Computer Aided Surgery 4:208 –228 (1999) Abstracts from the 4th Symposium on Computer Assisted Orthopaedic Surgery (CAOS’99), held in Davos, Switzerland on March 18th and 19th, 1999* REMARKS FROM THE SCIENTIFIC ORGANIZERS In 1995 we decided to organize a meeting on Computer Assisted Orthopaedic Surgery. When we sent out the announcements, we called it “First Symposium”, without knowing whether we would be successful enough to ever have a second one. Well, we have now reached symposium number four, and since 1997 our North American “twin”, CAOS/USA, has been held in Pittsburgh, Pennsylvania. This increase in popularity was one of the main reasons why we had to look for another location after CAOS III in Bern in November 1997. We found the ideal place at the Convention Center in Davos, Switzerland. The first three symposia relied upon an invited faculty, but this year we were overwhelmed by the more than 70 abstracts that we received as a result of our call for papers. This also indicates the growing attention that is being paid to Computer Assisted Orthopaedic Surgery by the medical community. Back in 1995, most lectures presented laboratory settings and CAOS systems under development. Besides ROBODOC, only a few pedicle screw systems had been used clinically up to that point. In this year’s program, however, in vivo experience from almost all anatomical areas was represented. Many systems have become commercial products, helping to improve accuracy and to reduce surgical risk to patients. It also appears that the obvious success of CAOS technology gave the researchers time to think about related issues and side aspects – Session VIII reflects this trend. We would like to thank the program committee for their assistance, and all sponsors that allowed us to set up CAOS this year. Special thanks go to all the participants, without whom the great success that the symposium has now achieved would not have been possible. Chairmen: Lutz-P. Nolte Reinhold Ganz Anthony M. DiGioia Scientific Coordinators: Marwan Sati Frank Langlotz * Abstracts from Sessions VI–IX are presented in this issue; those from Sessions I–V appeared in the previous issue. Abstracts from CAOS ’99 SESSION VI: IMAGE GUIDANCE FOR LONG BONE FRACTURES 209 COMPUTER ASSISTED FEMORAL NAILING PRINCIPLES OF PRECISE FLUOROSCOPY-BASED SURGICAL NAVIGATION C. Krettek1, M. Slomczykowski2, G. Köppen1, A. Hauptmeier1, R. Hofstetter2, M. Sati2, H. Tscherne1. 1Unfallchirurgische Klinik, Medizinische Hochschule Hannover, Hannover, Germany, 2M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland. R. Hofstetter, M. Slomczycowski, M. Sati, L.-P. Nolte. M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland. Objectives: To define a computer assisted technique for closed femoral nailing and evaluate its feasibility. Objectives: To highlight concepts of a computer assisted freehand navigation system that uses single intra-operatively acquired fluoroscopic images as a basis for real-time navigation of surgical tools, implants or bone fragments. Background: A significant disadvantage of closed femoral nailing is the extensive intraoperative radiation and the risk of axial malalignment (varus-valgus, ante-recurvatum, length, rotation). Using computer assisted navigation, a large part of intraoperative radiation could be avoided, and the precision of alignment and distal locking could be increased. Background: Intra-operative fluoroscopy is a valuable tool for visualizing underlying bone and surgical tool positions in orthopedics. Disadvantages of this technology include the need for continued radiation exposure for visual control and cumbersome alignment of the fluoroscope. Design/Methods: Optoelectronic markers are placed on surgical tools, patient references, and the fluoroscope to track their position in space. Projection properties of the fluoroscope are acquired through an initial pre-calibration procedure. Corrections are applied to compensate for both the fluoroscope’s image intensifier distortions and mechanical bending of the C-arm frame. This enables real-time simulation of surgical tool positions simultaneously in several single-shot fluoroscopic images. In addition, through optoelectronically tracked digitization of a target view point, the fluoroscope can be numerically aligned at exact angles relative to the patient without any X-ray exposure. Intra-operative definition of anatomic landmarks in two fluoroscopic views allows various static and real-time anatomic measurements. Results: For clinical evaluations (see subsequent abstracts) a prototype has been developed that achieved a navigation accuracy better than 1/-1 mm. Design/Methods: An optoelectronic tracking system (OPTOTRAK) was used. Light emitting diodes (LED) were fixed to the image intensifier (C-arm), to the proximal and distal main fragment, and to the instruments and implants. This allowed the establishment of a virtual, spatial link between the anatomy, the instruments and implants, and the 2D-C-arm images. A real-time visualization was possible by projection of the surgical instruments/implants onto the 2D-C-arm images on the monitor. After initial tests in plastic and cadaver bones without soft tissue, and in whole-body cadaver specimens, clinical applications followed. Distal locking was tested using a 4.0-mm drill in a model with a copper disk. Precision was analyzed using a caliper. Results: A C-arm-based navigation system for intramedullary nailing could be established and was successfully tested in cadaver specimens. Both initial problems, the reference-base-to-bone fixation and implant deformation, were solved using modified tools and protocols. The misfit between the center of the drill and the target was 0.91 mm (mean, SD 60.59 mm, range 0 to 2.1 mm, n 5 30). Conclusions: The developed 2D image-based computer assisted femoral nailing technique has great potential for the improvement of current techniques, especially in terms of reducing radiation, minimizing soft tissue disruption and increasing precision of alignment. 210 Abstracts from CAOS ’99 COMPUTER AIDED FRACTURE PLATE FIXATION J. Kowal, Y. Bourquin, M. Sati. M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland. Objectives: Minimally-invasive surgical techniques for fracture fixation are difficult. The objective of the present study is to develop a system to improve the navigation of plate positioning and fixation during biological osteosynthesis. Design/Methods: A fractured plastic bone, various osteosynthesis plates, a surgical drill, and screwdriver were fitted with optoelectronic tracking markers. Two-dimensional (2D) X-ray images from fractured bones were integrated within a 3D OpenGL computer graphics environment. X-ray projection parameters were imported from another CAS system (Hofstetter et al.). A projection model determined from these parameters allowed display of detailed three-dimensional (3D) geometric models of surgical tools and implants overlaid onto the X-ray images. With optoelectronic tracking, real-time display of these objects within the scene could be achieved. By digitizing the plate holes before surgical insertion, the system calculated the correct shape of the bent plate from a generic model. The system helps to (a) guide the surgeon to align the plate in the correct position relative to the underlying fracture site, (b) drill the screw holes, and (c) insert the screws. The current plate-database may easily be expanded with new implants for future applications. Results: The proposed prototype solution gives the surgeon a realistic and interactive feedback of normally hidden surgical actions. First attempts with broken plastic bones turned out to be very promising. Conclusions: The existing work shows that a minimally-invasive fracture fixation could be improved using advanced image interactive technology. Moreover, replacement of constant fluoroscopy with a few computer-integrated X-ray images promises significant reduction in radiation exposure. FLUOROSCOPY-BASED SURGICAL NAVIGATION FOR LOCKING OF PFN AND UTN IMPLANTS N. Suhm1, P. Messmer1, B. Baumann2, W. Steinbrich2, P. Regazzoni1, A.L. Jacob2. 1Department of Surgery, Trauma Division, University Hospitals Basel, Basel, Switzerland, 2Department of Radiology, University Hospitals Basel, Basel, Switzerland. Background: Until recently, surgical navigation systems required a 3D image dataset. This limited their usefulness to operations for which preoperative CT or MRI are the methods of assessment. The development of a fluoroscopy-based navigation system broadens the spectrum of applications for navigation techniques. Material and Methods: We have used a fluoroscopy-based navigation system (Surgigate, Medivision, Oberdorf, Switzerland) for distal interlocking of intramedullary implants. The system includes an imaging unit, equipment for optical tracking, hardware and software unit and surgical instrumentation. Before starting the operation, the reference base is attached to the insertion handle for the nail. Prior to distal interlocking, the image intensifier is placed such that the locking holes are perfectly circular in the lateral view. The image data are transferred to the navigation system and serve for continuous visualization of the locking hole and the drill tip. Once it can be clearly seen on the monitor that the longitudinal axis of the drill and locking hole are parallel, drilling can commence. 50 distal interlocking procedures of the PFN were performed with a test model without any failure. In 10 osteosyntheses (8 PFN, 2 UTN), 18 of 20 holes for locking bolts were successfully drilled. Two holes were drilled in the wrong place. All locking procedures were performed on the basis of a single fluoroscopic image. Conclusions: We conclude that distal locking of intramedullary implants performed with the navigation system requires minimal intraoperative fluoroscopy. The inaccuracies observed were primarily attributed to the learning experience and the necessary adaptations to the clinical application. Abstracts from CAOS ’99 211 AN EXTERNAL FIXATOR FOR COMPUTER ASSISTED FRACTURE TREATMENT COMPUTER ASSISTED MEASUREMENT OF FEMORAL ANTEVERSION K. Seide, D. Wolter, U. Schümann. Berufsgenossenschaftliches Unfallkrankenhaus Hamburg, Hamburg, Germany. M. Slomczykowski1, C. Krettek2, G. Köppen2, A. Hauptmeier2, H. Tscherne1. 1M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland, 2Unfallchirurgische Klinik, Medizinische Hochschule Hannover, Hannover, Germany. Objectives: An external fixator system, which allows a computer assisted modification of its shape and an in vivo measurement of fixator loads in 6 degrees-of-freedom has been developed. In a clinical study the capability of the system for the reduction of fractures of long bones and the significance of fixator load parameters to monitor callus formation were evaluated. Methods: A construction in the form of a hexapod (Stewart platform) is proposed. A Windows-based computer program was developed and optimized for its clinical use. The fixator was controlled either manually or with electro-motors. Using single-axis force transducers in line with the six linear actuators, axial and shear forces as well as rotational and bending moments were determined. The fixator was used in 49 patients, in which it was applied for primary or secondary fracture reduction on the lower leg, the femur and the humerus. In 9 patients treated for tibial fractures, axial force and bending moments in the fixator were evaluated 2, 4, 8 and 12 weeks after the osteosynthesis. Results: Using the assembly, reductions were possible with respect to any axis in space (maximum 60 mm and 35 degrees). Bone position accuracy was in the median 1 degree and 3mm. It was especially interesting that fracture reduction was painless to the patients, when performed sucessively with the hexapod fixator. In well-reduced fractures, axial forces in the fixator were already less than 10% of the external load, in the first measurement at 2 weeks after osteosysnthesis. As opposed to axial load, the fixator stabilizes significant bending moments at two weeks, which decreased due to callus formation. Thus, bending stiffness was found to be a good indicator for the monitoring of fracture healing. Conclusions: The clinical application showed that the hexapod fixator is a useful addition to external fixator systems for performing a planned fracture reduction. Additionally, the course of fracture healing can be monitored and callus load can be modified appropriately. In the next step, algorithms are developed to enable reduction and load adaptation to be done automatically, resulting in an “intelligent” fixator. Despite several methods applied, the precise estimation of the femoral anteversion still remains problematic. The goals of this study were: a) to introduce a definition of real AV angle (RAV); b) to evaluate the AV measurements performed by using computer assisted fluoroscopy method (CAFL, Hofstetter et al., see same proceedings) in comparison to the CT-based techniques. The angle between the plane created by the posterior region of the condyles, parallel to the femoral shaft axis and the line connecting the center of the femoral head and femoral neck defined the value of the RAV angle. In contrary to the standard CT-based technique, this angle lies in the plane sectioning the femoral neck along its mid axis and perpendicular to the “frontal” plane. The methods of the AV measurements used in this study were: a) CAFL; b) CT-based method with multiplanar reconstruction (CTMPR); c) traditional CT based method (CTTRAD). In the CAFL method, 4 fluoroscopic images for each bone were obtained. The posterior parts of the medial and lateral condyles, center of the femoral head and neck, and proximal and distal points for the femoral axis were reconstructed in space using bi-planar images obtained by a specially-equipped fluoroscope and utilizing custom written software. For the CTMPR method, the CT images were displayed in such a way to allow determination of the AV angle between the line connecting the center of the head and the middle point of the femoral neck in the mid-neck plane and the bicondyles-trochanter major plane. In contrast to the CAFL and CTMPR method, the CTTRAD AV angle was projected onto the transverse plane of human body. The AV angle of 24 pairs of bones was measured. No intra- or interobserver dependency for any tested method was found. High correlation coefficient between methods was proven. Statisticallysignificant difference was detected between CAFL and CTTRAD as well as between CTMPR and CTTRAD, and no difference between CAFL and CTMPR, since CAFL and CTMPR were based on the same concept of RAV. The CAFL technique is simple, provides reliable and precise results (61.1 degree) and can be used intraoperatively for monitoring of the AV changes during fracture reduction. The technique also allows AV angle to be compared pre- and postoperatively. 212 Abstracts from CAOS ’99 SESSION VII: POSTER SESSION PARTIAL SURFACE MATCHING FOR HIGH ACCURACY REGISTRATION K. Abdel-Malek1, D. McGowan2, L. Adhami3. 1Department of Mechanical Engineering, The University of Iowa, Iowa City, IA, USA, 2Nebraska Back and Spine Center, Good Samaritan Hospital, Kearney, Nebraska, USA, 3École Supérieure en Sciences Informatiques, Sophia-Antipolis, France. Objectives: The goal of this work is to develop an effective and robust method for surface matching leading to registration in computer assisted surgery. Background: Surface matching is the process that compares the shape of two surfaces in order to find a rigid motion of one object that makes a sufficiently large portion of its boundary lie sufficiently close to a corresponding portion of the boundary of the second object. Irregular and diverse objects whose surfaces cannot be described by fixed features are considered. Design/Methods: The models of two objects are represented using a proposed Grid Model method. Two models are fused to obtain an indication of the quality of the match. A sequence of translations and rotations is applied to one of the models until an appropriate match is reached. optimization techniques using genetics algorithms are implemented to select the appropriate motions from a predefined search space. Results: Accuracy calculations for the surface matching and registration of two objects were made. The surface of a vertebral body acquired using a 3D measuring-device is matched with a model obtained from CT models. Conclusions: A robust and accurate surface matching method was developed that uses two models obtained from two different sources. BIOMECHANICAL SIMULATIONS OF SCOLIOTIC SPINE INSTRUMENTATION C.-É. Aubin1,2, F. Poulin1,2, I.A.F. Stokes3, M. Gardner-Morse3, H. Labelle2. 1Department of Mechanical Engineering, École Polytechnique, Montréal, Canada, 2Sainte-Justine University Hospital, Montréal, Canada, 3Department of Orthopaedics and Rehabilitation, University of Vermont, Burlington, VT, USA. Objectives: To develop a preoperative biomechanical model to predict the effect of surgery on spine as a function of pre-operative geometry and surgical plan. Background: Surgical instrumentation of the scoliotic spine is a complex procedure implicating many parameters, such as the spinal segment to be instrumented, the number and position of the hooks/ screws, etc. Design/Methods: A biomechanical model combining rigid bodies (vertebrae) and flexible elements (intervertebral tissues) was designed using ADAMS software. Mechanical properties were taken from published data and the geometry was personalized using the patient’s 3D reconstruction built from biplane x-rays. Four steps were used to simulate the surgical maneuvers: 1) installation of hooks/screws on instrumented vertebrae; 2) attachment of the hooks/screws on the rod; 3) rod rotation, 4) hooks/screws lock-up and elastic spring-back. The surgical maneuvers were applied to 3 cases: 1) a physical model; 2) two scoliotic patients. Simulation results were compared with intraoperative measurements. Results: The trends observed during the simulations for geometric indices such as Cobb angles correspond to expected surgical results: gradual correction during surgical procedures, and partial release after hooks/screws lock-up to evacuate excessive forces induced by the correction maneuvers and stored in flexible elements. Discussion: This preliminary model of scoliosis surgery needs subsequent refinements to be a reliable tool to assist with preoperative planning. In the long term, this might be a practical tool for surgical correction predictions and more efficient (rationalized) instrumentation design. Abstracts from CAOS ’99 FRAMELESS COMPUTER-ASSISTED NAVIGATION SURGERY IN THE HEAD AREA R. Bächler1, M.D. Caversaccio2, K. Lädrach2, R. Häusler2, L.-P. Nolte1. 1M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland, 2Department of ENT, Head and Neck Surgery, University Hospital of Bern, Bern, Switzerland. Objective: Surgical treatments of the skull base and the paranasal sinuses are often difficult because of the complex structures. Computer-assisted navigation has increasingly proven its value in improving the safety of such operations. We present our system, which has been developed in Bern, and which has been in use since Spring 1997. Methods: Our CAS system is based on the intraoperative tracking of the patient’s head and the surgical instruments, each equipped with infrared LEDs, by an infrared position sensor. Both the CT acquisition and the surgery are performed frameless, i.e., without a head-holding device. During surgery, a dynamic reference base is attached on an upper jaw splint with a silicone adhesive. Because of the frameless approach, the mobility of the patient’s head is preserved during surgery. Results: So far, 70 surgeries have been performed. The system has been used for different surgical treatments at the anterior and lateral skull base (27), as well as for endonasal operations (43). The clinical accuracy of the system at the anterior skull base is between 0.5 and 2 mm and on the lateral skull base between 0.5 and 2.5mm. Using a new approach for surface-matching-based registration at the lateral skull base, the accuracy can be improved to range between 0.5 and 1.5mm. Installation time for the system was between 15 and 20 minutes. There were no complications due to the CAS system. Conclusions: The use of the CAS system assists the surgeon with the identification of important structures and permits safe surgery. In addition, the system permits efficient minimally-invasive procedures and has an educational effect. 213 CRIGOS – MODULAR ROBOTIC SYSTEM DESIGNED AS A GENERAL TOOL FOR COMPUTER ASSISTED ORTHOPAEDIC SURGERY G. