mechanically assisted 3d ultrasound guided prostate biopsy ......blood tests are common screening...

Mechanically assisted 3D ultrasound guided prostate biopsy systemJeffrey Baxa�

Robarts Research Institute, London, Ontario, N6A 5K8, Canada and Biomedical Engineering,University of Western Ontario, London, Ontario, N6A 5B9, Canada

Derek CoolRobarts Research Institute, London, Ontario, N6A 5K8, Canada and Department of Medical Imaging,University of Western Ontario, London, Ontario, N6A 5C1, Canada

Lori Gardi, Kerry Knight, David Smith, Jacques Montreuil, and Shi SherebrinRobarts Research Institute, London, Ontario, N6A 5K8, Canada

Cesare RomagnoliRobarts Research Institute, London, Ontario, N6A 5K8, Canada and Department of Medical Imaging,University of Western Ontario, London, Ontario, N6A 5C1, Canada

Aaron FensterRobarts Research Institute, London, Ontario, N6A 5K8, Canada; Department of Medical Imaging,University of Western Ontario, London, Ontario, N6A 5C1, Canada; and Biomedical Engineering,University of Western Ontario, London, Ontario, N6A 5B9, Canada

�Received 3 July 2008; revised 15 September 2008; accepted for publication 23 September 2008;published 10 November 2008�

There are currently limitations associated with the prostate biopsy procedure, which is the mostcommonly used method for a definitive diagnosis of prostate cancer. With the use of two-dimensional �2D� transrectal ultrasound �TRUS� for needle-guidance in this procedure, the physi-cian has restricted anatomical reference points for guiding the needle to target sites. Further, anymotion of the physician’s hand during the procedure may cause the prostate to move or deform toa prohibitive extent. These variations make it difficult to establish a consistent reference frame forguiding a needle. We have developed a 3D navigation system for prostate biopsy, which addressesthese shortcomings. This system is composed of a 3D US imaging subsystem and a passive me-chanical arm to minimize prostate motion. To validate our prototype, a series of experiments wereperformed on prostate phantoms. The 3D scan of the string phantom produced minimal geometricdistortions, and the geometric error of the 3D imaging subsystem was 0.37 mm. The accuracy of3D prostate segmentation was determined by comparing the known volume in a certified phantomto a reconstructed volume generated by our system and was shown to estimate the volume with lessthen 5% error. Biopsy needle guidance accuracy tests in agar prostate phantoms showed that themean error was 2.1 mm and the 3D location of the biopsy core was recorded with a mean error of1.8 mm. In this paper, we describe the mechanical design and validation of the prototype systemusing an in vitro prostate phantom. Preliminary results from an ongoing clinical trial show thatprostate motion is small with an in-plane displacement of less than 1 mm during the biopsyprocedure. © 2008 American Association of Physicists in Medicine. �DOI: 10.1118/1.3002415�

Key words: prostate, biopsy, 3D ultrasound, 3D TRUS, needle tracking

I. INTRODUCTION

Digital rectal exams and prostate-specific antigen �PSA�blood tests are common screening tests for prostate cancer�PCa� in asymptomatic men. Combined with the public’s in-creased awareness of PCa and PSA testing, the proportion ofmen diagnosed with early stage prostate cancer hasincreased.1,2 When diagnosed at an early stage, the disease istreatable,3,4 and even at later stages, treatment can be effec-tive. However, if a tumor has extended beyond the prostate,the risk of metastases increases significantly.

The definitive diagnosis of PCa requires histological as-sessment of tissue cores drawn from the prostate during abiopsy procedure. The physician uses a conventional 2Dtransrectal ultrasound �TRUS� probe to guide the needle into
the prostate. Biopsies are commonly performed with an 18-
5397 Med. Phys. 35 „12…, December 2008 0094-2405/2008/35„

gauge needle mounted on a spring-loaded biopsy “gun.” Thisgun is connected to an “end-firing” 2D TRUS probe, by aguide, which forces the needle to stay in the imaging planeso it is always visible in the US image. The TRUS probe isheld by the physician and inserted into the patient’s rectumin order to image the prostate through the rectal wall �Fig. 1�.Typically, an average of 10 biopsy cores are obtained in asingle biopsy procedure.5 Each core is separately identifiedas to its location within the prostate so that the pathologistcan report the extent and grade of cancer.

While the TRUS-guided prostate biopsy has become acommonly performed urological procedure, it is not withoutlimitations. Many patients’ initial biopsy will be negative forPCa; however, because the sensitivity of the procedure is
poor, PCa cannot be ruled out in these patients. Physicians
539712…/5397/14/$23.00 © 2008 Am. Assoc. Phys. Med.

http://dx.doi.org/10.1118/1.3002415

http://dx.doi.org/10.1118/1.3002415

5398 Bax et al.: Mechanically assisted 3D ultrasound guided prostate biopsy system 5398

currently try to obtain occult cancer samples by taking biop-sies from predetermined regions of the prostate, which havea high probability of harboring cancer. Not only is the vol-ume of the biopsy sample small but the presence of PCa isalso often multifocal, involving only a small part of the pros-tate in the early stages of the disease.3,6 As a result, a patientmay be informed of a negative biopsy result but in fact maybe harboring an occult early stage cancer. The managementof these patients, in addition to patients diagnosed with earlystage disease is currently a major challenge.

Since a negative result does not preclude the possibility ofa missed cancer,7 patients undergo repeat biopsies when in-dicated by clinical suspicion and in cases when a positivebiopsy for cancer would have therapeutic consequences.Since there are an appreciable number of men with false-negative biopsy who, in fact, harbor, curable PCa, the phy-sician is faced with a difficult challenge. Sometimes thesepatients are imaged with other modalities and undergo a sec-ond or even a third biopsy, in which the physician tries toavoid the location of the first negative biopsy.

Various reports have shown that the detection rate of PCafor repeat biopsy procedures ranges from 10% to 25% �afterinitial biopsy results were shown to be negative�.3,8–10 Be-cause PCa is present in at least 1 /10 to 1 /4 of patients whohave undergone an initial negative biopsy, current biopsyprocedures are still suboptimal.6,11,12 If an initial biopsy failsto detect cancer, how should a repeat biopsy be directed?Should the repeat �and initial� biopsy be lesion directed, ran-dom, or based on the details of the patient’s anatomy �e.g.,prostate regions, volume, shape�?3,13

Another important challenge facing physicians is men di-agnosed on biopsy to have premalignant lesions, i.e., high-grade prostatic intraepithelial neoplasia, and particularlyatypical small acinar proliferation �ASAP�.14 These are chal-lenging to manage as there is a 40%–50% chance of findingcancer on repeat biopsy with ASAP.15 Since coexisting can-cer might be present, especially with ASAP, where the pa-

Bladder

Needle

TRUS probe

Prostate

Urethra

FIG. 1. Schematic drawing of a biopsy procedure illustrates how a forwardviewing probe captures images of the prostate and nearby organs �bladderand urethra�. The needle guide constrains the needle path to stay in theimaging plane of the TRUS probe so it is always visible in the US image.

thologist finds only a small amount of histological “atypia”

Medical Physics, Vol. 35, No. 12, December 2008

but not enough material to confidently diagnose cancer, thesepatients require a repeat biopsy soon after the first. In thesesituations, it is vital to rebiopsy the same area.14 Currently,2D US provides only a vague location of the abnormal find-ings, and it is not possible to be certain that the same areahas been sampled by the repeat biopsy.

Due to the increasing number of younger men with poten-tially early and curable PCa undergoing repeated prostatebiopsy, it is important not to rebiopsy the same area if theoriginal biopsy was negative, and it is particularly vital torebiopsy the exact area if a possible abnormal area was de-tected on first biopsy as ASAP.14,15 Thus, the locations of thecores obtained in 3D from the prostate must be known accu-rately to help guide the physician during the repeatbiopsy16,17 to help in correlating any imaging evidence of thedisease, and to provide improved planning for any subse-quent therapy.

As there is a clinical need for a prostate biopsy systemallowing recording of the core locations and guiding tools toallow sampling of specific targets in 3D, special purposeTRUS biopsy systems have emerged in the marketplace. En-visoneering Medical Technologies �St. Louis, MO� has de-veloped a system, which uses a stationary endorectal ultra-sound probe for biopsy. The Voluson prostate biopsy system�General Electric� is a 3D hand held US imaging system thatallows for real-time imaging of the prostate. Several groupshave reported using 3D prostate systems in both phantomand patient studies.18–21 These studies have shown an im-provement in the positional and the diagnostic accuracy ofthe procedure.20,21 However, there is still a group of patientswith ultrasound occult cancers.20 As a result, the use of real-time 3D US probes suffers from the same problems as 2Dprobes.20

MRI imaging using high resolution T2 weighted images,and MRI spectroscopy are promising methods that can im-prove the detection of prostate cancer.22 Several groups haveproposed and developed a number of potential solutions thatuse CT or MRI for prostate biopsy.23–25 Fichtinger et al.integrated CT with a side-firing TRUS system to guide theneedle to a preplanned target using a robotic arm.23 MRI hasalso been used to identify potential malignant tumors while arobotic device has been used to guide the biopsy needle.24,25

The DaVinci robot �Intuitive Surgical Inc., Sunnyvale, CA�has been used extensively for prostate surgery, and can alsobe adapted to guide a biopsy needle. However, these solu-tions require physicians to use CT,23 MRI,24,25 or speciallymodified biopsy systems, which are costly. As well, theworkflow for each of the new designs differs significantlyfrom current biopsy protocols, requiring physicians to retrainthemselves. Furthermore, since there are more than one mil-lion biopsies performed each year,26 it would be beneficial todevelop a biopsy system that utilizes ultrasound equipmentcurrently in use for the procedure. Further, the use of 3D UScoupled to a mechanical navigation system has the potentialto provide a reproducible record and allow for image fusionwith other imaging modalities such as MRI, assisting thephysician in planning a repeat biopsy. This system can po-
tentially eliminate most of the user variability of using con-


ventional nonfixed hand-held probes that render them unsuit-able for precision biopsy, while preserving some of the userfamiliarity.

In this paper, we describe the development and testing ofa mechanically assisted 3D TRUS prostate biopsy system,which addresses the limitations of current prostate biopsyprocedures and also minimizes adoption cost and physicianretraining. We also report briefly on the initial clinical use ofthis system for prostate biopsy.

II. SYSTEM DESIGN

Our 3D TRUS system is an integrated 3D workstation,mechanical guidance system, and software tools to record thecore locations and guide the needle to its target. It allows realtime tracking and recording of the 3D position and orienta-tion of the biopsy needle as the physician manipulates theUS transducer. The system uses: �1� passive mechanicalcomponents for guiding, tracking, and stabilizing the posi-tion of a commercially available end-firing transreactal UStransducer, �2� software components for acquiring, storing,and reconstructing �in real time� a series of 2D US imagesinto a 3D US image, and �3� software that displays a modelof the 3D scene to guide and record the biopsy core locationsin 3D.

II.A. Mechanical tracking system

The mechanical assembly consists of a passive fourdegree-of-freedom tracking device, an adaptable transducercradle, and a hydraulic support �Fig. 2�. The transducercradle mechanically locks and fastens a commercially avail-able end-firing TRUS transducer. The tracking device andtransducer cradle are secured by the hydraulic support, whichstabilizes the mechanical assembly while the physician per-forms a biopsy. When the TRUS probe is maneuvered, thesoftware records the 3D position and orientation of the trans-

SprCou

TRUSTransducer

Needle Guide

HS

TrackingMechanism

Base

ducer tip via absolute encoders as described below.


The end-firing TRUS transducer �with the biopsy needleguide in place� is mounted to the mechanical trackingmechanism where the US probe is free to rotate and slidealong its long axis �Fig. 3, wrist joint axis of rotation: J3�.This allows the physician to insert the TRUS transducerthrough the patient’s rectum to view the prostate and to ro-tate it to acquire a 3D image. The tracking linkage containsangle-sensing encoders �Fig. 3� mounted to each joint to

oadedbalance

ulicizer

FIG. 2. Photograph of the tracking system to be usedfor 3D US guided prostate biopsy. The system ismounted at the base to a hydraulic actuated stabilizerwhile the linkage allows the TRUS transducer to bemanually manipulated about a remote center of motion,to which the center of the probe tip is aligned. Thespring loaded counterbalance was designed to fully sup-port the weight of the system throughout its full rangeof motions about the RCM.

DifferentialGearingMechanism

Encoders

Encoder

Encoder

J1

J2

J3

L1

RCM

L2

L3

FIG. 3. A front perspective view of the mechanical tracker, which in turn isattached to a multijointed stabilizer at L1. The RCM is at the intersection ofthe primary, secondary, and tertiary axes. The encoders mounted at the pivotJ1 is used to measure the relative angles between the two successive links L2

and L3. The encoder at pivot J2 measures the angle between L1 and L2. Thedifferential gearing mechanism, which is coupled to the tracking linkage,decouples the two DoF provided by the cylindrical joint J3 supporting theshaft, which in turn is mounted to the transducer cradle. These two degreesof mobility represent the probe penetration and angular orientation about itslongitudinal axis, respectively. Two additional encoders, mounted onto thewrist joint J3 are required to measure the angle and depth of penetration of

ing lnter

ydratabil

the probe through the differential gear train.


transmit the angle between the arms to the computer. Thisencoder arrangement allows for the continuous computationof the position of the transducer needle guide as the trans-ducer is manipulated in the rectum.

The mechanical tracking device is a spherical linkage as-sembly that consists of three links and three hinged connec-tions. The axis of each hinged connection converges to acommon point to produce a remote center of motion�RCM�.27–29 The base link �Fig. 3, L1� defines the referenceaxis of the proposed coordinate system and is fixed to amultijointed stabilizer �Fig. 3, L1� that attaches to a cart. Theangle between each hinged connection in the mechanism de-fines the size and shape of the operating envelope of thekinematics frame, which allows for two degrees of rotation�pitch and yaw� about the RCM.

The linkage assembly supports the TRUS transducerthrough the distant pinned connection �Fig. 3, wrist jointaxis: J3� such that the long axis of the probe passes throughits RCM. Therefore, the angular position of the axis of theprobe relative to the base is determined by measuring theangle between two successive joint axes. Two encodersmounted at the pivots �shoulder joint axis: J1� and �elbowjoint axis: J2� are used to measure the relative angles betweenthe two successive linkages �Fig. 3, links: L1 and L2� andbetween the link L1 and stabilizer, respectively.

The probe is supported through the revolute axis by asleeve and is allowed to pivot and slide freely along the axisof the revolute joint �Fig. 3, wrist joint axis: J3�. This com-pound joint gives the tracking mechanism an additional twoDoF where the probe penetration and relative rotation angleto the supporting frame are defined. As illustrated in Fig. 3,the differential gearing mechanism, which is coupled to thetracking mechanism, decouples the motion through the com-pound joint. These two degrees of mobility represent theprobe penetration and angular orientation about its long axis�i.e., roll angle�, respectively.

II.B. Forward kinematics equations of motion

The unit vector r̂ defines the 3D orientation of the TRUStip in spherical coordinates relative to its fulcrum �O� �Fig.4�:

r̂ = �x

y

z� = �cos � sin �

sin � sin �

cos �� . �1�

In this description, the position of the probe tip r̂= f�� ,�� is specified by the angle �, the angle long axis ofthe probe makes with the z-axis, and the angle �, which is theorientation of the TRUS transducer in the x-y plane. Therelationship between the tracker coordinate system r̂= f�� ,�� and the reference frame defined by the encodersconnected at each of the hinged connections is illustrated inFig. 4. The tracker arm configuration measured by the encod-ers is defined by the spherical triangle �ABC�, and is linked
to the global reference frame by the spherical triangle �APC�.

The forward kinematics equations of motion for the open-chain linkage �Eqs. �2a�, �2b�, and �2c�� were derived byapplying the Napier analogies to spherical triangle APC �Fig.4�:28

tan1

2�� + �� = cos

1

2�� −

�

2�sec

1

2�� +

�

2�cot

1

2� , �2a�

tan1

2�� − �� = sin

1

2�� −

�

2�csc

1

2�� +

�

2�cot

1

2� , �2b�

tan1

2� = tan

1

2�� −

�

2�sin�� + ��cos�� − �� . �2c�

This defines spherical coordinates of the vector r̂= f �� ,�� interms of the geometric configuration of the linkage angles �and � �see Fig. 4�.

Equations �3a� and �3b�, which define the configuration ofthe linkage in terms of the angles measured by the encodersat the shoulder and elbow joints, respectively, were derivedby solving the right spherical triangle �ABE�:28

tan1

� = cos � , �3a�

FIG. 4. Diagram illustrating the relationship between the tracker linkageorientations relative to the global Cartesian reference frame. The angle be-tween the joints A and B in link L2 �AOB� is � /4. Likewise, the anglebetween the joints B and C connections in link L3 �BOC� is also � /4.

2


cot1

2� =

12

tan . �3b�

The position of each arm �AB and BC in Fig. 4� in thelinkage was determined by measuring the spherical angles ateach of the pinned couplings A and B, respectively. The en-coder mounted at A measures the angle ��+�� between L1

and the x-z plane, and the encoder mounted at B measuresthe angle between the two arms �L1 and L2�. Equation �3b�is needed to decouple the values for �� and ��, required tosolve Eqs. �2a�–�2c�.

II.C. Software for acquiring and reconstructing 3D USimages

Before the physician acquires a 3D image of the prostate,the tracking arm is locked into place to prevent the TRUSprobe from changing its pitch, yaw, and depth of penetrationwhile the probe is rotated. As the physician rotates the trans-ducer manually about its long axis �Fig. 3, wrist joint axis:J3� to acquire a 3D scan composed of 200 images in 1 degincrements, the 2D US images from the US machine aredigitized using a frame grabber and reconstructed into a 3Dimage.30 The last 20 images in the overlap region �from180 deg to 200 deg and 0 deg to 20 deg of transducer rota-tion� are merged by averaging the duplicate images to re-move any slight discontinuities that may arise from imagelag31 or small patient motion. Our software has a graphicalsoftware interface to provide information to the physician onthe proper rotational speed. The software interface provides a

B

BBladder

ProstateBoundary

Prostate

visual queue to indicate if the rotation was too rapid and the


2D US images were not acquired properly. Once the 3DTRUS image has been acquired, the software then displaysthe image in a cube view �Fig. 5�.30

A model of the prostate is then generated from the 3Dimage by a semiautomatic 3D segmentation algorithm devel-oped in our laboratory and described in detail elsewhere.31,32

The segmentation algorithm requires that the physician se-lects four points around the boundary of a 2D prostate crosssection. The algorithm then segments the prostate by fitting adynamic deformable contour to match the boundary of theprostate. The contour is then used in the adjacent slice as atemplate and deformed around the prostate boundary; theresult is propagated through 180 deg and refined for eachsucceeding image slice.32 After the software reconstructs theimage and the model, the physician can slice through the 3DTRUS image to select target biopsy locations. If the proce-dure is a rebiopsy, then previous biopsy locations can beviewed for planning as shown in Fig. 6.

II.D. Software components for 3D tracking andrecording

Once the 3D image scanning and biopsy plan are com-plete, the system displays a 3D needle guidance interface tofacilitate the systematic targeting of each biopsy location.Throughout the biopsy procedure, the 3D location and orien-tation of the TRUS transducer is tracked and displayed inreal time. Figure 7 shows the biopsy interface, which is com-posed of four windows: the 2D TRUS video stream, the 3DTRUS image, and two 3D model views. The 2D TRUS win-

Urethra

er

er

FIG. 5. Axial �top left�, sagittal �topright�, and coronal views �bottomright� of the 3D TRUS image of a pa-tient’s prostate. The coronal view ofthe prostate is not possible using thecurrent 2D TRUS biopsy procedure.The �bottom left� shows the live videostream from the US machine. The im-age within the green bounding boxwas digitized by a frame grabber asthe physician rotated the TRUS trans-ducer and then reconstructed into the3D image shown in the other threewindows. The bladder �hypoechoic re-gion anterior to the prostate boundary�and urethra is visible within the axialand sagittal views.

ladd

ladd

dow �bottom left� displays the real-time 2D TRUS image


streamed from the ultrasound machine. The 3D TRUS win-dow �top left� contains the 3D TRUS image sliced in realtime to correspond to the orientation and position of theTRUS probe. This correspondence allows the physician tocompare the static 3D image with the real-time 2D image.Finally, the two 3D targeting windows show: �i� the coronaland perspective views of the 3D prostate model, �ii� the real-time position of the 2D TRUS image plane, and �iii� theexpected path of the biopsy needle as defined by the biopsyguide.

ProstateBoundary

BiopsyCore

BiopsyCore

TRUS

TargetingRing

US ImagePlane

Bladder

BladderNeedleAxis

TargetingRing


The targeting windows are used to aid the physician inguiding the needle to each preplanned target. The targetingcircle on the screen �Figs. 6 and 7� illustrates all accessibleneedle trajectories by rotating the US probe about its longaxis. This shows the physician the shortest distance to eachpreplanned target.

