related to the maximum of the FC For example the

slopes of the curves to cut two sheets of paper for theMayo and Metzenbaum scissors were 048 and 033Ndeg respectively (gures 4(a) and (c)) The corre-sponding maximum FC are 12 N and 10 N The slopesof the curves for cutting four sheets of paper were

056 and 075 Ndeg and the maximum FC are 20N and 25 N for the Mayo and Metzenbaum (gures4(b) and (d)) Therefore the scissors with the largestmaximum FC the Metzenbaum had the steepest slope

413 Reversals It is important to recall how anexperienced surgeon actually uses surgical scissors com-pared to a novice The surgeon rarely closes the scissorscompletely Once the sensation of having completelycut through a tissue is reached the scissors are openedagain This provides greater control A surgeon will nottypically use scissors opened by more than 20deg to makea cut because this results in reduced control while cut-ting and in the possibility of tissue sliding between theblades Instead of making one large cut many small cutsare made The only time the scissor is used wide open iswhen it is used to spread tissues apart as it is used dur-ing tissue preparation

Figure 3 Curves to analyze force versus position relationship (a) Force velocity and position u data for a typical cut with the

Metzenbaum dissecting scissors Phase (1) opening Phase (2) resting before closing Phase (3) closing Phase (4) resting before opening

(b) Force versus theta position data for a typical cut with the Metzenbaum dissecting scissors The vertical data segments are explained

in subsection 423

Greenish et al 631

Although the angle at which the surgeon stopsclosing or opening varies the behavior of the FC be-tween the endpoints was fairly constant The begin-

ning and the end of a cut were similar in that theyboth involved a rapid change in the FC but in oppo-site directions At the end of a close the force

Figure 4 Cutting of printer paper with Mayo (a) two and (b) four sheets with Metzenbaum (c) two and (d) four

sheets and with Iris scissors (e) two and (f) four sheets (student data)

632 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

amp Reidy 2002 Ottensmeyer amp Salisbury 2001Brower et al 2001)

The present study of using haptic recordings ts intoa larger class of research devoted to the use of real-world measurements in haptic rendering The mostcomprehensive system for acquiring data for reality-based modeling thus far is the Active Measurement Fa-cility (ACME) built by Pai Lang Lloyd and Woodham(2000) They have obtained reectance stiffness tex-ture and sound data by probing objects t these datato analytical and empirical models and generated realisticvirtual environments (Richmond amp Pai 2000) MacLean(1996) has created realistic virtual environments usingthe concept of a ldquohaptic camerardquo that captured hapticinformation about real environments Other examplesof empirical haptic models include work on friction byRichard Cutkosky and MacLean (1999) and vibrationsby Okamura Dennerlein and Howe (1998) and Oka-mura Hage Dennerlein and Cutkosky (2000)

3 Materials and Methods

Data were acquired with three instrumented surgi-cal scissors Mayo dissection scissors Metzenbaum dis-section scissors and the Iris scissors The last authorprovided general directions and professional expertise toprepare three separate animal experiments at the ap-proved McGill University animal-care facilities The datacollected are from his hand and that of his students Alldata collection was performed on animals that had beensacriced for other research and in accordance with ani-mal protection laws Because they were planned sacri-ces it was possible to perform the experiments imme-diately after death minimizing postmortem effects

3 1 Instruments

Among the thousands of surgical instrumentsthere exist distinct classes of basic handle designs Amost common class is that of the scissor handles Forscissors many variations exist in the handles and bladesThe blades can have different lengths and breadthsThey can have teeth or not be curved or straight and

pointed or blunt Furthermore the blades can be sharpor dull tight or loose depending on the joint screw andeach of these factors inuences the ldquofeelrdquo of the scissorFor the scissor the central aspect of the haptic experi-ence results from closing them around a tissue regionand hence the scissor angle is the primary state variable

It was learned that surgeons often prefer a curved tipwhich allows more visibility of the work area The in-struments used for experimentation were a 7 in (178cm) Metzenbaum scissor with tungsten carbide insertsa 675 in (171 cm) Mayo scissor with tungsten carbideinserts and a pair of 45 in (114 cm) Iris scissors (allwith curved tips) Piling-Weckrsquos gold-handled scissorswere selected for their smooth and sturdy feel (See g-ure 1)

3 2 Tissues

The tissues compared were from a sheep and tworats Tissues were selected because of their accessibilityand relative abundance but also because of their dis-tinct properties skin (abdominal region) the abdominalmuscle wall the liver and the calcaneal (or Achilles)tendon The three layers of the abdominal muscle wall

Figure 1 The instruments used (from left to right Mayo

Metzenbaum and Iris as in gure 5)

628 PRESENCE VOLUME 11 NUMBER 6

(oblique external internal and transversal abdominal)were cut as one tissue For the sheep experimentsmuscle-rectus fascia was also cut because it was availablein abundance in the fatty abdominal area of the sheep

321 Tissue Preparation For each run a pieceof tissue was prepared and isolated from the surround-ing layers This involved scraping away excess fascia tokeep the thickness relatively constant and pulling thetissue out so that it was accessible Care was taken toremove as much of the surrounding tissues howeverthere is an inherent tradeoff between tissue preparationand time and hence postmortem effects Thereforethe thickness varied as an innate property and as a func-tion of the amount of other tissues attached to the sam-ple of interest Hence it was only feasible to use ap-proximately even samples from run to run This wouldnot be a problem if the sample sizes were very large butthat could not be the case here This brings up anotherissue which is mentioned by Duck (1990) These ex-periments were performed on one sheep and two ratswhich did not allow for an average between animal agesor sizes The effect of the experimental limitations re-quired to minimize the postmortem effects are furtherdiscussed in subsection 426

322 Cutting Conditions Skin and muscle tis-sues were tested in the longitudinal (L) and transversal(T) directions which for the muscle was dened by theorientation of the contractile bers and for the skin bythe Langer lines Cuts were made at two speeds selectedby the physician one designated as fast and the otherslow Generally slow corresponded to less than 15degsec and fast to greater than 15 degsec Moredetails are provided in subsection 45

3 3 Instrumentat ion

The scissors were instrumented with strain gauges(Entran ESU-025-1000) and quarter-turn precision po-tentiometers (Midori CP-2UTX) so both displacementand force could be measured An Analog Devices 3B

Series signal conditioning unit was used with externalbridge balancing Half-bridge congurations were used

for the Mayo and Metzenbaum scissors because onegauge was in compression and the other in extension onthe scissor arm at the point of maximum bending Aquarter-bridge conguration was used for the space-limited Iris scissors Bridge completion was providedwith two dummy resistors (MicroMeasurement moduleMR1-10C-129) The signals from the potentiometers didnot require conditioning Identication of the mechani-cal transfer function of the scissors was conducted todetermine an adequate sampling frequency It was takento be ten times the maximum bandwidth found or 1kHz Twelve-bit analog-to-digital converters were usedA personal computer running QNX handled data acquisi-tion of the force and position data A graphical user in-terface was implemented to provide for simple operationin the eld

331 Calibration The free body diagram for thescissor assuming symmetrical loading is presented ingure 2(a)

The static equilibrium conditions are

O Fx 5 Rx 1 F1 sin u 2 F2 2 Fct cos u 1 Fcb sin u 5 0 (1)

O Fy 5 Ry 1 F1 cos u 2 Fct sin u 2 Fcb cos u 5 0 (2)

O Fz 5 Rz 1 F3 2 Fcz 5 0 (3)

O Moz 5 Mf 1 F1L1 1 F2L1 sin u 1 FcbL2 5 0 (4)

O Moy 5 F3L1 FczL2 5 0 (5)

The force F2 applied by the surgeon to push against thescissors was assumed to be small compared to F1 theforce to close the scissors and hence was neglected F2

may become important in unusual cases such as whenthe object being cut is very hard (cutting bone) or verythin (cutting a sheet) Assuming small angles ( u 30deg)equation (4) reduces to

O Moz lt Mf 1 F1L1 1 FcbL2 5 0 (6)

Neglecting friction Mf in the joint equation (6) reducesto

F1L1 5 FcbL2 (7)

Greenish et al 629

This equation was used to convert the strain gaugereadouts to a cutting force FC measured in Newtonsafter calibration using the setup in gure 2 with a cali-brated load cell (RDP Electrosense type E308)

Equation (7) is re-expressed as

FM OF k FC OC (8)

where k is a constant characteristic of each instrumentThe linear correlation r SSxy SSxyxxSSxyyy was

used as a goodness-of-t measure (Mann 1998) Therrsquos were found to be 100 for the Mayo 100 for theMetzenbaum and 096 for the Iris The forces com-pared in all subsequent discussion are resolved at thecutting point

3 4 Data Segmentat ion

The data were segmented with a velocity zero-crossing method rst by selecting the data points in thevelocity vector smaller than some (typically 005)and then eliminating those that are closer than g datapoints apart (typically g 100) to ignore segments ofsteady position It was found that there are consistentlyfour distinct phases in cutting (i) opening the scissor(ii) resting (iii) closing the scissor and (iv) resting be-fore opening again (See gure 3(a) and 3(b)) Negativeforces indicate scissor opening

4 R esu lt s and Discussion

4 1 Scissor Dif ferences and Inv ariants

In situ measurements prevented us from control-ling many variables To partially compensate for thisseveral controlled runs were performed cutting paperThese runs are presented in gure 4 for the cutting oftwo and four sheets of 20 lb standard printer paperEach graph is composed of more than ten cuts over-lapped

411 Breakthrough Angle In gures 4(a) and4(b) as the Mayo scissor cuts through the paper the FC

increased until close to 5deg after which the forcedropped off This dropoff for the Metzenbaum scissorsgures 4(c) and 4(d) was closer to 8deg whereas thisdropoff did not seem to exist for the Iris scissors untilthe scissors had fully closed and the cut was nished(gures 4(e) and 4(f)) This was related to the geometryof the blades both the Mayo and Metzenbaum haveblunt tips and therefore the scissor blades slide pasteach other before the handles are closed The Iris scis-sors have pointed tips so their blades close when thehandles collide This is shown pictorially in gure 5(b)the Metzenbaum blades just closed at the 8deg openingangle The Mayo scissors in gure 5(a) must close a bitmore before the blades completely close against one

Figure 2 (a) Free body diagram of scissors (b) Setup used to calibrate scissors

630 PRESENCE VOLUME 11 NUMBER 6

another which corresponds to the 5 deg mentionedearlier

Also depicted in gure 5 are the differences in thecontact surface areas of the two blades for the differentscissor types at the two angles 20 deg and 8 degThese differences are seen in ldquoblanksrdquo that corre-sponded to opening and closing the scissors in thin airIt was found that the Mayo scissors had the tightestjoint screw which is seen in the blank runs (gure 6)The tightness of the screw the sharpness of the bladeedges and the exibility of the arms vary considerablyfrom scissors to scissors

412 Scissor Characteristic Slope One mightexpect that the force required to cut four sheets of pa-per as shown in gures 4(b) (d) and (f) would requiretwice as much force as cutting two sheets of paper as ingures 4(a) (c) and (e) As the data show this is notso The ratio was approximately 16 for the Mayo 25for the Metzenbaum and 14 for the Iris

There was however a consistent increase in the FC asthe scissor closes Moreover the slope of the u -FC curveare related to the maximum of the FC For example the

slopes of the curves to cut two sheets of paper for theMayo and Metzenbaum scissors were 048 and 033Ndeg respectively (gures 4(a) and (c)) The corre-sponding maximum FC are 12 N and 10 N The slopesof the curves for cutting four sheets of paper were

056 and 075 Ndeg and the maximum FC are 20N and 25 N for the Mayo and Metzenbaum (gures4(b) and (d)) Therefore the scissors with the largestmaximum FC the Metzenbaum had the steepest slope

413 Reversals It is important to recall how anexperienced surgeon actually uses surgical scissors com-pared to a novice The surgeon rarely closes the scissorscompletely Once the sensation of having completelycut through a tissue is reached the scissors are openedagain This provides greater control A surgeon will nottypically use scissors opened by more than 20deg to makea cut because this results in reduced control while cut-ting and in the possibility of tissue sliding between theblades Instead of making one large cut many small cutsare made The only time the scissor is used wide open iswhen it is used to spread tissues apart as it is used dur-ing tissue preparation

Figure 3 Curves to analyze force versus position relationship (a) Force velocity and position u data for a typical cut with the

Metzenbaum dissecting scissors Phase (1) opening Phase (2) resting before closing Phase (3) closing Phase (4) resting before opening

(b) Force versus theta position data for a typical cut with the Metzenbaum dissecting scissors The vertical data segments are explained

in subsection 423

Greenish et al 631

Although the angle at which the surgeon stopsclosing or opening varies the behavior of the FC be-tween the endpoints was fairly constant The begin-

ning and the end of a cut were similar in that theyboth involved a rapid change in the FC but in oppo-site directions At the end of a close the force

Figure 4 Cutting of printer paper with Mayo (a) two and (b) four sheets with Metzenbaum (c) two and (d) four

sheets and with Iris scissors (e) two and (f) four sheets (student data)

632 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

(oblique external internal and transversal abdominal)were cut as one tissue For the sheep experimentsmuscle-rectus fascia was also cut because it was availablein abundance in the fatty abdominal area of the sheep

321 Tissue Preparation For each run a pieceof tissue was prepared and isolated from the surround-ing layers This involved scraping away excess fascia tokeep the thickness relatively constant and pulling thetissue out so that it was accessible Care was taken toremove as much of the surrounding tissues howeverthere is an inherent tradeoff between tissue preparationand time and hence postmortem effects Thereforethe thickness varied as an innate property and as a func-tion of the amount of other tissues attached to the sam-ple of interest Hence it was only feasible to use ap-proximately even samples from run to run This wouldnot be a problem if the sample sizes were very large butthat could not be the case here This brings up anotherissue which is mentioned by Duck (1990) These ex-periments were performed on one sheep and two ratswhich did not allow for an average between animal agesor sizes The effect of the experimental limitations re-quired to minimize the postmortem effects are furtherdiscussed in subsection 426

322 Cutting Conditions Skin and muscle tis-sues were tested in the longitudinal (L) and transversal(T) directions which for the muscle was dened by theorientation of the contractile bers and for the skin bythe Langer lines Cuts were made at two speeds selectedby the physician one designated as fast and the otherslow Generally slow corresponded to less than 15degsec and fast to greater than 15 degsec Moredetails are provided in subsection 45

3 3 Instrumentat ion

The scissors were instrumented with strain gauges(Entran ESU-025-1000) and quarter-turn precision po-tentiometers (Midori CP-2UTX) so both displacementand force could be measured An Analog Devices 3B

Series signal conditioning unit was used with externalbridge balancing Half-bridge congurations were used

for the Mayo and Metzenbaum scissors because onegauge was in compression and the other in extension onthe scissor arm at the point of maximum bending Aquarter-bridge conguration was used for the space-limited Iris scissors Bridge completion was providedwith two dummy resistors (MicroMeasurement moduleMR1-10C-129) The signals from the potentiometers didnot require conditioning Identication of the mechani-cal transfer function of the scissors was conducted todetermine an adequate sampling frequency It was takento be ten times the maximum bandwidth found or 1kHz Twelve-bit analog-to-digital converters were usedA personal computer running QNX handled data acquisi-tion of the force and position data A graphical user in-terface was implemented to provide for simple operationin the eld

331 Calibration The free body diagram for thescissor assuming symmetrical loading is presented ingure 2(a)

The static equilibrium conditions are

O Fx 5 Rx 1 F1 sin u 2 F2 2 Fct cos u 1 Fcb sin u 5 0 (1)

O Fy 5 Ry 1 F1 cos u 2 Fct sin u 2 Fcb cos u 5 0 (2)

O Fz 5 Rz 1 F3 2 Fcz 5 0 (3)

O Moz 5 Mf 1 F1L1 1 F2L1 sin u 1 FcbL2 5 0 (4)

O Moy 5 F3L1 FczL2 5 0 (5)

The force F2 applied by the surgeon to push against thescissors was assumed to be small compared to F1 theforce to close the scissors and hence was neglected F2

may become important in unusual cases such as whenthe object being cut is very hard (cutting bone) or verythin (cutting a sheet) Assuming small angles ( u 30deg)equation (4) reduces to

O Moz lt Mf 1 F1L1 1 FcbL2 5 0 (6)

Neglecting friction Mf in the joint equation (6) reducesto

F1L1 5 FcbL2 (7)

Greenish et al 629

This equation was used to convert the strain gaugereadouts to a cutting force FC measured in Newtonsafter calibration using the setup in gure 2 with a cali-brated load cell (RDP Electrosense type E308)

Equation (7) is re-expressed as

FM OF k FC OC (8)

where k is a constant characteristic of each instrumentThe linear correlation r SSxy SSxyxxSSxyyy was

used as a goodness-of-t measure (Mann 1998) Therrsquos were found to be 100 for the Mayo 100 for theMetzenbaum and 096 for the Iris The forces com-pared in all subsequent discussion are resolved at thecutting point

3 4 Data Segmentat ion

The data were segmented with a velocity zero-crossing method rst by selecting the data points in thevelocity vector smaller than some (typically 005)and then eliminating those that are closer than g datapoints apart (typically g 100) to ignore segments ofsteady position It was found that there are consistentlyfour distinct phases in cutting (i) opening the scissor(ii) resting (iii) closing the scissor and (iv) resting be-fore opening again (See gure 3(a) and 3(b)) Negativeforces indicate scissor opening

4 R esu lt s and Discussion

4 1 Scissor Dif ferences and Inv ariants

In situ measurements prevented us from control-ling many variables To partially compensate for thisseveral controlled runs were performed cutting paperThese runs are presented in gure 4 for the cutting oftwo and four sheets of 20 lb standard printer paperEach graph is composed of more than ten cuts over-lapped

411 Breakthrough Angle In gures 4(a) and4(b) as the Mayo scissor cuts through the paper the FC

increased until close to 5deg after which the forcedropped off This dropoff for the Metzenbaum scissorsgures 4(c) and 4(d) was closer to 8deg whereas thisdropoff did not seem to exist for the Iris scissors untilthe scissors had fully closed and the cut was nished(gures 4(e) and 4(f)) This was related to the geometryof the blades both the Mayo and Metzenbaum haveblunt tips and therefore the scissor blades slide pasteach other before the handles are closed The Iris scis-sors have pointed tips so their blades close when thehandles collide This is shown pictorially in gure 5(b)the Metzenbaum blades just closed at the 8deg openingangle The Mayo scissors in gure 5(a) must close a bitmore before the blades completely close against one

Figure 2 (a) Free body diagram of scissors (b) Setup used to calibrate scissors

630 PRESENCE VOLUME 11 NUMBER 6

another which corresponds to the 5 deg mentionedearlier

Also depicted in gure 5 are the differences in thecontact surface areas of the two blades for the differentscissor types at the two angles 20 deg and 8 degThese differences are seen in ldquoblanksrdquo that corre-sponded to opening and closing the scissors in thin airIt was found that the Mayo scissors had the tightestjoint screw which is seen in the blank runs (gure 6)The tightness of the screw the sharpness of the bladeedges and the exibility of the arms vary considerablyfrom scissors to scissors

412 Scissor Characteristic Slope One mightexpect that the force required to cut four sheets of pa-per as shown in gures 4(b) (d) and (f) would requiretwice as much force as cutting two sheets of paper as ingures 4(a) (c) and (e) As the data show this is notso The ratio was approximately 16 for the Mayo 25for the Metzenbaum and 14 for the Iris

There was however a consistent increase in the FC asthe scissor closes Moreover the slope of the u -FC curveare related to the maximum of the FC For example the

slopes of the curves to cut two sheets of paper for theMayo and Metzenbaum scissors were 048 and 033Ndeg respectively (gures 4(a) and (c)) The corre-sponding maximum FC are 12 N and 10 N The slopesof the curves for cutting four sheets of paper were

056 and 075 Ndeg and the maximum FC are 20N and 25 N for the Mayo and Metzenbaum (gures4(b) and (d)) Therefore the scissors with the largestmaximum FC the Metzenbaum had the steepest slope

413 Reversals It is important to recall how anexperienced surgeon actually uses surgical scissors com-pared to a novice The surgeon rarely closes the scissorscompletely Once the sensation of having completelycut through a tissue is reached the scissors are openedagain This provides greater control A surgeon will nottypically use scissors opened by more than 20deg to makea cut because this results in reduced control while cut-ting and in the possibility of tissue sliding between theblades Instead of making one large cut many small cutsare made The only time the scissor is used wide open iswhen it is used to spread tissues apart as it is used dur-ing tissue preparation

Figure 3 Curves to analyze force versus position relationship (a) Force velocity and position u data for a typical cut with the

Metzenbaum dissecting scissors Phase (1) opening Phase (2) resting before closing Phase (3) closing Phase (4) resting before opening

(b) Force versus theta position data for a typical cut with the Metzenbaum dissecting scissors The vertical data segments are explained

in subsection 423

Greenish et al 631

Although the angle at which the surgeon stopsclosing or opening varies the behavior of the FC be-tween the endpoints was fairly constant The begin-

ning and the end of a cut were similar in that theyboth involved a rapid change in the FC but in oppo-site directions At the end of a close the force

Figure 4 Cutting of printer paper with Mayo (a) two and (b) four sheets with Metzenbaum (c) two and (d) four

sheets and with Iris scissors (e) two and (f) four sheets (student data)

632 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

This equation was used to convert the strain gaugereadouts to a cutting force FC measured in Newtonsafter calibration using the setup in gure 2 with a cali-brated load cell (RDP Electrosense type E308)

Equation (7) is re-expressed as

FM OF k FC OC (8)

where k is a constant characteristic of each instrumentThe linear correlation r SSxy SSxyxxSSxyyy was

used as a goodness-of-t measure (Mann 1998) Therrsquos were found to be 100 for the Mayo 100 for theMetzenbaum and 096 for the Iris The forces com-pared in all subsequent discussion are resolved at thecutting point

3 4 Data Segmentat ion

The data were segmented with a velocity zero-crossing method rst by selecting the data points in thevelocity vector smaller than some (typically 005)and then eliminating those that are closer than g datapoints apart (typically g 100) to ignore segments ofsteady position It was found that there are consistentlyfour distinct phases in cutting (i) opening the scissor(ii) resting (iii) closing the scissor and (iv) resting be-fore opening again (See gure 3(a) and 3(b)) Negativeforces indicate scissor opening

4 R esu lt s and Discussion

4 1 Scissor Dif ferences and Inv ariants

In situ measurements prevented us from control-ling many variables To partially compensate for thisseveral controlled runs were performed cutting paperThese runs are presented in gure 4 for the cutting oftwo and four sheets of 20 lb standard printer paperEach graph is composed of more than ten cuts over-lapped

411 Breakthrough Angle In gures 4(a) and4(b) as the Mayo scissor cuts through the paper the FC

increased until close to 5deg after which the forcedropped off This dropoff for the Metzenbaum scissorsgures 4(c) and 4(d) was closer to 8deg whereas thisdropoff did not seem to exist for the Iris scissors untilthe scissors had fully closed and the cut was nished(gures 4(e) and 4(f)) This was related to the geometryof the blades both the Mayo and Metzenbaum haveblunt tips and therefore the scissor blades slide pasteach other before the handles are closed The Iris scis-sors have pointed tips so their blades close when thehandles collide This is shown pictorially in gure 5(b)the Metzenbaum blades just closed at the 8deg openingangle The Mayo scissors in gure 5(a) must close a bitmore before the blades completely close against one

Figure 2 (a) Free body diagram of scissors (b) Setup used to calibrate scissors

630 PRESENCE VOLUME 11 NUMBER 6

another which corresponds to the 5 deg mentionedearlier

Also depicted in gure 5 are the differences in thecontact surface areas of the two blades for the differentscissor types at the two angles 20 deg and 8 degThese differences are seen in ldquoblanksrdquo that corre-sponded to opening and closing the scissors in thin airIt was found that the Mayo scissors had the tightestjoint screw which is seen in the blank runs (gure 6)The tightness of the screw the sharpness of the bladeedges and the exibility of the arms vary considerablyfrom scissors to scissors

412 Scissor Characteristic Slope One mightexpect that the force required to cut four sheets of pa-per as shown in gures 4(b) (d) and (f) would requiretwice as much force as cutting two sheets of paper as ingures 4(a) (c) and (e) As the data show this is notso The ratio was approximately 16 for the Mayo 25for the Metzenbaum and 14 for the Iris

There was however a consistent increase in the FC asthe scissor closes Moreover the slope of the u -FC curveare related to the maximum of the FC For example the

slopes of the curves to cut two sheets of paper for theMayo and Metzenbaum scissors were 048 and 033Ndeg respectively (gures 4(a) and (c)) The corre-sponding maximum FC are 12 N and 10 N The slopesof the curves for cutting four sheets of paper were

056 and 075 Ndeg and the maximum FC are 20N and 25 N for the Mayo and Metzenbaum (gures4(b) and (d)) Therefore the scissors with the largestmaximum FC the Metzenbaum had the steepest slope

413 Reversals It is important to recall how anexperienced surgeon actually uses surgical scissors com-pared to a novice The surgeon rarely closes the scissorscompletely Once the sensation of having completelycut through a tissue is reached the scissors are openedagain This provides greater control A surgeon will nottypically use scissors opened by more than 20deg to makea cut because this results in reduced control while cut-ting and in the possibility of tissue sliding between theblades Instead of making one large cut many small cutsare made The only time the scissor is used wide open iswhen it is used to spread tissues apart as it is used dur-ing tissue preparation

Figure 3 Curves to analyze force versus position relationship (a) Force velocity and position u data for a typical cut with the

Metzenbaum dissecting scissors Phase (1) opening Phase (2) resting before closing Phase (3) closing Phase (4) resting before opening

(b) Force versus theta position data for a typical cut with the Metzenbaum dissecting scissors The vertical data segments are explained

in subsection 423

Greenish et al 631

Although the angle at which the surgeon stopsclosing or opening varies the behavior of the FC be-tween the endpoints was fairly constant The begin-

ning and the end of a cut were similar in that theyboth involved a rapid change in the FC but in oppo-site directions At the end of a close the force

Figure 4 Cutting of printer paper with Mayo (a) two and (b) four sheets with Metzenbaum (c) two and (d) four

sheets and with Iris scissors (e) two and (f) four sheets (student data)

632 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

another which corresponds to the 5 deg mentionedearlier

Also depicted in gure 5 are the differences in thecontact surface areas of the two blades for the differentscissor types at the two angles 20 deg and 8 degThese differences are seen in ldquoblanksrdquo that corre-sponded to opening and closing the scissors in thin airIt was found that the Mayo scissors had the tightestjoint screw which is seen in the blank runs (gure 6)The tightness of the screw the sharpness of the bladeedges and the exibility of the arms vary considerablyfrom scissors to scissors

412 Scissor Characteristic Slope One mightexpect that the force required to cut four sheets of pa-per as shown in gures 4(b) (d) and (f) would requiretwice as much force as cutting two sheets of paper as ingures 4(a) (c) and (e) As the data show this is notso The ratio was approximately 16 for the Mayo 25for the Metzenbaum and 14 for the Iris

There was however a consistent increase in the FC asthe scissor closes Moreover the slope of the u -FC curveare related to the maximum of the FC For example the

slopes of the curves to cut two sheets of paper for theMayo and Metzenbaum scissors were 048 and 033Ndeg respectively (gures 4(a) and (c)) The corre-sponding maximum FC are 12 N and 10 N The slopesof the curves for cutting four sheets of paper were

056 and 075 Ndeg and the maximum FC are 20N and 25 N for the Mayo and Metzenbaum (gures4(b) and (d)) Therefore the scissors with the largestmaximum FC the Metzenbaum had the steepest slope

413 Reversals It is important to recall how anexperienced surgeon actually uses surgical scissors com-pared to a novice The surgeon rarely closes the scissorscompletely Once the sensation of having completelycut through a tissue is reached the scissors are openedagain This provides greater control A surgeon will nottypically use scissors opened by more than 20deg to makea cut because this results in reduced control while cut-ting and in the possibility of tissue sliding between theblades Instead of making one large cut many small cutsare made The only time the scissor is used wide open iswhen it is used to spread tissues apart as it is used dur-ing tissue preparation

Figure 3 Curves to analyze force versus position relationship (a) Force velocity and position u data for a typical cut with the

Metzenbaum dissecting scissors Phase (1) opening Phase (2) resting before closing Phase (3) closing Phase (4) resting before opening

(b) Force versus theta position data for a typical cut with the Metzenbaum dissecting scissors The vertical data segments are explained

in subsection 423

Greenish et al 631

Although the angle at which the surgeon stopsclosing or opening varies the behavior of the FC be-tween the endpoints was fairly constant The begin-

ning and the end of a cut were similar in that theyboth involved a rapid change in the FC but in oppo-site directions At the end of a close the force

Figure 4 Cutting of printer paper with Mayo (a) two and (b) four sheets with Metzenbaum (c) two and (d) four

sheets and with Iris scissors (e) two and (f) four sheets (student data)

632 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

Although the angle at which the surgeon stopsclosing or opening varies the behavior of the FC be-tween the endpoints was fairly constant The begin-

ning and the end of a cut were similar in that theyboth involved a rapid change in the FC but in oppo-site directions At the end of a close the force

Figure 4 Cutting of printer paper with Mayo (a) two and (b) four sheets with Metzenbaum (c) two and (d) four

sheets and with Iris scissors (e) two and (f) four sheets (student data)

632 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

dropped off quickly and at the beginning of a cut theforce rose quickly

4 2 Tissue Dif ferences

The tissues compared were from a sheep and tworats Calling FC( u ) the cutting force recorded as a func-tion of u FC( u ) is the average force over several runsThe maximum of the average cutting force over all u rsquosmFC is then found and presented rst to establish gen-eral trends See table 1 The missing values are a resultof experimental limitations

421 Maximum Average Force Looking rstat the Mayo and Metzenbaum sections the sheep tis-sues generally required more force to cut than did therat tissues This agreed with our expectation The ten-don was the only tissue that did not change dramaticallyin the mFC required between sheep and rat This wasexplained by the fact that because the sheep Achillestendon was massive it was necessary to split it and cutonly the number of strands that could reasonably be cutby the scissors This resulted in a bundle close to thesize of the rat Achilles tendon

422 Comparing Blanks and Liver Cuts Onemight expect these runs had the lowest mFC This is truefor the Iris scissors but not for the Mayo and Metzen-

baum scissors the liver cut actually required less forcethan did the blank For example the average mFC forthe Mayo blank was close to 70 N whereas for the ratliver it is 5861 N (fastslow) Similarly for theMetzenbaum the blank had a mFC of approximately 33N but the rat liver registered 1625 N On the otherhand for the Iris scissors the rat liver required muchmore force to cut approximately 102 to 107 N versus5 N for the blank The results for the Iris scissors arediscussed rstFC of the Iris scissors All tissue cuts performed withthe Iris scissors required more force than did the Mayoand Metzenbaum scissors The shorter arm length hadno bearing on the measurement but the shorter bladelength did The Iris blade length was half that of theMetzenbaum blade The forces were typically two tothree times higher for the Iris than for the Metzenbaumscissors therefore factors must exist to account forthese differences

The Mayo and Metzenbaum scissor blades had tung-sten carbide inserts that made their blade edges sharperand stronger than the Iris blades Because the bladesthemselves are thicker for the Mayo and Metzenbaumscissors they were less likely to bend laterally and dis-turb the cutting process This does not explain how-ever why the FC for the Mayo and Metzenbaum scis-sors are lower than their respective blank cuts duringthe liver runs

Figure 5 Different overlap areas for different scissors blade types at same angles u

Greenish et al 633

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

FC of the Mayo and Metzenbaum scissors To explainwhy the Mayo and Metzenbaum liver runs registered alower force than did the blank cuts it was thought that

blood was lubricating the scissors However this effectwas not reproducible offsite spraying the scissors withlubricant (WD-40) gave virtually the same FC( u ) curves

Figure 6 Comparison of blank runs for Mayo scissors between (a) surgeon and (b) student Metzenbaum scissors

for (c) surgeon and (d) student and Iris scissors (e) surgeon and (f) student

634 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

which is consistent with the nature of friction Lubrica-tion in metal contacts is explained by hydrodynamiceffects that play no role here given the low velocitiesand the small contact areas involved

Upon closer inspection it was noticed that the blankruns performed by the student were different from theblank runs performed by the surgeon (gure 6) Thesurgeon registered a mFC of 8 N (table 1) which isshown graphically in gure 6(a) the student registerednearly half of that 41 N (gure 6(b)) Similar resultswere obtained for the Metzenbaum scissors and for theIris scissors 4 N versus 29 N (gure 6(c) and (d)) 72N versus 47 N (gure 6(e) and (f)) Therefore thegrip as well as the tissue and scissor types affected theforce experienced The grip used was probably a strongfunction of training and reveals that it is important toinvolve trained surgeons at these early stages of re-search

423 Grip Change Effects The grip affectedthe FC as there was an increase or a decrease in the forcewhen the scissor angle u was not changing In gure4(d) at 11 deg on the opening (lower) curve there is amomentary change in the cutting direction This causeda rapid increase in the force as a result of a brief closingof u to 10 deg (circled in gure) After this brief clos-ing the FC again went down to the force level of therest of the open curve as the scissors were completelyopened Clearly the short change in direction resultedin a momentary jump up to the corresponding blankclosing curve (gure 6(d)) for the student blank Met-zenbaum run These rapid rises in FC in the openingcurve are also visible in data where there is no closing ofthe scissor but only a pause (gure 4(a) at 11 deg cir-cled in gure) due to static friction

This indicates that there were at least two grips anopening grip and a closing grip interacting with differ-

Table 1 mFC the Maximum of the Average Applied Force at C for Different Tissue Types During Cuts by the Surgeon

Scissor

Mayo Metz Iris

Closing Opening Closing Opening Closing Opening

Blank (slow) 66 80 31 40 51 47Blank (fast) 64 73 36 26 30 72

Tissue 2 Animal 3 Rat Sheep Rat Sheep Rat Sheep

Skin (L slow) 80 176 70 175 226 448Skin (L fast) 87 205 59 169 234 326Skin (T slow) 84 164 mdash 132 173 mdashSkin (T fast) 78 164 81 193 182 mdashMuscle (slow) 102 mdash mdash mdash 166 mdashMuscle (fast) mdash mdash 24 mdash mdash mdashMuscle (L fast) mdash 58 mdash mdash mdash 140Muscle (T fast) mdash 91 mdash mdash 228 270Liver (slow) 61 mdash 25 mdash 102 mdashLiver (fast) 58 83 16 71 107 212Fascia (slow) mdash 78 mdash mdash mdash 125Fascia (fast) mdash 87 mdash 60 mdash 177Tendon 228 246 305 mdash mdash mdash

Greenish et al 635

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

ent regions of the rings Because an increase in FC oc-curred even when there was no closing action there alsoseemed to be a grip change when stopping This againimplies that the grip had a strong inuence on theforces read As a note it is also likely that a grip changecaused the abrupt increase in FC once the scissors hadbeen completely opened typically near 30 deg

