Computer Aided Surgery 5:359-370 (2000)

Abstracts from ISRACAS 2000 Third Israeli Symposium on Computer-Aided Surgery,

Medical Robotics, and Medical Imaging

We are pleased to present selected abstracts of the Third Israeli Symposium on Computer-Aided Surgery, Medical Robotics, and Medical Imaging (ISRACAS 2000), which was held in Haifa, Israel, on May 18, 2000. The meeting was jointly sponsored by the Belfer Symposium Fund, the Technion-Israel Institute of Technology, The Hebrew University of Jerusalem, GE Medical Systems Israel (Centers of Excellence for Magnetic Resonance, Nuclear Medicine, Ultrasound and Intraoperative Imaging), and Silicon Graphics Israel. It was endorsed by the International Society for Computer Aided Surgery (ISCAS) and the Israeli Society for Medical and Biolog- ical Engineering.

The goal of the Symposium was to convene in Israel clinicians, scientists and engineers actively interested in medical imaging, computer science, and robotics, and their application to the planning, monitoring and execution of medical surgeries. The symposium is the sequel to the two previous ones, which were held on May 4 1998 at the Technion, and on May 5, 1999 at the Hebrew University of Jerusalem. Each was attended by over 150 participants, mostly from Israel, evenly distributed between clinicians, industry, and engineering academia. The one-day events included four to six invited speakers, 15 oral presentations of refereed papers, industrial exhibits, and system demonstrations.

This year’s Symposium included five invited talks by internationally recognized experts, 14 oral paper presentations, an industrial perspectives session, and several system demonstrations. The papers were selected after review by the Local Program Committee and International Advisory Board.

Leo Joskowicz Moshe Shoham

Jerusalem and Haifa, May 2000

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

360 Abstramfim ISRACAS 2000

EVOLUTION AND FUTURE O F COMPUTER-ASSISTED NEUROSURGERY (Invited talk)

Prof. Gene M. Barnett, MD Department of Neurological Surgery, The Cleveland Clinic Foun- dation, Cleveland, OH, USA

The roots of computer-assisted neurosurgery reside in frame ste- reotaxy, a methodology that was introduced by Horsley and Clark in 1908 [l]. This landmark publication introduced the notion that specific locations in the brain could be accessed precisely using a three-dimensional (3D) coordinate system. Although their frame was designed for animals, a frame for human use was first reported by Spiegel and Wycis [2] almost 40 years later. Multiple frames were then devised, and frame stereotaxy became an important method of treating movement disorders such as Parkinson’s dis- ease. Targeting was performed using radiographs showing land- marks in the ventricular system of the brain, and correlating these with human atlases.

Frame stereotaxy almost became extinct in the 1960s. however, with the development of medical treatment for functional disorders. After a hiatus of a decade, a renaissance of stereotactic techniques emerged in the late 1970s when it became possible to extract spatial information from CT scans that could be used to direct a stereo- tactic apparatus [3,4]. More imaging modalities such as MRI and angiography became applicable, and a broad array of diagnostic and interventional procedures were developed or refined using com- puter image guidance, such as biopsy, minimal access craniotomy, depth electrodes, third ventriculostomy and radiosurgery [5,6,7]. The standards of morbidity and success of stereotactic techniques were defined during this period.

The transition from point to volumetric stereotaxy allowed de- velopment of those surgical techniques that would prove necessary to best use the next revolution in computer-assisted neurosurgery - the surgical navigation system. The availability of accurate 3D digitizers, fast imaging computers, and spatially accurate imaging came together in the mid-1980s and led to systems that provide real-time localization, orientation and guidance using preoperative imaging [8-111. They do so by correlating (“registering”) a volume of image data with the location of the patient in the operating room as determined by some type of 3D digitizer that also serves as a pointing device. A wide variety of tools, including operating mi- croscopes, endoscopes and drills, may be used in this fashion [12,13].

Today, these systems have emerged as the standard of care for many neurosurgical procedures, both in the head and spine [ 14-20], They have been shown to have success and complication rates equal or superior to frame procedures, and are more cost-effective than conventional surgery in some cases [21-231. They also provide uniquely useful views of functional anatomy and surface veins. Problems remain, however in constraints posed by digitizer tech- nologies (e.g., “line-of-sight”), registration procedures, image dis- tortion and, in particular, movement and distortion of the brain during surgery that renders the preoperative image data inaccurate.

Today, development focuses on methodology to provide intra- operative updates of imaging information, acquisition and interpre- tation of data, and use as part of effector systems ( i t . , robotics). Navigation systems may use multiple image data sets, co-registered (‘‘fused”) to provide more information than may be gleaned from one type of image set. For instance, the fine anatomic detail of volume MRI may not show low-grade tumors well, while fluid attenuation inversion recovery sequences show tissue changes as- sociated with a tumor, but detail is lacking and applied reference marks are not visualized. Combining these images provides for more robust navigation.

Similarly, coupling an ultrasound probe to a navigation system allows for generation of a 3D data set that is correlated with preoperative imaging. Work in our laboratory has shown this to be both feasible and useful. The next big advance in intraoperative navigation, however, is likely to be the use of a new generation of

miniature, low-field MRI devices. The first of these uses permanent ceramic magnets in a “C-arm” arrangement to provide intraopera- tive imaging (Polestar N- 10. Odin Medical Technologies, Israel), albeit with a restricted field of view. Integrated navigation uses either MRI technology or conventional passive infrared, and image output could be coregistered with preoperative imaging as de- scribed above.

Beyond imaging, navigation systems will coregister macroscopic atlas data (matched to the patient using deformation techniques) such as the Visible Human Data Set. By defining anatomy as objects, the identity of a structure can be immediately identified and myriad resources accessed, such as function, connections, vascular supply, surgical approaches and so on. Intraoperatively obtained physiologic data will also be archived and analyzed at central locations connected by the Internet. As such, robust physiologic and anatomic databases will be created with information at both the macro and microscopic level. This library of data will be used for clinical navigation as well as instruction.

Lastly, these platforms will be further integrated with robotic effectors that go beyond the simple guidance robots of today. These systems will be used to enhance dexterity in a “smart” fashion, knowing the nature and structural characteristics of tissues in the surgical field. Using magnetic manipulation, catheters and capsules of therapeutic agents (genes, growth factors, and other implants) will be directed through the brain and its vessels, augmenting the linear stereotactic and endovascular techniques of today.

Ultimately, these systems are just tools that augment, not replace, the surgeon’s judgment. Nonetheless, the “next level” of surgical navigation promises powerful enhancements to an already powerful tool.

References 1. Horsley V, Clarke RH. The structure and function of the

cerebellum examined by a new method. BRAIN 1908;31:45-124. 2. Spiegel EA, Wycis T. Stereoencephalotomy Part 1: Methods

and stereotaxic atlas of the human brain. Orlando, FL: Grune & Stratton; 1952.

3. Brown RA, Roberts TS, Osborn AG. Stereotaxic frame and computer software for CT-directed neurosurgical localization. In- vest Radio1 1980;15:308-312.