III. SYSTEM VALIDATION METHODS

We used an HDI 3500 US system and C9-5 �9–5 MHz�TRUS probe �Philips Medical Systems, Seattle, WA� in the

psyre

tateel

NeedleAxis

BiopsyCore

rostateodel

NeedleAxis

FIG. 6. For rebiopsy, previous biopsy locations in a pa-tient’s prostate can be viewed to plan which the physi-cian can slice to view the previous biopsy locations.This image also allows the physician to compare thepreviously segmented prostate boundary �red/gray line�with the currently segmented boundary �green/whiteline� �bottom left�. The coronal view of the patient’sprostate with the segmented boundary and the locationof the previous biopsy cores as circles �top and bottomright�. The 3D locations of the biopsy core are dis-played within the 3D prostate models. The targetingring in the bottom right window shows all the possibleneedle paths that intersect the selected target by rotatingthe TRUS about its long axis. The vertical white linethrough the targeting ring shows the orientation of theUS imaging plane. The bladder �hypoechoic region an-terior to the prostate boundary� and urethra is visiblewithin the axial view �top left�.

ostatedel

RUS

ostateodel

iopsyarget

FIG. 7. The 3D biopsy system interface is composed offour windows: �top left� the 3D TRUS image dynami-cally sliced to match the real-time TRUS probe 3D ori-entation �bottom left�, the live 2D TRUS video stream�right side�, and the 3D location of the targets is dis-played within the 3D prostate models. The targetingring in the bottom right window shows all the possibleneedle paths that intersect the preplanned target by ro-tating the TRUS about its long axis. This allows thephysician to maneuver the TRUS to the target �high-lighted by the red/gray dot� in the shortest possible dis-tance. The biopsy needle �arrow� is visible within thereal-time 2D TRUS image. The bladder �hypoechoicregion anterior to prostate� and calcifications within theprostate �arrow head� illustrate anatomical correspon-dence between the real-time and static 3D image.

BioCo

ProsMod

PM

PrMo

T

PrM

BT


following experiments. The resolution of the Philips scanhead was determined experimentally by measuring the diam-eter of a wire �full width at half-maximum� within a presetfocal zone 3 cm deep in a bath of distilled water and 7%glycerol by weight�, giving a speed of sound of 1540 m /s ata temperature of 20 °C. The axial, lateral, and elevationalresolution was 1.2, 1.4, and 2.5 mm respectively.

III.A. Geometric reconstruction

The accuracy of 3D geometric reconstruction of theTRUS image based on manual rotation of the transducer wasverified using a string phantom,33 which was immersed in abath of distilled water and 7% glycerol �by weight�. Thestring phantom consists of four layers of orthogonally inter-secting monofilament nylon string with a diameter of0.10 mm, which are anchored to the sides of a brass frame ofthe following dimensions: 12.2 cm �length� by 12.2 cm�width� by 3.0 cm �height�. The intersections of the stringsand each of the layers are spaced 10 mm apart, and eachlayer is offset by 2.5 mm so that the upper layers of stringsdo not shadow the subsequent lower layers.

The tracking apparatus was positioned with the tip of theTRUS transducer placed below the surface of the glycerolsolution. The transducer was then manually rotated 200 degaround its long axis �z-direction� to acquire a 3D US imageof the string phantom. Using this 3D image, the spacingbetween the strings was measured in the three principalviews �x ,y�,�x ,z�, and �y ,z��. The measurements for eachplane were then repeated 15 times, and the mean, standarddeviation, and the standard error for �x, �y, �z were thendetermined.

III.B. 3D segmentation

A certified ultrasound calibration phantom, which is con-structed of Zerdine®, �Computerized Imaging ReferenceSystems Inc., Norfolk, Virginia� was used to assess the ac-curacy of volumetric measurements. The phantom had anembedded 21.5 cm3 egg-shaped object 1 cm below the sur-face of the phantom simulating the location and size of asmall-medium prostate.

To verify the 3D segmentation accuracy of our system,the reconstructed volume of the embedded egg-shaped objectwas compared to that of its known volume. A 3D TRUSimage of the phantom was acquired and the semiautomatedsegmentation algorithm was then used to generate a 3Dmodel of the embedded object. The volume of the segmentedmodel was then compared to the certified volume of the ob-ject to quantify the volume measurement error of the 3Dsegmentation. The phantom was independently scanned 15times and each object was semiautomatically segmented us-ing the software described in Sec. II C to calculate mean andstandard deviation values for the volume error. The volumeerror Verror�V ,V*� was defined as the percentage differencebetween the reconstructed volume V and the known volume

*
of the scanned object V :

Verror =V − V*

V*. �4�

To determine the impact the semiautomated segmentationalgorithm has on the volume error, the following volumemetrics were used to compare the reconstructed volume to amanually segmented volume as a gold standard.34 The sensi-tivity S�V ,Vman� was used to determine the portion of thereconstructed volume V that intersects the manually seg-mented volume Vman:

S =V � Vman

Vman . �5a�

The difference D�V ,Vman� was used to measure the propor-tion of the segmented volumes that is not correctly over-lapped:

D =V � Vman − V � Vman

Vman . �5b�

III.C. Mock biopsy on prostate phantoms

Mock biopsy procedures were performed on an agar-based tissue-mimicking phantom to quantify the accuracy ofthe 3D TRUS system for biopsy guidance and recording the3D location of the biopsy core. The agar phantom consistedof an agar-based prostate model35 embedded in surroundingbackground agar.34,35 The background and prostate modelwere constructed by adding 7% by mass of glycerol solutionwith agar powder to produce a speed of sound similar to thatof human tissue �1540 m /s�.30 The mold of the prostatemodel was generated from a segmented 3D image of a hu-man prostate �volume=21.5 cm3�. The background materialcontained cellulose �15% by weight� to create acoustic back-scattering in US, making the surrounding region appearbright relative to the prostate. Tungsten powder �1% byweight� was added to the prostate model to increase the x-rayattenuation within the prostate, making it visible in the CTimage. Finally, stainless steel ball bearings of 1 mm diamwere placed in a geodesic configuration within the back-ground material, surrounding the prostate model. These ballbearings served as fiducials for coordinate registration be-tween the 3D TRUS and CT images.

Six independent mock biopsies were performed on fiveprostate phantoms for a total of 30 cores using the 3D TRUSbiopsy system. To start the procedure, the tracking apparatuswas positioned with the tip of the TRUS transducer contact-ing the surface of the agar phantom and aligned with thecenter of the prostate gland and immediately adjacent to itsposterior aspect. A 3D TRUS image of the prostate was thenacquired �and reconstructed� by manual 200 deg rotation ofthe TRUS probe about its long axis and was reconstructed ina Cartesian reference frame with isometric voxel dimensions�0.194 mm3�. A semiautomated segmentation was then usedto generate a 3D segmented model of the prostate.30–32,34 Atraditional sextant biopsy plan was followed, with contralat-
eral biopsy targets placed at the base, midgland, and apex of


the prostate. A systematic biopsy of each target was per-formed and the location of each biopsy core was recordedwithin the 3D system �e.g., Fig. 7�.

The agar phantom was imaged, postbiopsy, in an eXploreLocus Ultra Pre-Clinical CT scanner �GE Healthcare, Lon-don, ON�, which is used for small animal microimaging. Thehigh resolution CT image �isometric voxel dimensions of0.15 mm3� served as the gold standard for the “true” locationof each biopsy core. In the CT image, a biopsy core wasevident as an air track left from the insertion of the 18-gaugeneedle into the agar phantom. The location of a biopsy corewas extracted from a CT image by selecting multiple pointsalong the center of the air track, from the base to the tip. Alinear regression of the points was used to define the 3Dtrajectory of the needle and the location of the 19 mm biopsycore was defined relative to the tip of the air track. Thecenters of the 1 mm fiducial markers surrounding the pros-tate were manually selected from corresponding CT and 3DTRUS images to coregister the two coordinate systems.