This grip effect may also explain the high variabilityexperienced with the Iris scissors (gure 6(e)) whereinfour cuts performed by the surgeon are plotted to-gether The high degree of variability was not seen inthe student runs however (gure 6(f) wherein morethan ten cuts are plotted together

424 Contact Between Object and ScissorsAnother observation can be made from gures 4(a)through (d) For all incident angles the force leveljumped to the same closing curve The inner-most curveclosed at around 15 deg where the force level jumpedfrom approximately 2 N to 1 N When the scissorsstarted closing the force remained on the blank closingcurve then near 12 deg contact occurred with thetissue and force rose almost instantly to 20 N to meetthe closing curve

425 Longitudinal versus Transversal Noclear relationship emerged from comparing the valuesfor the longitudinal (along the body axis) and transver-sal cutting directions (table 1) Statistical analysis wastherefore performed to determine whether there weresignicance in the data from sum of squares error analy-sis (SSE) (Mann 1998) A signicant difference betweencurves was taken as a difference in the SSE between thetwo directions SSEL T which was greater than the vari-ability between the runs in any one direction SSEL orSSET The average force curves in each direction wererst calculated

FDave u 51N O

i 1

N

FDi( u ) (9)

where N is the number of runs and D is L or T Thenvariability in the two directions is calculated as

SSED 5 Oi 1

ND

(FDave u 2 FDi u )2 (10)

The sum square of the error could then be calculated as

SSEL T Oj min( u )

max( u )

(FLave( j ) FTave( j ))2 (11)

The results from this analysis are shown in table 2For the Mayo and Metzenbaum scissors the differ-

ences between transversal and longitudinal cuts for boththe rat and sheep skin were not signicantly differentHowever for the Iris scissors the rat skin cut slowlyregistered as signicantly different for the transversaland longitudinal directions This can be due to theknown variability of these scissors or to their higher sen-sitivity Unfortunately the fast skin cuts do not verifythis result From table 1 the mFC for the longitudinaldirection is 23 N for the Iris scissors and 178 N for thetransversal skin direction Because skin bers are alignedwith the transversal axis of the body the longitudinalforce should be higher The more delicate Iris scissorsmay indeed be sensitive to the cutting directions of skin

The two cutting directions were signicantly differentfor the muscle cuts with both the Mayo and Iris scissorsBecause muscle bers are highly oriented strands this isexpected These results are indeterminate for the Met-zenbaum scissors because there were no data availablefor the slow muscle cuts For the simulation of skin cut-ting with these scissors it seems reasonable that thesame model could be used for both cutting directionsHowever a muscle cutting simulation should accountfor orientation

426 Tissue VariabilityTissue Toughness The differences between tissues ofthe same type are due to the fact that the properties ofthe layers changed from region to region in the animalFor example the skin near the outside of the stomacharea was tougher (and thicker) than the skin closer tothe center During experimentation care was taken touse tissue from only a small region However therewere many different experimental cutting conditionswhich required a large quantity of uniform tissue

636 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

Tissue Thickness and Preloading To understand theeffect of thickness and tissue strength on the cuttingforce one must recall how tissue is cut The appliedforce is a result of interaction with the tissue at andahead of the cutting point With closing the amount oftissue deformed between the blades increases This re-sults in a gradual increase in the cutting force

Although exact thickness control is not critical in thiswork where general trends are sought it is an importantfactor to keep in mind for future experiments A prob-lem encountered was the folding over of the preparedtissues The tissues had to be exposed so they were heldwith forceps clamped near the corners Initially the as-sisting surgeon was holding the tissues just below thecorners and therefore the skin had no support in themiddle This resulted in the edge folding over on itselfresulting in a much larger FC because the tissues weretwice as thick This effect can be seen in gure 7 wheretwo runs were performed under the same conditionslongitudinal and fast but give very different results TheFC measured when the skin was folded over was nearlytwice as much approximately 12 N versus 6 N Againfor reasons of time limitation the possible effect of pre-load applied to the tissue could not be controlled asmuch as it should have been

Unfortunately this problem was not identied untilpart way through experimentation Although most ofthe tests were repeated similar tissue and time restric-tions prevented all experiments from being repeatedThis accounts for some of the values absent in table 1The other missing data values include the tendon cutsfor the Iris scissors These experiments were not per-formed for fear of ruining the delicate Iris scissors withthe tough tendon As well no liver experiments wererun in slow speed for the sheepEffect of Sticking Another variable was ldquostickinessrdquo(as distinguished from friction) as a result of the blooddrying on the blades Because the animals were drainedbefore experimentation there was less blood in the tis-sues than when the animal is alive Stickiness betweenthe blades probably appears in the data as an increase inFC Spraying with saline solution was performed whendeemed necessary but it is possible that certain datales would require a slight correction This error is notconsidered signicant for the determination of generaltrendsSummary Overall an accurate description of the tis-sues and experimental conditions is required for an ex-act force measure of one tissue type It is not sufcientto simply specify ldquoskinrdquo or ldquofasciardquo The thickness of

Table 2 Comparing Transversal and Longitudinal Directions for Cuts (Surgeon Data)

Scissor TissueMax(SSET SSEL) SSEL T

Statisticallydifferent

Mayo Skin (sheep slow) 31 15 NSkin (sheep fast) 61 14 NSkin (rat slow) 13 09 NSkin (rat fast) 23 11 NMuscle (sheep) 06 24 Y

Metz Skin (sheep slow) 44 29 NSkin (sheep fast) 61 14 NSkin (rat fast) 35 22 N

Iris Skin (rat slow) 34 86 YSkin (rat fast) 46 41 NMuscle (sheep) 45 109 Y

Greenish et al 637

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

the tissue and moisture content play an equally impor-tant role in the resistance of the tissue and thereforeshould be measured As well future experiments shouldinclude more exacting tissue preparation and larger sam-ple sizes

4 3 Frequency Analysis

The power spectral density (PSD) of the force sig-nal was obtained for the four task phases described insubsection 34 The signals have little high-frequencycomponents for the four phases for each of the threescissors The magnitude of the PSD response (in dB) wasthe highest for phase 3 or closing the scissors For all ofthe tissue types (skin liver tendon fascia) most of thefrequency components are below 4ndash5 Hz with a peaktypically between 10ndash39 Hz Figures comparing thepower spectral density curves for the three scissors havebeen included for the cutting of sheep skin slowly (Seegure 8) It is likely that the 10ndash39 Hz componentswere a property of the tissue being cut From a model-ing point of view this suggests that only low-frequencysignals need to be reproduced for realism Furthermorethere must be differences other than the range of thefrequency components to explain why the cuts for dif-ferent tissues feel different

4 4 Comparat iv e A nalysis

It is useful to compare some tissue-cutting curvesas in gure 9 along with table 1 Looking rst at theblank and liver cuts the closing curves for these runs arerelatively smooth especially compared to the cuts forthe muscle and skin Comparing the muscle run to theblank the differences are seen in the magnitude of theforce during the closing phase as well as in the bumpi-ness or texture of the muscle cut The maximum of theaverage force curve mFC for the blank cuts was around 6N versus 9 N for the muscle Whereas the skin cutsshowed a similar behavior the force magnitude differ-ence was much less pronounced when compared to theblank run (reaching as low as 6 N)

Given the small change in mFC for skin and liver runsas compared to blank runs it was also possible to inferthat for these tissues cut with Mayo scissors the tactilesensation was due to irregularities This was opposite forthe tendon however for which all of the tactile feed-back seemed to be due to the high force magnitudewithout any texture Looking at gure 9(c) it is clearthat the tendon had a more elastic force prole and thestress increases until breakthrough occurs with proba-bly little cutting before that point After the break-through the force drops off rapidly to zero

Figure 7 The effect of having the tissue layer folded over Both curves are for the fast longitudinal cutting of rat skin with

the Metzenbaum scissors where the skin is (a) normal and (b) folded

638 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

441 Textural Components Because some ofthe tissues expressed little or no magnitude force differ-ence from the blank the difference felt for these tissuesmay strictly result from their nonuniformity Based onthe power spectral analysis textural components werelow frequency and hence should perhaps more appropri-ately be expressed in deg 1 rather than Hertz In otherwords the question whether the synthesis of texturalcomponents should be done temporally rather than spa-tially is unresolved

To see whether these components could explain thedifferences in feel for other samples the results from the

sheep experiments were investigated for the same tis-sues (See gure 10)

One difference that stands out is that for the sheepthe muscle cut seemed smoother than the liver cutwhereas this relationship was opposite for the rat tis-sues This could be a result of the fact that the sheepliver is much larger and actually engulfs the scissors inthe tissue as they cut through it enhancing texturalcomponents The sheep skin was similar to the ratskin however it naturally had a considerably largerforce magnitude component (18 N versus 8 N) Thesheep tendon like the rat tendon displayed a behav-

Figure 8 Comparison of power spectral density curves for the cutting of sheep skin slowly with (surgeon data) (a) Mayo

scissors (b) Metzenbaum scissors and (c) Iris scissors

Greenish et al 639

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

ior in which the stress increases to a large value untilbreakthrough Some of the tissues like the skin andtendon had a similar texture between the rat and

sheep for the Mayo scissors We now compare theMetzenbaum cutting of rat tissues by looking at g-ure 11

Figure 9 Comparison of curves for different rat tissue cuts made with the Mayo scissors (surgeon data) (a) blank

(b) muscle (c) tendon (d) skin and (e) liver

640 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

The same trends were visible for the Metzenbaumcutting of rat tissues as for the Mayo scissors Theblank for the Metzenbaum scissors was relatively

smooth as is the liver cut The tendon cut also in-volved a steep rise in the force until breakthroughdisplaying the same elastic behavior Furthermore

Figure 10 Comparison of curves for different sheep tissue cuts made with the Mayo scissors (surgeon data) (a)

blank (b) muscle (c) tendon (d) skin and (e) liver

Greenish et al 641

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

the skin and muscle cuts with the Metzenbaum con-tained textures similar to the ones obtained with theMayo scissors

Although all of these sheep tissue trials were availablefor the Metzenbaum scissors some comparisons are stillpossible For the cutting of sheep skin with the Metzen-

Figure 11 Comparison of curves for different rat tissue cuts made with the Metzenbaum scissors (surgeon data)

(a) blank (b) muscle (c) tendon (d) skin and (e) liver

642 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

baum scissors a behavior similar to that of sheep skincut with the Mayo scissors could be seen The fast cut-ting of fascia with the Metzenbaum and the sheep livercut closely resemble the Mayo cuts As with the Mayothe magnitudes for cutting the sheep tissues are muchlarger when compared to the rat tissues For examplethe force to cut skin for sheep is 20 N versus 6 N for therat skin As for the Mayo scissors the liver for the sheepwas more textured than the rat liver and required alarger force 7 N versus 2 N

In summary it seems that the differences in hapticsensations resulted from a combination of varying tex-tural components and magnitude force differencesThese textures were not high frequency because approx-imately 99 of the power is in the frequency compo-nents below 5 Hz

4 5 Ef fect of Velocity

To test the inuence of the velocity u on the forceexperiments were performed at two speeds fast andslow which were determined by the surgeon while cut-ting As mentioned it was found that slow corre-sponded to a speed smaller than 15 degsec with anaverage of 7 degsec and fast to a speed greater than15 degsec with an average of 44 degsec The aver-age speeds depended on the material being cut For ex-ample the average fast speed for the blank runs wasgreater than 100 degsec whereas the average fastspeed for cutting skin was 25 degsec

To determine whether the results were signicantlydifferent between the two speeds another sum ofsquares analysis was used to determine the variability inthe slow runs SSES the fast runs SSEF and the variabilitybetween the average slow and fast runs SSES F Againentire curves were compared point by point along u be-cause no accurate model for the closing curve exists

The results of the SSE analysis are presented in table 3There did not seem to be a statistically signicant differ-ence between the slow and fast cutting speeds for any ofthe Mayo and Metzenbaum experiments but the Irisscissors gave a different result The fast versus slowspeed registered as signicantly different for the sheepskin cut longitudinally the rat skin cut longitudinally

the rat liver and the sheep muscle It seems prematureto draw concrete conclusions about the invariance ofthe force measure with respect to velocity except in thecase perhaps of the more sensitive Iris scissors

5 Pilot Study o f Cu tt ing Simulat ions

The ldquohaptic recordingrdquo approach was used to cre-ate cutting simulations A custom-designed haptic de-vice was developed to mimic Metzenbaum scissor han-dles and three cutting tasks (rat skin liver and tendon)and one blank condition could be displayed Despite thepresent limitations of the data and of the hardware it isshown that realistic simulations are possible

Two preliminary experiments were conducted to ex-amine the effectiveness of virtual simulations of scissorcutting The rst examined the realism provided by acustom-designed haptic display by comparing it to realMetzenbaum scissors The second experiment examinedthe performance of subjects while indentifying varioustissues using haptic feedback alone Their performancewas assessed both during real and virtual cutting replay-ing selected haptic recordings Quantitative and qualita-tive feedback indicated that the simulations were effec-tive at simulating the haptic experience of cuttingbiological tissues

5 1 Apparatus and Data Preparat ion

The haptic device consisted of a pair of Metzen-baum scissor arms one stationary and the other at-tached to a wheel driven by a motor by a frictionlesstransmission of ratio 351 as shown at the top of gure12 The system includes the device a power ampliercomputer interface boards and a control computer asshown in gure 13 The actuator used was a Maxon RE25 mm dia20 W motor with an HP HEDS 5540 en-coder Calibration of the angle between the scissor armsyielded a resolution of 0053 deg A PWM amplier(25A-series Model 12A8) from Advanced Motion Con-trol was used A Servo To Go card (Model 2) was usedto read the encoders and a Measurement Computingcard (CIO-DAS 160012) was used to provide DA

Greenish et al 643

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

The computer had a Pentium III 800 MHz processorA multithreaded control program displayed graphics at50 Hz and haptics at 1 KHz

While selecting the force proles suitable for cuttingsimulations it was observed that the forces differedwidely depending on the controlled variables of tissuetype animal type and scissor type We selected theMetzenbaum scissors and rat tissues because the forceswere generally lower thereby decreasing the possibilityof clipping the force signal The data collected fromempty scissors and three tissues (liver skin and tendon)were used Each data set shown in gure 14 displaysseveral cuts resulting in overlapping curves Recall thatwhen surgeons cut a repetitive motion is used thatnever closes fully the scissors

Before the data could be replayed for simulation it

was rst segmented as described in subsection 34 Us-ing these raw haptic recordings would not yield a realis-tic simulation because the subjects could use scissor an-

Table 3 Comparing Fast and Slow Cutting Speeds (Surgeon Data)

Scissor TissueMax(SSEF SSES) SSES F

Statisticallydifferent

Mayo Blank 16 11 NSkin (sheep L) 60 17 NSkin (sheep T) 25 16 NSkin (rat L) 22 08 NSkin (rat T) 12 09 NFascia (sheep) 15 14 NLiver (rat) 12 08 N

Metz Blank 11 09 NSkin (sheep L) 49 30 NSkin (sheep T) 63 49 NSkin (rat L) 26 12 NLiver (rat) 13 06 N

Iris Blank 20 12 NSkin (sheep L) 78 96 YSkin (rat L) 65 52 NSkin (rat T) 51 57 YFascia (sheep) 65 45 NLiver (rat) 41 09 YMuscle (sheep) 45 109 Y

Figure 12 The haptic scissors and a pair of real Metzenbaum

scissors are mounted on a stationary platform

644 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

gles not contained in the data Thus to limit theamount of ldquofabricatedrdquo data outside of the measuredangles we limited the motion to closing only extractingthe closing segment of one data loop for each tissueThis also helped to eliminate the effect of different cut-ting styles in identication experiments To provideforces for complete closure of the scissors the currentforce trend was maintained from the minimum angle inthe data to zero degrees

Within one segment of the data there were multipleforce measurements for each potentiometer readingThus the force measurements ranging from one totwenty data points at a single angle were averaged foreach angle Next the data were ltered with an expo-nentially weighted moving average lter to removenoise that felt clearly unnatural in preliminary simula-tions The data were then scaled to reect the correctforce at the handles of the scissors as the original datarepresented forces at the cutting point between the scis-sor blades A nal change was also made to lower thetendon forces the rat used to acquire data had stiffertendons than any of the rats to which we had access forperceptual experiments Thus we lowered the peakforce of the tendon simulation signicantly The naldata used in simulation are shown overlaid with theoriginal data in gure 14

Once the simulation data were obtained a lookuptable was created providing a list of angles and forceswith a one-to-one correspondence If the angle mea-

sured by the haptic scissors encoder was between theangles listed in the table linear interpolation was usedto determine the correct force for display The softwareenabled the experimenter to choose which simulationdata set to display (none blank liver skin or tendon)as shown in gure 15 This visual display was hiddenfrom the subjects during experiments

5 2 Met hods and R esu lt s

521 Realism of Haptic versus Real ScissorsIn this test the haptic scissors were programmed to sim-ulate the feel of blank Metzenbaum scissors during clos-ing Five male subjects ranging in age from 22 to 38were selected with the criterion that they had no previ-ous knowledge of the haptic scissors nor of the purposeof the test They also had no previous experience withsurgical or haptic devices The subjects were blindfoldedand asked to use three pairs of scissors mounted on astationary platform the haptic scissors (programmed toreproduce a blank cut) and two pairs of complete Met-zenbaum scissors labeled Scissors 1 and 2

Each subject was informed that there were at leastone real pair and at least one simulated pair After test-ing each pair by closing the scissor arms only once thesubjects had to decide whether each of the three scissorswere real or simulated One pair of the real Metzen-baum scissors had never been used before (Scissors 1)resulting in a slightly different feel from the other pairof real scissors which had been previously used for cut-ting rat tissues and cleaned (Scissors 2)

All subjects reported that the Scissors 1 were real butonly 60 thought that the Scissors 2 were real Fortypercent of the subjects judged that the haptic scissorswere real In other words the haptic scissors were some-times judged to be real but some subjects also thoughtthat the real scissors were articial

A likely reason why the haptic scissors did not feel realwas that they did not reproduce the correct behavior ofthe blades near 0 deg because we did not have data inthis angle range to use as a haptic recording An impor-tant comment was that the haptic scissors and thesmoother pair of Metzenbaum scissors felt very similarAnother observation was that real scissors of the same

Figure 13 The haptic system consists of the haptic device an

amplier computer interface boards and a control computer

Greenish et al 645

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

type could feel signicantly different from each other Inthis case the increased force in the second pair was dueto the fact that they had been used and washed whichprobably increased friction in the joint

522 Identi cation of Real and Virtual Tis-sues In the second experiment subjects cut real andvirtual rat tissues with real Metzenbaum or simulatedscissors and attempted to identify tissue types by hapticfeedback alone These tissues were by necessity fromanimals different from those used for the initial datagathering

During phase I the subjects rst visually inspectedthe prepared animals and made several cuts in each tis-sue with a pair of Metzenbaum scissors in liver skinand tendon They also tried blank cuts Then each sub-ject was blindfolded and proceeded to make blind cutsin blocks of four cases presented in a randomized orderThe samples were held between the blades of themounted scissors by a surgical assistant who also con-trolled the quality of the cut The subjects were in-formed that there was one sample of each tissue in eachblock Their task was to label the four cases by feelalone They were allowed to repeat some cuts to make a

Figure 14 Data used for haptic simulations The simulation data uses ltered data from one

closing of the scissors over rat tissues or blank It is represented in the gure as a thick gray line

Figure 15 Visual display for the haptic scissors

646 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

decision and to relabel them until they felt condence intheir judgment

During phase II the subjects repeated the tissue iden-tication with the virtual tissues approximately oneweek after they tested the real tissues in phase I Of thesix subjects participating in phase I four were availablefor phase II Because of the limitations of the recordand replay approach (as further discussed in subsection53) subjects were limited to experience haptic feed-back during closing only but could repeat the closingaction several times as in phase I The only difference isthat subjects were asked to limit the scissors opening tothe largest angle for which we had data

Fifty-four percent of the subject tissue identicationswere correct for real tissues (phase I) For virtual tissues(phase II) the percentage of correct identications was44 Results are shown in table 4 It was expected thatthe matching would be better with the real tissuesHowever the low success rates show that subjects ingeneral are not good at matching tissues by feel aloneThe t-value for comparing the means in the real andvirtual environments was 062 (eight degrees of free-dom p-value 029) so we conclude that the differ-ence is not statistically signicant That is subjects didnot perform signicantly worse with the virtual tissuesthan with the real tissues but this could be due to thelimited data from this pilot study

Qualitative feedback also indicated that completelyrealistic tissue display is not necessary to convince sub-jects of the ldquorealismrdquo of virtual environments involving

cutting However as several subjects commented afterthe experiments removal of acoustic noise from themotor forces when opening the scissors and transla-tional motion of the scissors while cutting are necessaryfor a true sense of ldquoimmersionrdquo in a virtual environ-ment

5 3 Discussion

Rendering simplicity is the main advantage of hap-tic recordings A lookup table is all that is needed forreal-time haptic display Any other data manipulationsuch as ltering and scaling can be performed as a pre-processing step In general subjective responses in theexperiments indicated that the haptic simulations feltrealistic although not ldquoexactly likerdquo the real rat tissues

Yet there are several limitations to using haptic re-cordings First appropriate display of the recorded hap-tic data required that we modify the original data signif-icantly Filtering scaling and fabricating data forldquoemptyrdquo angles were required to create simulations thatcould be compared against real tissues It is hard to de-termine the precise amount or type of ltering scalingor additional data that should be used without moredetailed empirical modeling Because these factors weredetermined ldquoby feelrdquo by the experimenter a certainamount of arbitrariness was injected into the nal simu-lation although we used direct haptic recordings

In addition we had to limit the subjects to a singleclosing of the scissors Although the data were recorded

Table 4 Experimental Results for Matching Tissue Types to Real and Virtual Tissues

Subject

Real Tissues (Phase I) Virtual Tissues (Phase II)

Blank Liver Skin Tendon Correct Blank Liver Skin Tendon Correct

1 S B L T 1 B L T S 22 L B S T 2 S B L T 13 B L T S 2 L B S T 24 B L S T 4 L B S T 25 B S L T 2 Did not perform experiment6 B L T S 2 Did not perform experiment

Greenish et al 647

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

in loops of opening and closing an even larger amountof fabricated data would have been needed to simulatethese loops The stationary mounting limited the simu-lation to the closing action If subjects were to performcontinuous cutting motions it would be more naturalto move the scissors forward This could be accom-plished by mounting the scissors on a translational hap-tic device or on a haptic device that was designed forhaving this capability (Hayward et al 1998)

A nal limitation is that a haptic recording is essen-tially a snapshot of the ldquofeelrdquo for one particular tissue atone instance Depending on the history of the tissuesmeasured the data may not reect the average tissueforces It was clear from our experiments that the ten-don forces of the comparison rats were much lower thanthose originally measured Because we scaled a portionof the high forces down to match the available tissues byfeel there was no guarantee that our scaling was percep-tually correct These issues underscore the importanceof continued data collection analysis and modeling

Although the authors believe that more completemodeling is the ideal approach using haptic recordingsprovided signicant insight Using the direct data wewere able to determine the mechanical and control re-quirements of the haptic system We were also able toverify that realistic simulations were indeed possible atleast in some cases before embarking on a lengthymodeling procedure

6 Conclusions

The forces involved in the cutting of tissues forthree surgical scissors were acquired and analyzed Al-though the experimental method allowed for the deter-mination of general trends in the data exact quantita-tive measures for the forces required to cut tissuesremain indeterminate Several improvements to the dataacquisition method have been identied and proposedThe acquired data were displayed on a specialized hapticinterface using the ldquohaptic recordingrdquo rendering ap-proach Although straightforward and computationallyefcient this method lacks the exibility for surgicalsimulation This result motivates the need for further

experiments and parameter-based models describing thecutting forces Equally important is the necessity to pur-sue careful psychological testing to relate these parame-ters to the performance of subjects

6 1 Future R ecommendat ions

Improvements include eliminating the grip effectfrom future data collection because it partially inhibitsthe determination of the innate tissue properties Con-versely this raises the subtle issue that a simulator mightneed to account for the grip At least two grip typeshave been identied the pull-open and push-closedgrips but others may exist especially when the scissorsare at rest The surgeonrsquos grip is quite different fromthat of the studentrsquos which indicates the necessity toinvolve a trained surgeon in experiments at these earlystages (or conversely to use measurements as a perfor-mance assessment technique during training)

Larger sample sizes are desirable to average out someof the sources of error and variability between tissuesamples The tissue variability is a result of innate tissuedifferences from region to region especially for largetissues such as skin or muscle and from the varying tis-sue structures attached to the tissue of interest Tissuemust be prepared carefully to obtain the greatest unifor-mity possible between runs

Equally important for the determination of the exactforces to cut tissues is the moisture content of the tis-sue This should be controlled for or accounted forVideotaping and having assistants in the operating roomwould allow appropriate measure to be taken to controlthe tissue thickness and the visible differences betweentissue samples

Furthermore it was determined that the transverseand longitudinal cutting conditions may not be neces-sary for all of the experimental conditions Once thisresult is validated the cutting direction could be elimi-nated as a parameter for skin cuts Limiting the data setis especially important as uniform tissue is difcult toobtain Further experiments in the cutting directions ofother tissues are warranted

The scissor dynamics themselves are among the fac-tors that inuence what the surgeon feels The length of

648 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

the scissor arms determines the mechanical advantagethat inuences the amount of force the user must applyat the handle The blade characteristics such as edgesharpness joint screw tightness and degree of curvaturechange the feel of the scissors For these reasons it isimportant to test the scissors before purchasing them andto determine quantiable metrics for these parameters

An improvement to the force measurement would bea better rejection of off-axis stresses and thermal effectsThis might prove worthwhile in reducing the variabilityin the data

6 2 Important Trend s

Although the data collection method would bene-t from these improvements several observations arestill possible First looking at the average maximumforce the measurable difference between tissues includea larger magnitude in the force applied for thicker andstronger tissues In descending order forces required tocut the rat tissues were found to be greatest for the ten-don then the muscle skin and nally the liver For thesheep tissues this order was determined to be tendonskin muscle fascia and nally liver Although there is amagnitude force difference between tissues an equallyimportant difference may be in the textural componentsof the force versus angle curves Whereas the liver exper-iments for the rat and muscle experiments for the sheepare relatively smooth the skin and fascia curves are moreridged or textured

This texture consists of only low-frequency signalcomponents because approximately 99 of the signalcomponents are below 5 Hz Therefore a simulatormay provide realistic haptic interaction with low band-width For the tendon however there is no measurabletextural components The force increases until break-through after which the force drops off quickly Bycomparing two cutting speeds the force was found tobe independent of the cutting rate for the Mayo andMetzenbaum scissors

Looking into invariants between either tissues or scis-sors provided the observation that the force resolved atthe cutting point required to cut a tissue consistentlyincreases as the scissor angle closes This can be ex-

plained by the amount of deformed tissue due to thechange in contact geometry as the scissors close Theresponse to cutting thus highly depends on the detailsof tissue deformation and rupture which in turn bothdepend on the macro and the micro geometry of thescissor blades As well the cutting curve is highly re-peatable within a run even if the cutting incident anglevaries which indicates that the tissue cutting force char-acteristics can be determined with the measured forceand scissor angle alone These ndings are consistentwith the cutting model recently introduced by Mahvashand Hayward (2001) Other cutting dynamics identiedin the data include the breakthrough angle and the con-tact area between the blades

A ck now ledgments

At McGill University initial funding for this research was pro-vided by the project ldquoHaptic Devices for Teleoperation andVirtual Environmentsrdquo (HMI-6) supported by IRIS (Phase 2)the Institute for Robotics and Intelligent Systems part ofCanadarsquos National Centers of Excellence program (NCE) andan operating grant ldquoHigh Performance Robotic Devicesrdquo fromNSERC the Natural Science and Engineering Council of Can-ada Additional funding for this research was provided by anNSERC UniversityIndustry Technology Partnership GrantMcGillMPBT ldquoA Balanced Haptic Device for ComputerUserInteractionrdquo

At Johns Hopkins University Drs Homayoun Mozaffariand Romer Geocadin both of the Johns Hopkins MedicalInstitutions were very helpful in conducting experiments withrat tissues Torrey Bievenour designed the initial prototype ofthe haptic scissors Ms Chial was supported by the NSF Re-search Experience for Undergraduates Program through TheJohns Hopkins University NSF Engineering Research Centerfor Computer Integrated Surgical Systems and Technology(grant EEC9731478)

R eferences

Ackerman M J (1994) The visible human project Medicinemeets virtual reality II Interaction technology and health-care (pp 5ndash7) Amsterdam IOS Press

Greenish et al 649

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

Baumann R Maeder W Glauser D amp Clavel R (1998)Force feedback for minimally invasive surgery Medicinemeets virtual reality IV Interaction technology and health-care (pp 564ndash579) Amsterdam IOS Press

Basdogan C Ho C Srinivasan M A Small S amp DawsonS (1998) Force interactions in laparoscopic simulationsHaptic rendering of soft tissues Medicine meets virtual real-ity IV Interaction technology and healthcare (pp 385ndash391)Amsterdam IOS Press

Basdogan C Ho C-H amp Srinivasan M (1999) Simula-tion of tissue cutting and bleeding for laparoscopic surgeryusing auxiliary surfaces Medicine meets virtual reality VII(pp 38ndash44) Amsterdam IOS Press

Bicchi A Canepa G De Rossi D Iacconi P amp ScilingoE P (1996) A sensorized minimally invasive surgical toolfor detecting tissutal elastic properties Proceedings of theIEEE Int Conf on Robotics and Automation 884 ndash888

Biesler D amp Gross M H (2000) Interactive simulation ofsurgical cuts Proceedings of IEEE Pacic Graphics 2000116ndash125

Brower I Ustin J Bentley L Sherman A Dhruv N ampTendick F (2001) Measuring in vivo animal soft tissueproperties for haptic modeling in surgical simulation Medi-cine Meets Virtual Reality 2001 (pp 69 ndash79) AmsterdamIOS Press

Carter F J Frank T G Davies P J amp Cuschieri A(2000) Puncture forces of solid organ surfaces Surg En-dosc 14(9) 783ndash786

Delp S Loan P Basdogan C amp Rosen J (1997) Surgi-cal simulation An emerging technology for training inemergency medicine Presence Teleoperators and VirtualEnvironments 6(2) 147ndash159

Duck F A (1990) Physical properties of tissue A comprehen-sive reference book London Academy Press

Green P S Jensen J F Hill J W amp Shah A (1993)Mobile telepresence surgery (Tech Rep) SRI International

Hannaford B Trujillo J Sinanan M Moreyra M RosenJ Brown J Lueshke R amp MacFarlane M (1998)Computerized endoscopic surgical grasper Medicine meetsvirtual reality IV Interaction technology and healthcare (pp265ndash271) Amsterdam IOS Press

Hayward V Gregorio P Astley O Greenish S DoyonM Lessard L McDougall J Sinclair I Boelen SChen X Demers J-P Poulin J Benguigui I AlmeyN Makuc B amp Zhang X (1998) Freedom-7 A highdelity seven axis haptic device with application to surgicaltraining Experimental robotics V In Casals A de

Almeida A T (Eds) Lecture Notes in Control and Infor-mation Science vol 232 (pp 445ndash456) New YorkSpringer-Verlag

Hirota K Tanaka A amp Kaneko T (1999) Representationof force in cutting operation Proceedings of IEEE VirtualReality p 77

Hunter I Jones L Sagar M Lafontaine S amp Hunter P(1995) Ophthalmic microsurgical robot and associated vir-tual environments Computers in Biology and Medicine25(2) 173ndash182

Kuppersmith R B Johnston R Jones S B amp HermanA J (1996) Virtual reality surgical simulation and otolar-yngology Arch Otolaryngol Head Neck Surg 122 1297ndash1297

Longnion J Rosen J Sinanan M amp Hannaford B(2001) Effects of geared motor characteristics on tactileperception of tissue stiffness Medicine Meets Virtual Reality2001 (pp 286 ndash292) Amsterdam IOS Press

MacFarlane M Rosen J Hannaford B Pellegrini C ampSinanan M (1999) Force feedback endoscopic grasperhelps restore the sense of touch in minimally invasive sur-gery Journal of Gastrointestinal Surgery 3 278 ndash285

MacLean K (1996) The ldquoHaptic Camerardquo A technique forcharacterizing and playing back haptic properties of real en-vironments Proceedings of ASME Dynamic Systems andControl Division 58 459ndash467

Mahvash M amp Hayward V (2001) Haptic rendering ofcutting A fracture mechanics approach Haptics-e The Elec-tronic Journal of Haptics Research 23 [Available on-linewwwhaptics-eorg]

Mann P S (1998) Introductory statistics 3rd ed New YorkJohn Wiley amp Sons

Okamura A M Dennerlein J T amp Howe R D (1998)Vibration feedback models for virtual environments Pro-ceedings of the IEEE International Conference on Roboticsand Automation 1 674 ndash679

Okamura A M Hage M W Dennerlein J T amp Cutko-sky M R (2000) Improving reality-based models for vi-bration feedback Proceedings of the ASME Dynamic Systemsand Control Division 69 1117ndash1124

Ottensmeyer M P amp Salisbury J K (2001) In vivo dataacquisition instrument for solid organ mechanical propertymeasurement In MICCAI 2001 LNCS 2208 (pp 975ndash982) New York Springer-Verlag

Pai D K Lang J Lloyd J E amp Woodham R J (2000)ACME A telerobotic active measurement facility In PCorke and J Trevelyan (Eds) Experimental Robotics VI

650 PRESENCE VOLUME 11 NUMBER 6

Lecture Notes in control and information sciences 250 (pp391ndash400) New York Springer-Verlag

Picinbono G Delingette H amp Ayache N (2000) Nonlin-ear and anisotropic elastic soft tissue models for medicalsimulation Proceedings of the IEEE International Conferenceon Robotics and Automation 1370ndash1375

Richard C Cutkosky M R amp MacLean K (1999) Fric-tion identication for haptic display Proceedings of theASME Dynamic Systems and Control Division 67 327ndash334

Richmond J L amp Pai D K (2000) Active measurementand modeling of contact sounds Proceedings of the IEEEInternational Conference on Robotics and Automation2146ndash2152

Rosen J Hannaford B MacFarlane M amp Sinanan M(1999) Force controlled and teleoperated endoscopicgrasper for minimally invasive surgery Experimental perfor-mance evaluation IEEE Transactions on Biomedical Engi-neering 46 1212ndash1221

Satava R M (1996) Advanced simulation technologies forsurgical education Bull A Coll Surg 81(7) 77ndash81

Satava R M amp Jones S B (1997) Virtual environments formedical training and education Presence Teleoperators andVirtual Environments 6(2) 139 ndash146

Vuskovic V Kauer M Szekely G amp Reidy M (2002)Realistic force feedback for virtual reality based diagnosticsurgery simulators Proceedings of the IEEE InternationalConference on Robotics and Automation 1592ndash1598

Greenish et al 651