4. Leksell L. The stereotaxic method and radiosurgery of the brain. Acta Chir Scan 1951;102:316-319.

5. Kelly PJ. Applications and methodology for contemporary stereotactic surgery. Neurological Res 1986;8:2- 12.

6. Gomez H, Barnett GH, Estes ML, Palmer J , Magdinec M. Stereotactic and computer-assisted neurosurgery at the Cleveland Clinic. Cleveland Clin J Medicine 1993;60:399-410.

7. Leksell L. Stereotactic radiosurgery. J Neurol Neurosurg Psychiatry 1983;46:797-803.

8. Roberts DW, Strohbehn JW. Hatch JF, Murray W, Ketten- berger H. A frameless stereotaxic integration of computerized to- mographic imaging and the operating microscope. J Neurosurg

9. Watanabe E, Watanabe T, Manaka S, Mayanagi Y, Takakura K. Three-dimensional digitizer: new equipment for computed to- mography-guided stereotaxic surgery. Surg Neurol 1987;27:543- 547.

10. Barnett GH, Kormos DW, Steiner CP. Weisenberger J. In- traoperative localization using an armless, frameless stereotactic wand. J Neurosurg 1993;78:5 10-5 14.

11. Barnett GH, Kormos DW, Steiner CP, Weisenberger J. Use of a frameless, armless stereotactic wand for brain tumor localiza- tion with 2D and 3D neuroimaging. Neurosurgery 1993;33:674- 678.

12. Rhoten RL, Luciano MG, Barnett GH. Computer-assisted endoscopy for neurosurgical procedures: Neurosurgery 1997;40: 632-637 (discussion 638).

13. Luciano MG, Rhoten RLP, Barnett GH. Hydrocephalus. In: Barnett GH, Roberts DW, Maciunas RJ, editors: Image-Guided Neurosurgery. Clinical Applications of Surgical Navigation. St. Louis, MO: Quality Medical Publishing, Inc.; 1998.

1986;65:545-549.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

Abm-acisp-m ISRACAS 2000 361

14. Barnett GH, Kaakaji W. Intracranial meningiomas. In: Bar- nett GH, Roberts DW, Maciunas RJ, editors: Image-Guided Neu- rosurgery. Clinical Applications of Surgical Navigation. St. Louis, MO: Quality Medical Publishing, Inc.; 1998. p 87-100.

15. Barnett GH, Miller DW, Weisenberger J. Brain biopsy using frameless stereotaxy with scalp applied fiducials: Experience in 21 8 cases. J Neurosurg 1999;91:569-576.

16. Barnett GH, Miller DW. Brain biopsy and related proce- dures. In: Barnett GH, Roberts DW, Maciunas RJ, editors: Image- Guided Neurosurgery. Clinical Applications of Surgical Naviga- tion. St. Louis, MO: Quality Medical Publishing, Inc.; 1998. p

17. Barnett GH, Steiner CP, and Weisenberger J . Target and trajectory guidance for interactive surgical navigation systems. Stereotact Funct Neurosurg 1996;66:91-95.

18. Barnett GH. Steiner CP, Kormos DW, Weisenberger J. In- tracranial meningioma resection using interactive frameless ste- reotaxy-assistance. J Image Guided Surg 1995; 1:46-52.

19. Barnett GH. Transsphenoidal hypophysectomy. In: Barnett GH, Roberts DW, Maciunas RJ, editors: Image-Guided Neurosur- gery. Clinical Applications of Surgical Navigation. St. Louis, MO: Quality Medical Publishing, Inc.; 1998. p 101-104.

20. Kalfas IH, Kormos DW, Murphy MA, McKenzie RL, Bar- nett GH, Bell GR, Steiner P, Trimble MB, Weisenberger JP. Ap- plication of frameless stereotaxy to pedicle screw fixation of the spine. J Neurosurg 1995;83:64-647.

21. Barnett GH, Walsh JG, Steiner CP, Weisenberger JP. One- year outcome data after resection of malignant gliorna. In: Barnett GH. Roberts DW, Maciunas RJ, editors: Image-Guided Neurosur- gery. Clinical Applications of Surgical Navigation. St. Louis, MO: Quality Medical Publishing, Inc.; 1998. p 251-257.

22. Bingaman WE, Barnett GH. Social and economic impact of

nas RJ, editors: Image-Guided Neurosurgery. Clinical Applications of Surgical Navigation. St. Louis, MO: Quality Medical Publishing, Inc.; 1998. p 231-249.

23. Murphy MA, Barnett GH, Kormos DW, Weisenberger J. Astrocytoma resection using an interactive frameless stereotactic wand. An early experience. J Clin Neurosci 1994;1:33-37.

181-191.

surgical nmigarion sy3Em flr-*mm&-

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

362 Abstt-amfrom ISRACAS 2000

SESSION 1: NEUROSURGERY AND NEUROLOGY

A DIGITAL BRAIN ATLAS - EVALUATION BY FUNCTIONAL MRI

H. Dickhaus’, K.A. Ganser’, A. Staubert’, R. MetzneP, C.R. Wirtz*, M.M. Bonsanto’, V.M. Tronnie?, and S. Kunze’ ’Department of Medical Informatics, University of Heidelberg, Heilbronn University of Applied Sciences, Max-Planck-Str. 39, D 7408 1 Heilbronn, Germany; ’Department of Neurosurgery, Univer- sity of Heidelberg; ’Department of Radiology, German Cancer Research Center, Heidelberg, Germany E-mail: dickhaus@fh-heilbronn.de

Abstract: Due to the requirement for high precision in neurosur- gical interventions, setting up a surgical strategy is indispensable. Valuable tools for the planning procedure are brain atlases, which assist the neurosurgeon in interpreting the three-dimensional, pre- operatively acquired MRI datasets of the patients. To improve the inconvenient handling of printed atlas books, we have developed a digital version of the well-established stereotactic brain atlas of Talairach and Tournoux. Our implementation has several advan- tages in comparison with the original book. For example, the computerized atlas system can be matched with MRI and functional MRI data of patients. By means of this feature, we evaluate the accuracy of the atlas matching in the cortical region, especially at the primary motor cortex.

Methodology: We digitized the well-established stereotactic at- las of Talairach and Toumoux [ I ] and calculated a 3D reconstruc- tion of the atlas structures. We included these data in a computer- ized atlas system that allows matching of the atlas with individual MR images of patients (DICOM file format). The matching proce- dure follows the paradigm of the so-called “proportional grid”, which is piecewise linear; it is supported by a convenient user interface. The atlas structures can be displayed as 3D surface models along with orthogonal cross-sections through the MR image stack. If an MR slice intersects with an atlas structure, the inter- section contour can alternatively be drawn onto the grey-value image. Furthermore, the original atlas plates contained in the printed atlas book may be transparently inserted at their appropriate spatial location in the scene [Z]. It is possible to match the atlas with functional MR images (tMRI) as well. These images give informa- tion about the location of certain functional brain regions. These regions, visible as hyperintensive spots in the fMRI, can be com- pared with the corresponding cortical regions in the atlas, which are delimited by means of the Brodmann areas.

In particular, we measured the distances between the primary motor cortices in the atlas (Brodmann area 4) and in fMR images (indicated as active spots on the motor cortex caused by finger tapping). Additionally, we evaluated the congruence of the shapes of the motor cortices in both modalities applying a three-scored scale.

Results: We have developed a computerized atlas system for Windows NT which allows convenient application of the Talairach atlas to MR images of patients. It offers several visualization options and is suitable for assisting the neurosurgeon in the preop- erative planning phase. We evaluated the matching accuracy in the cortical region by considering the primary motor cortex, which is located on the gyrus praecentralis. The estimated distances showed an RMS of about 5 mm, while their mean was about 0 mm. This means that the observed differences are mainly due to anatomical variations. The congruence was rated as “good“ or “very good” in 73% of the cases.

References I . Talairach J, Tournoux P. Co-Planar Stereotaxic Atlas of the

Human Brain. Stuttgart: Georg Thieme Verlag; 1988. 2. Dickhaus H, Ganser KA, Staubert A, et al. Three dimensional

reconstruction of the stereotactic atlas of Talairach and Toumoux for neurosurgical planning. To appear in IEEE Transactions on Information Technology in Biomedicine 2000.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

Abstructsfi-om ISRAGlS 2000 363

CLINICAL EXPERIENCE WITH ROUTINE APPLICATION OF FRAMELESS SURGICAL NAVIGATION SYSTEM IN 60 NEUROSURGICAL PROCEDURES

Yigal Shoshan, MD, Guy Rosenthal, MD, Shmuel Igus. MD. Mufid Yaakub, MD, Eyal Itshayek, MD, Samuel Tobias, MD, and Felix Umansky, MD Department of Neurosurgery, Hadassah University Hospital, Jerusalem, Israel E-mail: ys@md2.huji.ac.il

The recent availability of accurate 3-D imaging, accurate 3-D digitizers, and, above all, fast, inexpensive computers permitted the development of a Frameless Surgical Navigation System (FSNS). This technology enables real-time presentation of localization, ori- entation, and guidance data during surgery. FSNS has been utilized at the Hadassah University Hospital on a routine basis in neuro- surgical procedures since August 1999.

Material and Methods: Fifty-eight consecutive patients with various neurosurgical pathologies underwent 60 procedures with the assistance of the FSNS (Stealth StationTM) between August 1999 and April 2000. The FSNS was operated with optical digitizer technology and scalp-applied fiducial markers for registering the image coordinate system with the operating room space. System capabilities include real-time display of tool location, orientation, and relationship to nearby structures using multiplanar 3D presen- tation and oblique views of MRI and/or CT data obtained preop- eratively.

Results: Twenty-four craniotomies (40%) were performed for

astatic tumors (2 pts.), and radiation necrosis (2 pts.). One patient each underwent resection of a cavemoma, a histiocytic lesion, and extraction of a bullet from the brain. Minimal-access craniotomy was obtained in all patients in this group. Macroscopic resection (> 95%) was achieved in 19 patients (79%). Subtotal resection (80%- 95%) was performed in 5 patients (21%). Surgery-related compli- cations included rebleeding into the operative bed following re- moval of a metastatic melanoma that required re-operation. One patient suffered a wound infection. Frameless stereotactic brain biopsy was carried out in 21 patients (35%) for the diagnosis of glial tumors (I4 pts.), metastatic tumors (3 pts.), radiation necrosis (1 pt.). lymphoma ( 1 pt.) and leukoencephalopathy (1 pt.). Histo- logical diagnosis was achieved in 20 patients (95%). One biopsy was non-diagnostic (5%). There were no permanent biopsy-related complications. No patient had evidence of intracerebral bleeding on immediate post-biopsy CT scan. FSNS was also used for minimally invasive insertion of an Ommaya reservoir catheter into the lateral ventricles in ten lymphoma patients (17%). Accurate catheter im- plantation was obtained in all 10 patients, as confirmed by imme- diate postoperative CT scan, in a single brain pass. There were also three spinal surgeries, performed for pedicular screw fixation, re- moval of small pedicular osteoblastoma, and C1-2 fixation. Two patients underwent transsphenoidal surgery. The average postoper- ative hospital stay was 4.1, 2.1, and 1.5 days for craniotomy, biopsy, and the Ommaya group respectively, compared to 8.5 hospitalization days among patients undergoing traditional neuro- surgery in our department. Subjectively, there was a high level of patient satisfaction with the relatively small surgical scar and min- imal hair shaving.

Conclusions: The acquisition of an image-guided neurosurgery system in the Department of Neurosurgery at Hadassah University Hospital has been successfully completed. It has been utilized routinely by most staff, residents, and fellows in various neurosur- gical procedures. Frameless stereotactic craniotomy, biopsy and ventricular catheterization appear to be safe and effective for treat- ing a wide spectrum of neurosurgical lesions, with accurate regis- tration provided by scalp-applied fiducial markers. Application of

-- ' mm-(-l2-@- ' ' m ~ ~

FSNS significantly reduces postoperative hospital stay and mark- edly increases subjective patient comfort following neurosurgical operations.

THE VISION OF THE OPERATING ROOM O F THE FUTURE (Invited talk)

Prof. Ferenc Jolesz, MD Director, Division of MRI, and Director, Image-Guided Therapy Program, Department of Radiology, Brigham and Women's Hos- pital, Harvard Medical School, Boston, MA 02115, USA

There is now an ongoing effort to develop and implement MRI- guided therapy for interventional radiology and surgical applica- tions. The current progress does not represent a significant para- digm-shift from the original concept of image-guided therapy. Nevertheless, there are new, potentially revolutionary methods which have enormous potential for a series of breakthrough appli- cations. Among these, the most important are MRI-guided thermal ablations (laser, RF, cryo- and focused ultrasound ablations), intra- operative MR imaging-based open surgeries and endoscopies. So far. neurosurgery (including brain and spine applications) has ben- efited most from intraoperative MRI guidance. This relatively fast success is due to the fact that frameless stereotaxy. navigational tools, and multimodality image-fusion are already accepted in neu- rosurgery, and the introduction of real-time image updates has improved localization and targeting as well as the completeness of tumor resections. This early success of MRI-guided therapy in the field of neurosurgery should be extended to other fields. New applications have been described for tumor treatment in the breast, prostate and liver, and the preliminary results are encouraging.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

364 Abstractsfiom ISRACAS 2000

SESSION 2 COMPUTER-AIDED ORTHOPAEDIC SURGERY

COMPUTER ASSISTED SPINE SURGERY (Invited talk)

Prof. Phillipe Merloz Univ. Dept. of Orthopaedic Surgery, CHU A. Michallon; BP 217; 38043, Grenoble Cedex 09, France

The use of clinical systems in image-guided spinal surgery has many applications in which it is necessary to reach a deep target that cannot be visualized directly, or to increase the accuracy of the operative procedure. Provided that a CT scan of the patient with numerous thin slices can be acquired, it is possible to define a linear trajectory on CT images and drill a hole exactly where it has been planned. Transpedicle instrumentation has been used more often in the lumbar andor thoracic spines of adults for rigid segmental fixation after decompression and arthrodesis for various disorders, including scoliosis, spondylolisthesis, and iatrogenic or degenera- tive instability. During a posterior approach to the spine, the back part of the vertebra is exposed and the surgeon's anatomical knowl- edge guides the drilling direction.

Pilot holes are prepared and screws are inserted into the pedicle without any direct visual control. A slight error in direction may result in a significant error in the position of the tip of the screw. Drilling is performed at least twice, and sometimes 3 or 4 times, during surgery, with no direct visibility of the crucial structures (spinal cord, lungs, vessels and nerves). To check the correct placement of the screw in the pedicle. image intensification is used, but, because of radiation exposure to the patient and the difficulty of using this system, it cannot be. applied during the entire screw insertion procedure. The variability in width, height and spatial orientation of spinal pedicles, especially in the surgical procedure for scoliosis, consequently leads to a considerable rate of misplaced screws. Biomechanical studies suggest an optimal position of the screw tip to be ai a near to the anterior cortex of the vertebral body as possible. A correct pedicle screw insertion has to take into account 2 parameters: the spatial orientation of the screw and its length.

To look for the correct spatial orientation on the one hand and the correct length of the pedicle screw on the other hand, and, more generally, to look for a crucial anatomical structure without any direct visual control, we developed a novel technique that combines preoperative tomographic imaging with principles of stereotaxis, because it is important to increase safety by more precise interven- tion. This technique follows a general tendency of computer-as- sisted medical interventions. A passive system with only a 3D optical localizer can be used during surgery for both registration and guidance.

The purpose of the method is to reduce complications, which proves to be feasible according to the first clinical results obtained in the lumbar and thoracic regions. There is currently no limitation for using the method in the lower back part of the spine, and osteosynthesis could therefore be performed at higher levels, in- cluding cervical vertebrae, with a high level of confidence. The method may be extended to removal of intervertebral disk, biopsy of the cancellous bone of the vertebral body for tumors, or, more generally, to reaching a deep target that cannot be visualized directly.

In the future, we believe it will be possible to use such a system percutaneously. The acquisition of radiographic or ultra-sound im- ages obtained in calibrated conditions makes it possible to register preoperative CT images with the surgical space percutaneously. Such work may contribute to the reduction of invasiveness in orthopedic surgery.

PLANNING OF TOTAL KNEE REPLACEMENT: ANALYSIS OF THE CRITICAL PARAMETERS INFLUENCING THE IMPLANT

M. Marcacci, MD, S. Martelli, L. Nofrini, F. La Palombara, and F. Iacono, MD Istituti Ortopedici Rizzoli, Lab. Biomeccanica, Bologna, Italy

This paper describes planning software developed for computer- assisted Total Knee Replacement (TKR), aimed at improving the accuracy and reliability of the preoperative phase of the interven- tion and therefore resulting in fewer surgical mistakes and adjust- ments.

The software allows the surgeon to simulate the implant on a 3D model of the patient's limb, computed from C T images, and con- sists of two steps. First, the surgeon interactively identifies the limb mechanical axis and articular line, setting the position of the hip, knee and ankle centers on frontal and lateral scout views of the leg. The surgeon identifies anatomical characteristics, and the system automatically computes the position/orientation of the implant in order to restore the mechanical axis, minimize bone loss, and preserve the articular line. The surgeon can then interactively adjust the position of the prosthesis component on sections or 2D projec- tions.

The system was repeatedly used by 3 surgeons on 10 patients to analyze the repeatability of the surgeons' planning, the efficacy of the automatic computations, and the influence of the different steps of the planning process on the result. The computation of anatom- ical parameters from the user's defined features resulted in a final mean repeatability of the implant of 1" and 2 mm (the tibia1 component may occasionally show 2.5" and 3 mm). while manual repeatability was around 6" and 5 mm.

New information, such as cross-sections and accurate sizes and angles in the patient's images, were considered by all the surgeons to be a very useful tool for analyzing the patient's anatomy, and provided deep knowledge of the global effects of the implant, the amount of resected bone, and the contact areas between the pros- thesis component and the bone.

The computer planning was considered useful and easy to use by surgeons: on average, initial tests and 6 trials were necessary to become fully acquainted with all the program features and to become confident in the final result, even for non-expert surgeons.

The use of accurate images and computer assistance for planning TKR has shown great potentiality and satisfactory accuracy for improving the surgeons' insight and confidence, and provides a new protocol for TKR planning.

1. Moreland JR. Mechanism of failure in Total Knee Arthro- plasty. Clin Orthop Re1 Res 1988 ;226:49-64.

2. Fadda M, Bertelli D, Martelli S, Marcacci M, Dario P, Pag- getti C, Caramella D, Trippi D. Computer assisted planning for Total Knee Arthroplasty. In Computer Science Notes 1205, Pro- ceedings of CVRMed-MRCAS'97. Grenoble: Springer; 1997, p

3. Bargar WL, Bauer A, Borner M. Primary and revision total hip replacement using the Robodoc system. Clin Orthop Re1 Res

4. Smith SL, Mosier JN. Guidelines for designing user interface

620-628.

1998;354:82-91.

software. Edited by MITRE Corporation, 1986.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

Abstrmixfrorn ZSRACAS 2000 365

ROBOT-ASSISTED REGISTRATION FOR TOTAL KNEE REPLACEMENT

Daniel Glozman, Moshe Shoham and Anath Fischer Robotics Laboratory, Department of Mechanical Engineering, Technion-Israel Institute of Technology, Haifa, 32000, Israel

Successful implementation of robot-assisted surgery (RAS) re- quires coherent integration of spatial image data with sensing and actuating devices, each having its own coordinate system. Hence, accurate estimation of the geometric relationships between relevant reference frames, referred to as registration, is a crucial procedure in all RAS applications. The purpose of this article is to present a new registration scheme and experimental results of a robot-as- sisted registration method for RAS applications in orthopedics.

The registration method is based on a surface-matching algo- rithm that does not require marker implants, hence reducing surgi- cal invasiveness. Points on the bone surface are sampled by the robot, which in turn directs the surgical tool. This technique elim- inates additional coordinate transformations to external devices (such as the digitizer), resulting in increased surgical accuracy. The registration technique was tested on an RSPR six-degrees-of-free- dom parallel robot specifically designed for medical applications. A six-axis force sensor attached to the robot’s moving platform en- ables fast and accurate acquisition of positions and surface normal directions at sampled points.

Sampling with a robot probe was shown to be accurate, fast, and easy to perform. The robot performs most of the registration pro- cedures, leaving the surgeon’s hands free. The whole procedure takes about 2 minutes.