III.C.1. Biopsy system errors

The 3D biopsy system was evaluated for its ability to: �a�guide a biopsy needle to a 3D target and �b� record the loca-tion of the biopsy core in 3D following the methods de-scribed by Cool et al.34 The process of directing a biopsyneedle toward a target using the 3D TRUS system has hu-man, machine, and tissue factors as all potential sources oferror. The needle guidance error NGE measured the meancomplete error associated with a user guiding the biopsyneedle to predefined targets. This error was quantified for allbiopsy sites n as the perpendicular distance between the ithtarget location �ai� and the corresponding “true” biopsy coreidentified in the 3D CT image �bi

CT�:

NGE =�i=1

n D�ai,biCT�

n, �6�

where D� � is a function measuring the 3D minimum dis-tance between a 3D target point ai and a 3D biopsy core linesegment, bi

CT. NGE is comprised of two quantifiable errors:one related to human guidance error NGHE and the otherrelated to the needle trajectory error NTE. Conceptually,NGHE represents the ability of the user to align the biopsytarget within the center of the expected needle path of thebiopsy system. NGHE was measured by comparing the 3DEuclidean distance between the biopsy targets ai and the ex-pected path of the biopsy needle pi within the 2D TRUSimage:

NGHE =�i=1

n ai − pi n

. �7�

It should be noted that the preplanned biopsy targets ai

were virtual targets, fixed within the 3D TRUS system track-ing space, not within the real 3D world. Therefore, both thetarget ai and the expected needle path pi were within thesame 3D tracking space and NGHE was only influenced by
human errors and not by errors in the 3D system, such as

positional tracking error, inaccurate system calibration, poor3D TRUS image to world correspondence, etc.

Needle trajectory error NTE was calculated to determinehow well the biopsy needle traveled along the expected bi-opsy needle path pi, defined by the TRUS manufacturer.NTE was measured within the 2D TRUS image plane anddefined as the minimum distance between the actual needletrajectory bTRUS and the expected needle trajectory pi:

NTE =�i=1

n D�pi,biTRUS�

n. �8�

The accuracy of the biopsy system to record biopsy corelocations in 3D �biopsy localization error �BLE�� was quan-tified using three metrics: BLEmin, BLEcenter, and BLE�.BLEmin is defined as the minimum distance between the“true” biopsy core from CT, bi

CT, and the record 3D TRUSbiopsy core, bi

TRUS:

BLEmin =�i=1

n D�biCT,bi

TRUS�n

, �9�

where D� � measures the minimum distance between two 3Dbiopsy core line segments. BLE� is the angle between thecores when projected on a plane perpendicular to BLEmin:

BLE� =�i=1

n ��biCT,bi

TRUS�n

, �10�

where ��u ,v� is the function that calculates the minimumangle between vectors u and v. BLEcenter is the center-to-center distance between corresponding biopsy cores:

BLEcenter =�i=1

n biCT�0.5� − bi

TRUS�0.5� n

, �11�

where b�0.5� represents the center point of the 19 mm biopsycore b.

III.D. Clinical evaluation

The 3D biopsy system was tested clinically on three pa-tients to determine the extent of prostate motion and defor-mation and the system’s impact on workflow. All subjectsprovided written consent to the study protocol approved bythe University of Western Ontario Standing Board on HumanEthics prior to imaging. The workflow of the current 2Dstandard procedure was maintained with the addition of a10 sec 3D US scan at the start and end of the procedure.Anatomical landmarks such as the prostate boundary or cal-cifications within the patient’s prostate were used for quali-tative comparison of the real-time 2D TRUS image with the3D TRUS image to identify correspondence issues resultingfrom prostate deformation, patient motion, or tracking inac-curacies.

An in-plane quantitative assessment of prostate motionand deformation was evaluated by measuring the distancebetween calcifications found in the 2D slices of a 3D TRUSimage, which is generated at the start of the procedure to the
real-time 2D US images obtained at a later time. We defined


the in-plane displacement as the absolute distance betweenthe true location of the landmark and its location within theinitial 3D image.

IV. RESULTS

IV.A. Geometric reconstruction

A 3D TRUS image of the string phantom was successfullyreconstructed without significant visible discontinuity orprobe alignment artifacts. Table I shows the distances be-tween the strings within the phantom measured in the threeorthogonal views �x ,y� , �x ,z� , �y ,z��. The distance mea-sured in each plane overestimated the known manufacturedinterstring distance, with measured errors varying from 3.1%to 4.0% �Table I�.

IV.B. 3D segmentation

Table II shows the results from the scan of the phantomand the segmentations of the 3D object. Multiple 3D TRUSimages were successfully reconstructed from the certified in-dustrial US phantom. The egg-prostate shaped object wassegmented using both the semiautomatic segmentation algo-rithm with a mean volume error of 4.7%. Furthermore, thereconstructed prostate models overlapped with the gold stan-dard volumes with a sensitivity 95.4%, and difference errorof 7.5%.

TABLE II. The volume of the segmented model was compared to the certifiedvolume of the object to quantify the volume measurement error of the 3Dsegmentation. The volume error �Eq. �4�� compares the difference betweenthe reconstructed volume V and the known volume of the scanned objectV*=21.5 cm3. The following volume metrics were used to compare thesemiautomated segmented volume to a manually segmented. The sensitivity�Eq. �5a�� was used to determine the portion of the reconstructed volume Vthat intersects the manually segmented volume Vman, and the difference �Eq.�5b�� measured the proportion of the segmented volumes that is not correctlyoverlapped.

Volume-based metrics

Semiautomatic segmented volume �cm3� 20.49�0.16Manually segmented volume �cm3� 20.83�0.16

Overlap volume �cm3� 19.88�0.26Volume error �%� 4.69

Sensitivity �%� 95.44Difference �%� 7.50

TABLE I. Results from the 3D geometric reconstructistrings, the measurements error � u-u0 �, standard dev

Y-Z plane

y z

Mean �mm� 10.38 10.65 u-u0 0.38 0.35

Error �%� 3.80 3.40STD �mm� 0.11 0.14

n 45 45


IV.C. Needle guidance

The quantitative results for 3D biopsy needle guidanceand recording of biopsy cores are summarized in Table III.Navigation of the biopsy needle to each target had a meanNGE=2.13�1.28 mm. A one-tailed t-test showed theneedle-guidance error was statistically less than 5 mm �p0.001, B=0.2, n=29�, which is the smallest tumor consid-ered clinically significant.36 Figure 8 shows a plot of theNGE, which is decomposed into in- and out-of-plane errors.The needle deflection is primarily in-plane with little out-of-plane error. Human navigation error in directing the probe tothe biopsy target NGHE had a mean error of0.54�0.41 mm. Trajectory error of the biopsy needle devi-ating from the expected needle path NTE was2.08�1.59 mm. Figure 9�a� shows a comparison of the 3Dbiopsy cores for an 18-gauge needle. The biopsy cores over-lapped with each target �white spheres� showing that theneedle was successfully guided to each biopsy target loca-tion. Figure 9�b� shows a qualitative comparison of the 3Dbiopsy cores recorded using the 3D TRUS system �black�with the gold standard cores �white�. The biopsy cores wereaccurately localized to 1.51�0.92 mm �BLEmin� and with amean angulation difference of 6.68�2.23 deg �BLE��. Thecenter-to-center distance �BLEcenter� between correspondingbiopsy cores was larger then BLEmin, at 3.87�1.81 mm.

IV.D. Clinical evaluation

Figure 5 shows axial, sagittal, and coronal views of the3D TRUS image of a patient’s prostate obtained during abiopsy procedure. There were no obvious discontinuities orartifacts within the 3D image, even in the coronal plane,which lies perpendicular to the axis of rotation for 3D scan-ning.

Figure 10 shows a sequence of images that were capturedduring a sextant biopsy at the start �top row�, midpoint�third� biopsy �middle row�, and at the end �bottom row� ofthe procedure. The 2D slice from the 3D static image �Fig.10, left column� is shown with the prostate boundary ingreen. The corresponding 2D real-time video stream �Fig.10, middle column� displays a dotted outline of the prostateboundary from the 3D image. The images in the right col-umn show the real time 2D TRUS images after needle inser-tion �Fig. 10, white arrow�. There is a strong correlation

periment illustrating the mean distance between then �STD�, and the number of data points �n�.

X-Z plane X-Y plane

z x y

.40 10.62 10.37 10.38

.40 0.32 0.37 0.38

.00 3.11 3.70 3.80

.13 0.14 0.14 0.1245 45 45

on exiatio

x

10040

45

between the real-time 2D TRUS image �middle column� and

�.


the original 3D image �left column� at each time point, sug-gesting that there were no major changes in prostate locationand morphology during the biopsy procedure. Alignment be-tween the initial 3D and real-time image was maintainedthroughout the entire procedure, even after six biopsies �bot-tom row�. Similar anatomical landmarks �Fig. 10, calcifica-tions� were visible in both the 3D and real-time 2D images,indicating a good correlation. The mean in-plane displace-ment of calcifications used as anatomical markers identifiedin images obtained from three patients within the prostatewas 0.90�0.97 mm �see Table IV�. The measured in-planedisplacements of calcifications as well as maintaining goodcorrelation between the images in each row suggest that theprostate did not move or deform to a prohibitive extent.

TABLE III. 3D biopsy system accuracy based on the biopsy core analysis descerror, NGHE �Eq. �7��, and needle trajectory error, NTE �Eq. �8��, all evaluaand BLE� �see Eqs. �9�–�11�� indicate the errors in recording the 3D locationof the values are reported as mean�STD. The results from the experiment

Experiment

System accuracy

NGE�mm�

NGHE�mm� �

1 3.79�0.86 0.37�0.51 4.42 0.70�0.32 0.27�0.12 1.43 1.95�0.39 0.19�0.09 1.94 2.21�1.39 0.99�0.08 2.05 1.99�0.77 0.90�0.11 0.5

Mean 2.13�1.28a 0.54�0.41 2.095% CI �1.67, 259� �0.39, 0.69� �1.5

aStatistically less than 5 mm using one-sided t-test �p0.001, B=0.2, n=29

a

FIG. 8. �a� Scatter plot of the in-TRUS plane vs. out-of-TRUS plane needleTRUS to the target and the scatter plot. The scatter plot in is a cross-secticorresponds to the x-axis of the TRUS image �positive and negative x valuesy-axis is parallel to the normal of the 2D TRUS image plane. The origin o

CT
location bi plotted with an “x.” This diagram illustrates the relative orientation

Figure 11 shows a case where the patient had moved dur-ing the procedure. The 2D axial slice from the initial 3Dimage does not correlate well with the 2D TRUS image fromthe US machine. The prostate has moved from its originalscanned position and the shape of the prostate boundary hasalso changed.

V. DISCUSSION

Volume calculations from the ultrasound phantoms werewithin 5% of the certified standard, and the shape-based met-rics �sensitivity=95%, error=4.7%� show a high degree ofoverlap correspondence. Together with the string phantomresults suggest that geometrical distortion due to manual ro-

in Sec. III C. Needle guidance error, NGE �Eq. �6��, needle guidance humanopsy targeting accuracy. The biopsy localization metrics: BLEmin, BLEcenter,he biopsy cores, 95% CI represents the confidence interval of the mean. Allighted in bold are illustrated in Fig. 9.

Localization error metrics

BLEmin

�mm�BLEcenter

�mm�BLE�

�deg�

27 1.93�0.75 3.98�1.45 5.65�2.0446 1.08�0.61 3.82�2.52 5.61�3.3059 2.39�0.98 4.07�1.22 7.76�2.0194 1.20�0.67 4.34�3.17 6.46�2.5551 0.97�0.92 3.13�2.00 7.93�1.2059 1.51�0.92 3.87�1.81 6.68�2.235� �1.18, 1.84� �3.22, 4.52� �5.88, 7.48�

b

ance error NGE. �b� This diagram illustrates the relative orientation of thelane, perpendicular to the 2D TRUS image plane, where the plot’s x-axisore lateral and more medial in the TRUS image, respectively� and the plot’splot is centered on each biopsy target point xi with the “true” biopsy core

ribedte biof t

highl

NTEmm�

1�0.4�0.7�0.6�1.0�0.8�1.1, 2.6

guidonal pare mf the

of the TRUS.


tation of the transducer and semiautomatic segmentation al-gorithm were not significant.

The 3D system was able to accurately guide a biopsyneedle in vitro to predefined targets with an average error of2.1 mm. This error was primarily due to the large NTE as aresult of needle deflection in our prostate phantom.34 In spiteof this error, the targeting accuracy was within the 5 mmradius of the smallest tumors considered clinically significant�tumor radius less than 0.5 cm is considered insignificant�.36

From the results of our ongoing patient study, we wereable to guide the biopsy needle to predetermined targets andmaintain proper correspondence of our real-time 2D and 3Dimages by minimizing prostate motion through a variety ofmechanical and software mechanisms. By providing me-chanical means to stabilize and support the US probe, con-stant pressure between the probe and rectal wall was main-tained. The remote fulcrum �anus� further stabilized thepressure by minimizing motion between the RCM and probetip. Further, constant pressure on the prostate was maintainedwith the probe tip by simply pivoting about the RCM. Tocomplement this mechanical feature, we implemented a vi-sual ring representation in the software interface where the

a b

c

ProstatePhantom

Phantom Box

ApexBase

TRUS

Apex

Base Base

Agar Phantom

CT Core

Target

US Core

CT Core

axis of the needle intersects the depth of the preselected tar-


get �Fig. 5�. This ring provides visual guidance in manipu-lating and orienting the probe to the target in the shortestdistance possible, while minimizing motion of the probe andmaintaining constant pressure on the prostate. The shortestpath is the minimum yaw and pitch angle needed to align theneedle guide to a predetermined target. This is accomplishedby rolling the probe and changing the pitch angle to align theneedle axis with the target. Although minimizing probe mo-tion and its effects on the prostate would be preferred, it isalso important that the physician avoids the urethra andnearby organs �bladder�. This imposes restrictions on the en-try points as it is necessary for the physician to be able to seesensitive areas so they can be avoided.

Patient motion on the biopsy table, which cannot be con-trolled �see Fig. 11� or prostate motion would result in thereal-time 2D image not matching the 2D corresponding slicefrom the 3D TRUS image. This would provide a visual warn-ing to the physician that the preselected targets are no longervalid, requiring the physician to perform another 3D scan.Prostate motion can be corrected by registering the plannedand recorded biopsy locations in the 3D US image to the one

ex FIG. 9. �a� Coronal view of the prostate as viewed fromthe end of the end of the TRUS probe shows the CTgold standard biopsy cores �black cylinders� and thecorresponding biopsy targets �gray dots�. �b� The imageon the right shows the coronal view of the prostatemodel with the US biopsy cores recorded within the 3Dbiopsy system �black� and the corresponding CT goldstandard biopsy cores �white�. The cylinders representthe 18-gauge biopsy core tissue samples. The data high-lighted in Table III were used to construct the phantomimages. �c� Illustration of the simulated prostate orien-tation within the agar phantom with respect to theTRUS probe.

Ap

after the prostate has moved. Although patient motion during


the procedure is infrequent, correction for patient motion er-rors is a focus of our future work.

VI. CONCLUSIONS

By adding 3D information to the prostate biopsy proce-dure, our system should improve the recording procedure aswell as the physician’s ability to accurately guide the biopsyneedle to selected targets identified using other imaging mo-

FIG. 10. A sequence of images captured at the beginning �top row�, midpointThe 2D slice from the 3D static image is shown with the prostate boundarylines show the estimated location of the biopsy core �center column�. The cboundary from the 3D image in the left column. The location of the preplabiopsy notch �19 mm� are designated by the two horizontal lines. Spatial cothe real-time 2D TRUS images �right column�. White speckled calcificationsthe location of the segmented prostate boundary show good correspondethroughout the entire procedure.

TABLE IV. Mean in-plane displacement of anatomical landmarks identifiedin three patients participating in an ongoing clinical trial. The standard de-viation �STD�, maximum, and minimum displacement, and the number oflandmarks �n� identified in the images are reported.

Displacement�mm�

Mean 0.90Maximum 4.60Minimum 0.04

STD 0.97n 21


dalities. This would be beneficial in cases where the patientwas diagnosed on biopsy to have ASAP and requires thephysician to rebiopsy the same area. Using the 3D TRUSimage, the physician was able to observe the patient’s pros-tate in views currently not possible in 2D procedures. Over-all, our 3D system should result in prostate biopsy proce-dures that are stereotactic and more reproducible, which maylead to higher cancer detection rates and improve the yieldon repeat biopsy.

Finally, integration of our system into the current prostatebiopsy procedure requires minimal physician retraining asprocedural workflow is maintained. By adhering to the im-aging tools and protocols of current biopsy procedures, clini-cal integration of our 3D system should be cost effective.

ACKNOWLEDGMENTS

The authors gratefully acknowledge funding from the On-tario Institute of Health Research �CIHR�, and Eigen Medi-cal �13366 Grass Valley, CA 95945� for this work. Specialthanks to Vaishali Karnik for analyzing the patient data and

dle row�, and end �bottom row� of a sextant biopsy procedure �left column�.ented in green. The circle represents the preplanned target, and the parallel

ponding 2D real time video stream displays a dotted outline of the prostatetarget is highlighted by the circle on the 2D image, and the extents of theondence is maintained even after needle insertion �white arrow� as seen inin the prostate boundary �small arrow heads�, cysts �large arrow heads�, andetween the real-time �center column� and static 3D image �left column�

�midsegmorresnnedrrespwith

nce b

Kayley Ma for the many late nights she worked on this


project. JB was supported by a Canada Graduate Scholarshipfrom the Natural Sciences and Engineering Research Councilof Canada. AF holds a Canada Research Chair in BiomedicalEngineering and acknowledges the funding from the CanadaResearch Chair Program.

a�Author to whom correspondence should be addressed. Electronic mail:[email protected]

1C. Mettlin, G. W. Jones, and G. P. Murphy, “Trends in prostate cancercare in the United States, 1974–1990: Observations from the patient careevaluation studies of American College of Surgeons Commission on Can-cer,” Ca-Cancer J. Clin. 43, 83–91 �1993�.

2A. S. Fink, G. Hutarew, B. Esterbauer, A. Pytel, A. Jungwirth, O. Dietze,and N. T. Schmeller, “Evaluation of transition zone and lateral sextantbiopsies for prostate cancer detection after initial sextant biopsy,” Urol-ogy 61�4�, 748–753 �2003�.

3W. G. Nelson, A. M. De Marzo, and W. B. Isaacs, “Review article: Pros-tate cancer,” N. Engl. J. Med. 349, 366–381 �2003�.

4J. A. Eastham and P. T. Scardino, “Early diagnosis and treatment of pros-tate cancer,” DM, Dis.-Mon. 47, 421–459 �2001�.

5L. A. Eskew, R. L. Bare, and D. L. McCullough, “Systematic 5 regionprostate biopsy is superior to sextant method for diagnosing carcinoma ofthe prostate,” J. Urol. �Baltimore� 157�1�, 199–203 �1997�.

6B. Djavan, B. Remzi, C. C. Schulman, M. Marberger, and A. R. Zlotta,“Repeat prostate biopsy: Who, how, and when?,” Eur. Urol. 42�2�, 93–103 �2002�.

7J. C. Presti, Jr., G. J. O’Dowd, M. C. Miller, R. Mattu, and R. W. Veltri,“Extended peripheral zone biopsy schemes increase cancer detection ratesand minimize variance in prostate specific antigen and age related cancerrates: Results of a community multi-practice study,” J. Urol. �Baltimore�169�1�, 125–129 �2003�.

8B. Djavan et al., “Is one set of sextant biopsies enough to rule out prostateCancer? Influence of transition and total prostate volumes on prostatecancer yield,” Eur. Urol. 38�2�, 218–224 �2000�.

9P. I. Karakiewicz et al., “Outcome of sextant biopsy according to glandvolume,” Urology 49�1�, 55–59 �1997�.

10D. W. Keetch, W. J. Catalona, and D. S. Smith, “Serial prostatic biopsiesin men with persistently elevated serum prostate specific antigen values,”J. Urol. �Baltimore� 151�6�, 1571–1574 �1994�.

11S. J. Park, H. Miyake, I. Hara, and H. Eto, “Predictors of prostate canceron repeat transrectal ultrasound-guided systematic prostate biopsy,” Int. J.Urol. 10�2�, 68–71 �2003�.

12R. Kranse, P. Beemsterboer, J. Rietbergen, D. Habbema, J. Hugosson, andF. H. Schroder, “Predictors for biopsy outcome in the European random-ized study of screening for prostate cancer �Rotterdam region�,” Prostate39�4�, 316–322 �1999�.

13D. E. Coplen, G. L. Andriole, J. J. Yuan, and W. J. Catalona, “The abilityof systematic transrectal ultrasound guided biopsy to detect prostate can-cer in men with the clinical diagnosis of benign prostatic hyperplasia,” J.Urol. �Baltimore� 146�1�, 75–77 �1991�.

14K. A. Iczkowski, H. M. Chen, X. J. Yang, and R. A. Beach, “Prostatecancer diagnosed after initial biopsy with atypical small acinar prolifera-

Bladder

Ureth

ConcaveBoundary

ConvexBoundary

tion suspicious for malignancy is similar to cancer found on initial bi-


opsy,” Urology 60�5�, 851–854 �2002�.15K. A. Iczkowski et al., “Diagnosis of ‘suspicious for malignancy’ in pros-

tate biopsies: Predictive value for cancer,” Urology 51�5�, 749–757�1998�.

16I. San Francisco, W. DeWolf, S. Rosen, M. Upton, and A. Olumi, “Ex-tended prostate needle biopsy improves concordance of Gleason gradingbetween prostate needle biopsy and radical prostatectomy,” Urology 169,136–140 �2003�.

17P. Thorson and P. A. Humphrey, “Minimal adenocarcinoma in prostateneedle biopsy tissue,” Am. J. Clin. Pathol. 114�6�, 896–909 �2000�.

18G. Andriole, T. Bullock, J. Belani, E. Traxel, Y. Yan, D. Bostwick, and P.Humphrey, “Is there a better way to biopsy the prostate? Prospects for anovel transrectal systematic biopsy approach,” Urology 70�6�, 22–26�2007�.

19J. A. Long, V. Daanen, A. Moreau-Gaudry, J. Troccaz, J. J. Rambeaud,and J. L. Descotes, “Prostate biopsies guided by three-dimensional real-time �4-D� transrectal ultrasonography on a phantom: Comparitativestudy versus two dimensional transrectal ultrasound-guided biopsies,”Eur. Urol. 52, 1097–1105 �2007�.

20F. T. Abul, N. Arun, Mona A. Abu-Assi, and A. M. Asbeutah, “Transrec-tal ultrasound guided biopsy for detecting prostate cancer: Can randombiopsies be reduced using the 4-dimensional technique?,” Int. Urol. Neph-rol. 39, 517–524 �2007�.

21I. Megwalu, G. Ferguson, J. Wei, V. Mouraviev, T. Polascik, S. Taneja, L.Black, G. Andriole, and A. Kibel, “Evaluation of a novel precisiontemplate-guided biopsy system for detecting prostate cancer,” Br. J. Urol.102, 546–550 �2008�.

22H. Hricak, P. Choyke, S. Eberhardt, S. Leibel, and P. Scardino, “Imagingprostate cancer: A multidisciplinary perspective,” Radiology 243�1�,28–53 �2007�.

23G. Fichtinger, T. L. DeWeese, A. Patriciu, A. Tanacs, D. Mazilu, J. H.Anderson, K. Masamune, R. H. Taylor, and D. Stoianovici, “System forrobotically assisted prostate biopsy and therapy with intraoperative ctguidance,” Acad. Radiol. 9�1�, 60–74 �2002�.

24A. Krieger, R. C. Susil, C. Menard, J. A. Coleman, G. Fichtinger, E.Atalar, and L. L. Whitcomb, “Design of a novel MRI compatible manipu-lator for image guided prostate interventions,” IEEE Trans. Biomed. Eng.52�2�, 306–313 �2005�.

25A. S. Barnes, S. J. Haker, R. V. Mulkern, M. So, A. V. D’Amico, and C.M. Tempany, “Magnetic resonance spectroscopy-guided transperinealprostate biopsy and brachytherapy for recurrent prostate cancer,” Urology66�6�, 131–193 �2005�.

26American Cancer Society. “Detailed guide: Prostate cancer.” http://wwwcancer.org/docroot/CRI/CRI�2�3x.asp?dt 36 �accessed November7, 2007�.

27R. Taylor, J. Funda, B. Eldridge, K. Gruben, D. LaRose, S. Gomory, M.talamini, L. Kavoussi, and J. Anderson, “A telerobotic assistant for lapro-scopic surgery,” IEEE Eng. Med. Biol. Mag. 14, 539–548 �1995�.

28J. Bax, D. Cool, L. Gardi, and A. Fenster, “Apparatus for guiding amedical tool,” International Application under the Patent CooperationTreaty �PCT� WO 2007/147232 A1 �2007�.

29J. Bax, D. Cool, L. Gardi, J. Montreuil, E. Gil, J. Bluvol, K. Knight, D.Smith, C. Romagnoli, and A. Fenster, “3D transrectal ultrasound prostate

FIG. 11. Comparison of �a� the 2D axial slice of the 3DTRUS image with �b� the real-time 2D TRUS imagecaptured immediately after the biopsy gun was fired;the needle is still visible in the image on the right �ar-row�. If the patient moves during the procedure, notonly does the prostate move from its original scannedposition, but also its shape has changed. The prostateboundary next to the TRUS tip is concave in the 2Daxial slice, and convex in the real-time snap shot. Thediscontinuity in the 3D image highlighted by the arrowheads was the result of patient motion during a 3D scan.The bladder �hypoechoic region anterior to the prostateboundary� and urethra is visible in each image.

ra

biopsy using a mechanical imaging and needle-guidance system,” Medi-

http://dx.doi.org/10.1016/S0090-4295(02)02502-5

http://dx.doi.org/10.1016/S0090-4295(02)02502-5

http://dx.doi.org/10.1016/S0022-5347(01)65322-9

http://dx.doi.org/10.1016/j.urology.2007.06.1128

http://dx.doi.org/10.1148/radiol.2431030580

http://dx.doi.org/10.1109/TBME.2004.840497


cal Imaging 2008: Visualization, Image-Guided Procedures, and Model-ing, M. Miga, and K. Cleary, eds., Proc. SPIE Vol. 6918�25�, 1–11 �2008�.

30A. Fenster, D. Downey, and N. Cardinal, “Three dimensional ultrasoundimaging,” Phys. Med. Biol. 46�5�, R67–R99 �2001�.

31H. Ladak, F. Mao, Y. Wang, D. Downey, D. Steinman, and A. Fenster,“Prostate boundary segmentation from 2D ultrasound images,” Med.Phys. 27�8�, 1777–1788 �2000�.

32Y. Wang, N. Cardinal, D. Downey, and A. Fenster, “Semi-automatic 3Dsegmentation of the prostate using 2D ultrasound images,” Med. Phys.30�5�, 887–897 �2003�.

33Z. Wei, G. Wan, L. Gardi, G. Mills, D. Downey, and A. Fenster, “Robot


assisted 3D TRUS guided prostate brachytherapy,” Med. Phys. 31�3�,539–548 �2004�.

34D. Cool, S. Sherebrin, J. Izawa, J. Chin, and A. Fenster, “Design andevaluation of a 3D transrectal ultrasound prostate biopsy system,” Med.Phys. 35, 4695–4707 �2008�.

35D. W. Rickey, P. A. Picot, D. A. Christopher, and A. Fenster, “A wall-lessvessel phantom for doppler ultrasound studies,” Ultrasound Med. Biol.21�9�, 1163–1176 �1995�.

36J. I. Epsein, H. Sanerson, H. B. Carter, and D. O. Scharfstein, “Utility ofsaturation biopsy to predict insignificant cancer at radical prostatectomy,”Urology 66�2�, 356–360 �2005�.

http://dx.doi.org/10.1088/0031-9155/46/5/201

http://dx.doi.org/10.1118/1.1286722

http://dx.doi.org/10.1118/1.1286722

http://dx.doi.org/10.1118/1.1568975

http://dx.doi.org/10.1118/1.1645680

http://dx.doi.org/10.1118/1.2977542

http://dx.doi.org/10.1118/1.2977542

http://dx.doi.org/10.1016/0301-5629(95)00044-5

mechanically assisted 3d ultrasound guided prostate biopsy ......blood tests are common screening...

Documents