The simplicity and accuracy of the proposed registration is appropriate for specified orthopedic surgical applications. Robotic

DEVELOPMENTS OF THE HULL COMPUTER- ASSISTED ORTHOPAEDIC SURGERY SYSTEM (CAOSS) PROJECTS

R. Phillips’, W.J. Viant’, M.S. Beilby’, A.M.M.A. Mohsen’, M. Chawda’. and K.P. Sherman’ I Dept of Computer Science, University of Hull, Hull. United Kingdom, HU6 7RX * Department of Traumatology and Orthopae- dics, Hull Royal Infirmary, Anlaby Road, Hull, United Kingdom, HU3 2JE, E-mail: R.Phillips@dcs.hull.ac.uk

The University of Hull and hospitals in the Hull region have pursued research into Computer-Assisted Orthopaedic Surgery (CAOS) for over seven years. Two milestone projects (CAOSS-1 and CAOSS-2) have been undertaken. Both projects provide com- puter-based intra-operative surgical planning using a fluoroscopic image intensifier, and both feature a self-supporting arm to position surgical tools according to the surgical plan.

CAOSS-I (1992-1995) was a laboratory prototype developed to understand problems and investigate solutions for computer-aided surgery. This system assisted orthopaedic surgeons to drill bones accurately for the insertion of the compressiorddynamic hip screw (DHS) into the femoral head and for distal locking of intramedul- lary femoral nails [l]. A grid of 32 x 32 balls is used preoperatively to calibrate the fluoroscopic image. Anatomy is localized intraop- eratively by an optical tracking system that tracks the C-arm via LEDs attached to its X-ray receptor. After computer-based opera- tion planning on two orthogonal fluoroscopic images, the surgeon then places a guiding cannula along the required trajectory. This placement is computer guided, and an arm that is self-supporting holds the cannula; all the arm’s joints have brakes and position encoders. .- . ..

I . registration iinKS nawiessiy between preoperative planning and robotic assistance during surgery.

L.KCSSS-2 was a pre-production prototype used to evaluate the efficacy of our intra-operative image-guided surgery approach via clinical trials. Besides CAOSS-I operations, CAOSS-2 also caters for the insertion of cannulated screws for repair of neck of femur fractures [2]. Clinical trials started in May 2000. The main differ- ences from CAOSS-I were improved accuracy [3], a new method of localizing anatomy, and a much simpler arm for holding the guiding cannula. To localize anatomy, a small phantom with 21 balls is placed in the fluoroscopic image space when imaging the patient. Back-projection of the phantom’s balls and optical tracking of the position of the phantom enables the anatomy to be localized. A commercially available passive arm with a purpose-built end- effector holds the guiding cannula. The arm has three joints, which are locked by tightening a single knob. The position of the end- effector is obtained by optical tracking of LEDs attached to the end-effector. This passive arm is attached to rails on the operating table by standard theater clamps.

The CAOSS-2’s approach to anatomy localization reduces the volume observed by the optical tracking cameras. Both our systems use a self-supporting arm to maintain the position of surgical tools. This contrasts with other approaches using hand-held surgical tools, robot manipulation of surgical tools, and anatomy-fitting templates.

1. Viant WJ, Phillips R, Griffiths WJ, Mohsen AMMA, Cain TJ, Karpinski MRK, Sherman KP. A computer assisted onhopaedic surgical system for distal locking of intramedullary nails. Proc Institute of Mechanical Engineers 1997;211 Part H:293-300.

2. Viant WJ, Phillips R, Mohsen A. Computer assisted position- ing of cannulated hip screw. Computer Assisted Radiology and Surgery (CARS’99). Paris, June 1999.

3. Viant WJ, Phillips. R, Beilby MS, Zhu Y, Griffiths JG, Mohsen AMMA, Sherman KP. A technique for a very high accu- racy image intensifier calibration. Proceedings of Medicine Meets Virtual Reality, January 20-23 1999. p 379-380

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

366 Abstractsfrom ZSRACAS 2000

SIMULATOR AND DISTAL TARGETING DEVICE FOR IN-VITRO TRAINING AND EXPERIMENTATION IN COMPUTER-AIDED CLOSED MEDULLARY NAILING

L. Joskowicz’, C. Milgrom’, A. Simkin’, S. Kimchi4, 2. Yaniv’, 0. Sadowsky’ E-mail: josko@cs.huji..ac.il School of Computer Science’, and Precision Machining Work- shop4, The Hebrew University of Jerusalem, Israel, Depts. of Or- thopaedic Surgery2 and Experimental Surgery3, Hadassah Univer- sity Hospital, Jerusalem, Israel

Introduction: Reducing the surgeon’s cumulative radiation ex- posure and improving the positioning accuracy are key issues in computer-assisted orthopaedic surgery. We have developed FRA- CAS, a computer-integrated system for closed long bone fracture reduction. The system replaces uncorrelated static fluoroscopic images with a virtual reality display of 3D bone models created from preoperative CT and tracked intraoperatively in real time. Two important issues for system acceptance are distal locking and training. Because the nail bends when inserted, distal locking requires automatic recognition of nail holes and an adjustable drill guide to ensure that the holes are drilled in the right position. Training devices help familiarize surgeons with computer-gener- ated real-time multiple view displays, and with the requirements of optical tracking, and permit study of system ergonomy.

Materials and Methods: We have developed two devices and software for in-vitro experimentation and training for computer- aided closed medullary nailing: a fracture reduction simulator and an adjustable drill guide for distal locking. The simulator consists of two adjustable bone-fragment holders mounted on a radiolucent base, whose positions are followed by optical tracking. Each bone- fragment holder is a spring-loaded lockable spherical joint to which distal and proximal bone fragments are attached. The distal frag- ment holder also translates, simulating the action of the muscles on the bones. The device allows surgeons to practice bone alignment and distal nailing based on computer-generated images. It also serves as a platform to evaluate the ergonomy and accuracy of the system.

The adjustable drill guide is a radiolucent, 5DOF device for assisting the surgeon in drilling the holes for distal locking screws. The guide attaches to the nail’s head like the proximal targeting fixture. The drill is tracked in real time, and its position with respect to the nail holes is determined by image processing and registration software. The axes of the nail holes are automatically identified in the fluoroscopic images and registered to the bone model. The surgeon can then adjust the position and orientation of the drill guide until its axis and the axis of the nail are identical. The guide, whose position and orientation is tracked in real time, can also be used independently as a targeting device.

Results: Both devices have been integrated into the current FRA- CAS system. Preliminary experiments show improvement in the acceptance of the computer-aided system and allow for accuracy evaluation of image-based methods.

Reference : 1. Joskowicz L, Milgrom C, Simkin A, Tockus L, Yaniv Z.

FRACAS: A system for computer-aided image-guided long bone fracture Surgery. Comp Aid Surg 1999;3:271-288.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

Abstractsfiom ZSRACAS 2000 367

TELESURGERY: FROM DEVELOPMENT TO CLINICAL APPLICATION (Invited talk)

Jon C. Bowersox, MD, PhD, FACS Department of Surgery University of California, San Francisco San Francisco, California 94143-1674, USA E-mail: jcbsx@itsa.ucsf.edu

When the potential of telesurgery was first introduced to surgeons in 1995, few believed that robotic systems would actually be used to operate on humans. It was hard to imagine that an electrome- chanical device could replicate the precise motions necessary for dynamic tissue handling, and concerns were raised regarding safety and patient acceptance. Less than three years later, the first clinical use of a surgical telemanipulator occurred in France, when surgeons working from a console positioned across the operating room repaired a damaged heart valve. Since then, more than one hundred patients have undergone surgery with the assistance of robotic surgical systems. The successful introduction of surgical telema- nipulators has been based on the coordinated development in three key areas: system design, user validation, and clinical need.

System design: Remote teleoperation has been used in hazardous material handling for almost fifty years. Consequently, the princi- ples of master-slave manipulators were well described when tele- surgery development began in the 1980s. The forces and working volumes required for precision handling of living organs, however, are substantially different from those used in industrial applications. Thus, in the telesurgery program started at Stanford Research Institute (SRI International), under the direction of Phil Green, robotic manipulators and control systems were specihcally de- signed for use in operating rooms, on delicate tissues. Equally critical was the design of the user interface. Soft-tissue surgery is performed in a complex and dynamic task environment. Three- dimensional spatial orientation is necessary for precision tissue handling, as are haptic cues including perception of tissue mass and elasticity and instrument forces. Proper eye-hand orientation (ocu- lovestibular axis) must be maintained to create a natural, intuitive work environment. These principles were incorporated into the surgeon’s console of the SRI Telepresence Surgery System. The operator looks down at a mirrored screen, on which a three- dimensional image of the remote environment is projected. Hands are placed in standard surgical instrument handles attached to the manipulator masters. Surgical instrument tips, attached to the ma- nipulator slaves, are oriented in such a manner as to appear as though they are extending from the surgeon’s hands. Gravity com- pensation and force feedback are incorporated, closely approximat- ing the feel of instruments encountered in the operating room. The integrated user interface instills a sense of immersion in the remote environment. allowing existing surgical skills to be applied without special training or practice.

User validation: The potential of remote telepresence surgery was demonstrated in a series of experiments on live, anesthetized swine, using a prototype four-degree-of-freedom (4DOF) telema- nipulator system. Procedures were chosen to replicate the range of task elements encountered in soft-tissue surgery. These fundamen- tal maneuvers included incision. grasping, dissection (tissue spread- ing and cutting), suturing, and knot tying. Surgeons performed common operations, including organ removal (cholecystectomy, nephrectomy), blood vessel repair and replacement, catheter ma- nipulation, and intestinal anastomoses, operating from a remote console. All procedures were completed successfully without com- plications. Procedures required 2.5-2.8 times as long to complete using the telemanipulator, as compared with conventional surgical techniques.

A 4DOF remote center-of-motion system was designed for min- imally invasive surgery. Using the same intuitive operator interface, intracorporeal suturing and knot tying, as well as tubular anasto-

moses, were performed under ex vivo conditions. Performance was significantly better than that achieved using standard laparoscopic techniques for all task conditions. More recently, a prototype mi- cromanipulator has been developed and used in experimental mi- crovascular procedures. Again, from the same operator interface, surgeons were able to view highly magnified, stereoscopic images of 1 mm diameter rat femoral arteries, and perform precise, end- to-end anastomoses using standard microsurgical techniques.

Clinical need: The rapid progression of telesurgery from proto- type demonstration to clinical application was based on evidence that patients recovered more rapidly, with fewer complications, from minimally invasive surgery. Loss of intracorporeal wrist mo- tion, and the unnatural orientation of video displays and instrumen- tation have created a challenging environment for surgeons. For straightforward procedures not requiring suturing or complex tissue manipulations (e.g., gallbladder removal), minimally invasive tech- niques have been widely adopted. Special training. frequent prac- tice, and lengthy operating times are required for more complex laparoscopic procedures, including suturing and knot tying, and have limited their application to relatively few centers and sur- geons. By restoring wrist motion at the tissue level, and a natural work orientation, commercially-developed surgical telemanipula- tors offer surgeons an opportunity to apply minimally invasive techniques to more challenging cases, such as myocardial revascu- larization and valve repair. To date, the use of robotic systems in cardiac surgery has been described in more than a dozen peer- reviewed articles from Europe and North America. Approval to market such systems has been received in Europe, and clinical trials are underway in the United States, Canada and Japan.

Future opportunities for surgical telemanipulator development include component miniaturization, directed energy delivery, and special-purpose Instrumentation. Microscale 6DOF instruments are needed for fetal, congenital heart, and neonatal surgery. Alternative approaches to suturing and knot tying for joining tissues, such as laser welding or biological gluing, would increase telemanipulator value. Finally, less versatile, but less expensive, systems could be developed for specialized applications, such as ophthalmic proce- dures and image-guided tumor biopsy or ablation.

__ -

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

368 Abstractsfroom ZSRAolS 2000

SESSION 3 EMERGING TECHNOLOGIES AND APPLICATIONS

PLACEMENT OF ENDOSTEAL IMPLANTS

Wolfgang Birkfellner, Franz Watzinger, Felix Wanschitz, Farzad Ziya, Judith Kremser, Andreas Potyka, Robert Mayr, Klaus Huber. Franz Kainberger, Rolf Ewers, and Helmar Bergmann Departments of Biomedical Engineering and Physics, Oral and Maxillofacial Surgery, Anatomy, Diagnostic Radiology, and the Ludwig-Boltzmann Institute of Nuclear Medicine, University of Vienna, Vienna, Austria

Endosteal implants used in edentulous patients also facilitate obtu- rator prosthesis fixation in tumor patients after radical resection of carcinoma in the maxilla. Previous clinical studies have shown, however, that survival of implants placed into available bone after maxillectomy is generally poor. Nevertheless, implants positioned optimally in residual zygomatic bone provide superior stability from a biomechanical point of view. In a pilot study, we assessed the precision of VISIT, a computer-aided surgical navigation sys- tem dedicated to placement of endosteal implants in the maxillo- facial area. Five cadaver specimens underwent hemimaxillectomy. The cadaver head was matched to a preoperative high resolution CT by using implanted surgical microscrews as fiducial markers. The position of a surgical drill relative to the cadaver head was deter- mined with an optical tracking system. Implants were placed into the zygomatic arch. where maximum bone volume was available. The results were assessed using tests for allocation accuracy and postoperative CT-scans of the cadaver specimens. The localization accuracy of landmarks on the bony skull was 0.6 +/- 0.3 mm (average and standard deviation), determined with a 5-degree-of- freedom pointer probe. The allocation accuracy of the tip of the implant burr was 1.7 +I- 0.4 mm. The accuracy of the implant position compared to the planned position was 1.3 +I- 0.8 mm for the external perforation and 1.7 +I- 1.3 mm for the internal perfo- ration of the zygoma. Eight out of 10 implants were inserted with maximum contact with surrounding bone, and two implants were located unfavorably. However, reliable placement of implants in this region is difficult to achieve. The technique described in this paper may be very helpful in the management of patients with poor support for obturator prostheses after maxillary resection. Further- more, we conclude that this first cadaver study has proven that VISIT, our modular software system for fast development of novel applications of computer-aided surgery, is suitable for use in the operating theater.

IMAGE-GUIDED DIAGNOSIS A h 9 TREATMENT OF RETINAL DISEASES: PROGRESS AND CHALLENGES

Jeffrey W. Berger, MD, PhD and Bojidar Madjarov, MD Computer Vision Laboratory, Scheie Eye Institute University of Pennsylvania, Philadelphia, PA, USA 19104 Phone: 215-662-8675, Fax: 215-662-0133 E-mail: jwberger@mail.med.upenn.edu

We have recently proposed, and have made progress towards im- plementing, an augmented reality environment for real-time track- ing and overlay of previously-stored photographic and angio- graphic images directly onto the real-time slit-lamp eye fundus (retinal) biomicroscopic image in order to allow for real-time image comparison, measurement and correlation, as is required for diag- nosis and treatment of retinal diseases. In this report we describe progress and challenges relevant to design and deployment of an ophthalmic augmented reality environment.

We are developing a slit-lamp-based platform for ophthalmic image overlay. The image of one of the oculars is imaged onto a CCD camera, and sent to a framegrabber and digitizer interfaced to a personal computer. Previously stored images can be selected in a graphical user interface environment, and the real-time slit-lamp biomicroscopic image is registered with prior images based on fundus features (usually blood vessels) present in both images. Once registered, the previously stored image is superimposed onto the real-time biomicroscopic image.

In order to limit algorithmic requirements and demonstrate proof of principle, we have reduced the multi-modal, real-time registra- tion problem to two steps. First, a real-time biomicroscopic fundus image is acquired. This biomicroscopic fundus image is registered with the previously stored fundus-camera-derived photographic or angiographic image in non-real-time. A landmark, for example, a vessel crossing or bifurcation, is then identified in the captured biomicroscopic image. The incoming video image sequence is then searched for the presence and position of the identified landmark, and the previously stored image is rendered based on the highly accurate, non-real-time, non-global initial registration. Our system is built around a Windows NT, Pentium I11 personal

computer. A graphical-user interface has been designed to allow for interactivity, access to previously stored images, and control over environment functionality. The system was first evaluated in model eyes where the “blood vessels’’ are of maximum contrast and visibility. Following initial development and refinement, the algo- rithmic and ergonomic issues involved in migrating from model eyes to human subject are being explored. Image sequences were obtained for five human subjects. Tracking with search over trans- lation-only performed at -5-10 Hz. When searching over transla- tion and rotation, tracking occurred at - 3 Hz. Misregistration errors are approximately 1-5 pixels.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

Abstractsfrom ZSRACAS 2000 369

MEDICAL ROBOTICS: PRINCIPLES AND APPLICATIONS (Invited talk)

Paolo Dario, Maria Chiara Carrozza, Giuseppe Megali, and Oliver Tonet Scuola Superiore Sant'Anna - MiTech Lab Via Carducci 40 - 56127 - Pisa, Italy dario(chiara, peppe, oly)@mail-arts.sssup.it

The objective of Computer-Assisted Surgery (CAS) is to provide the surgeon with benefits deriving from the integration of advanced image processing techniques and computer graphics with new robotic and mechatronic surgical instruments, in order to achieve more effective planning and more accurate execution of surgical interventions.

A variety of new surgical tools are currently under development as part of CAS systems and as stand-alone devices. According to our vision of this field, all new surgical tools can be classified conceptually into one of three main classes, based on their technical characteristics and on the aspect of surgeon performance they aim to enhance: a) robotic tools; b) teleoperated dexterous tools; and c) manually operated mechatronic tools.

Robotic tools are integral parts of CAS systems. They execute very accurately and autonomously (though under the supervision of the surgeon) the tasks pre-planned by the surgeon using preopera- tive data on the patient. The main advantage of the robotic tool, as compared to direct human operation, is much higher accuracy due to the intrinsic stiffness of the robot manipulator and the resulting precise execution of desired trajectory and force patterns.

The distinctive feature of teleoperated dexterous tools is dexter- ity. This class of surgical tools comprises teleoperated instruments supporting different levelsof direct surgeon control. These instru- ments allow the surgeon to access sites in the human body that are difficult to reach, and to use a variety of miniaturized tools for different types of intervention and therapy.

The third class of surgical instruments, the mechatronic tools, is perhaps the least well known and investigated in the field of CAS. Mechatronic surgical tools include integrated and miniaturized pre- cision mechanisms, sensors, actuators, preprocessing electronics, embedded microcontrollers, and a humadmachine interface. Sim- ple and intuitive operation, little or no need for complex training of the surgeon and operating room personnel, real-time compensation for perturbations, easy integration with CAS systems, and low cost are the main advantages of smart mechatronic tools, as compared to both traditional surgical tools and to robotic tools, which may result in their widespread use in clinical practice.

In this presentation. the authors will discuss the working princi- ples and objectives of each class of robotic systems for CAS, and will present examples of their present applications in clinical prac- tice, along with some considerations on future principles.

SESSION 4 MEDICAL IMAGE PROCESSING AND VIRTUAL REALITY

PATH EXTRACTION IN 3D MEDICAL IMAGES FOR VIRTUAL ENDOSCOPY

T. Deschamps' and L. Cohen2 ' Professional Imaging System Group, LEP, 22, avenue Descanes, BP 15, F-94453 Limeil-Brevannes Cedex, France * Ceremade UMR CNRS 7534, Universitt Paris IX Dauphine Place du Marechal de Lattre de Tassigny, 75775 Paris Cedex 16, France E-mails: deschamps@lep-philips.fr, cohen@ceremade.dauphine.fr

We have developed a fast and efficient algorithm that computes a path for guiding endoscopic viewing. The inputs are just the start point and end point of the path.

This algorithm is based on previous work [I] for extracting paths in 2D images, given the two extremities of the path, using a front propagation equation. This technique maps the active contours [2] into a minimal path problem minimizing only a feature potential P term. It makes global minimization, reduces the user interaction, and the front propagation is solved using Fast Marching [3]. Our original contribution is to extend this technique to 3D. We also introduce a method to extract a centered path in tubular structures, which is very useful for objects with complex shapes.

This work finds its motivation in the particular case of 3D medical images. We applied this technique to the problem of finding a path in tubular anatomical structures with minimum interactivity, and more particularly to virtual endoscopy. Usually,

- p a t h - c o n s ~ c t i o ~ ~ e f t - ~ ~ e - u s e r ; - w h u i d e -the-virtual- endoscope by hand. For a complex structure, the path construction in 3D images becomes a very tedious task. Using our 3D front propagation method, we propose an automatic path tracking method to overcome this drawback: we are able to build a path given one of the two endpoints. Successful results were obtained for various anatomical regions (colon, brain vessels etc.) with images from different 3D imaging modalities (CT. MR).

References: 1. Cohen LD, Kimmel R. Global minimum for active contours

models: a minimal path approach. Int J Computer Vision 199724:

2. Kass M, Witkin A, Terzopolous D. Snakes: Active contour models. Int J Computer Vision 1988;1:321-331.

3. Sethian JA. Level set methods: evolving interfaces in geom- etry, fluid mechanics, computer vision and materials sciences. Cam- bridge University Press, UC Berkeley, 1996.

57-78.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.

370 Abstracts- ZSRACAS 2000

COMPUTERIZED PROCESSING OF DIGITAL BONE RADIOGRAPHS AS A TOOL TO ESTIMATE FRACTURE RISK

I. Leichterl, A. Simkin’, V. Neeman’. 1. I l sd . 0. Mano?, and M. ~ i e b e r g a ~ ’ ‘Dept. of Electro-optics, The Jerusalem College of Technology, ’Dept. of Orthopaedic Surgery. Hadassah University Hospital, ’School of Public Health, The Hebrew University of Jerusalem, Jerusalem, Israel

Objectives: radiographs of the proximal femur by computerized processing.

Background: DXA is the most widely used method for measur- ing bone mineral content. This method is. however, costly and hence not readily available to the population at risk. The present study examines a computerized method that quantifies mineral density and trabecular organization from routine bone radiographs, independent of the conditions under which the X-rays were ob- tained.

To predict bone strength by features derived from

Material and Methods: Seventeen fresh human femora were excised and x-rayed, next to a calibration wedge, in the A-P position at two exposures (40 KV at 80 mAS and 500 mAS). The radiographs were digitized and four regions, identical to those defined by the DXA software, were defined on the digital image. The absolute average gray levels (AGL) were calculated for each region. The images were then processed, based on the values of the calibration wedge, in order to obtain normalized gray level values (NGL) that reflect the mineral density independently of the X-ray exposure. Bone dimensions were also measured on the radiographs: the minimal width and the length of the femoral neck, the diameter of the femoral head, and the neck-shaft angle. Optical Fourier analysis of the trabecular pattern on the radiograph reflected the trabecular organization in the proximal femur. The power spectrum represented by the light distribution in the Fourier image was analyzed to yield the trabecular bone index (TBI), which depends on the ratio of high- and low-frequency components. The bone mineral density (BMD, g/cm2) at the proximal femur was mea- sured. The strength of the proximal femur was measured in a loading configuration which simulates a fall on the greater trochan- ter. Mechanical load was applied to the femoral head in a Material Testing Machine, at a speed of 500 mdmin. The load at fracture was obtained for each specimen. Multiple regression analysis was used to assess the ability of a model combining the variables derived from the digital bone image and those measured by DXA to predict the fracture load.

Results: The correlation between the BMD values and the fracture load ranged between 0.81 for the femoral neck and 0.92 for the trochanteric region. The correlations of the AGL values with the fracture load were much lower, ranging between 0.28 for the femoral neck and 0.53 for the trochanteric region. Following normalization of the gray levels, the cornlation with the fracture load improved, ranging between 0.76 and 0.88 for the respective regions. The correlation between the TBI values and the fracture load ranged between 0.32 for the femoral neck and 0.82 for the intertrochanteric region. Multivariate analysis of a combination of the NGL values (representing bone mass at the various regions of the proximal femur) and of the TBI values (representing the trabecular organization of the same regions) yielded a much higher prediction of the fracture load (R = 0.97; p<O.OOOI). The inclusion of bone dimensions in the analysis improved the corre- lation significantly to R = 0.99.

Conclusions: The information that can be extracted by comput- erized image processing of routine bone radiographs provides vari- ables related to the mineral density and trabecular organization of bone tissue. The combination of these two bone-features at different locations within the proximal femur results in reliable and relevant assessment of bone strength.

CLINICAL EVALUATION OF A SEE-THROUGH DISPLAY FOR INTRAOPERATIVE PRESENTATION OF PLANNING DATA

J. Brief’. S. Hassfeld’, T. Salbz, 0. Burgert’, J. Miinchenberg’, H. Grabowski’. T. Redlich’, C. Ziegler’, I. Raczkowsky’. R. Krem- pied, H. Worn*, R. Dillmann’, and J. Miihling’ I Department of Oral and Maxillofacial Surgery, University of Heidelberg, Germany; Institute of Real-Time Computer Systems & Robotics, University of Karlsruhe, Germany; ’ Department of Radiology, University of Heidelberg, Germany E-mail: jbrief@med.uni-heidelberg.de

Introduction: Computer Aided Surgery has evolved rapidly in recent years. Nowadays, computer-based tools for preoperative planning and simulation are widely used. While preoperative tasks are already supported by computer technology, the surgical proce- dure itself still lacks computer-based assistance. To perform surgi- cal interventions with the necessary accuracy, intraoperative sup- port in the application of previously planned and simulated actions is indispensable. Our approach for closing this gap is an augmented reality-based system for intraoperative presentation of planning and simulation results.

Augmented or Enhanced Reality refers to systems that merge virtual, computer-generated data with real imagery into a single coherent perception of the scene configuration under consideration. This superimposition of computer graphics on real data provides an enhanced view of the scene. The aim of intraoperative presentation of image data is to offer relevant information to the physician directly in the operation field. We suggest using a see-through head-mounted display for the superimposition of patient and virtual data. In the display, we want to present the surgeon with the result from preoperative planning, e.g, the ideal plan for cutting and drawing the bone, by presenting the cutting plan in the see-through display.

Methods: In this paper, we present our approach to evaluating which color would be best for presenting virtual planning data in the see-through head-mounted display during surgical operations. We chose the Sony Glasstron LDI IOOE device. The display offers SVGA resolution (800 x 600 pixels) with 1.55 million dot LCD. It is lightweight (120 g) and offers a wide range of functions. It is possible to switch between see-through mode and non-see-through mode in an infinitely variable manner.

We compared different colors for the planning data itself and different colors for the background. Geometrical figures in different colors were projected into the see-through display while looking at the operation scene.

We measured the quality of the intraoperative presentation by giving scores between 1 and 6 ( 1 = best. 6 = worst). We compared the subjective judgment of 14 persons during different surgical interventions in the operating theater. All persons had to look at the operation field through the see-through glasses and were asked to evaluate how well or poorly they were able to recognize the (different coloured) planning data combined with the real view of the patient.

Results: The majority of test persons judged that yellow would be the best color for the plan itself, black for the background. Second was violet for the plan. Third choice was yellow on a blue background, and, just to mention the first four, green on a black background. During testing, we recognized that some test persons complained of being unwell or sickness, so not all surgeons will be able to use such an Enhanced Reality device. This phenomenon, as well as accuracy tests and experiences with clinical application, will be the focus of our future research.

Summary: We present an overview of our experimental setup for the intraoperative presentation, including a see-through head- mounted display.

Com

pute

r A

ided

Sur

gery

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Heb

rew

Uni

vers

ity o

n 08

/04/

15Fo

r pe

rson

al u

se o

nly.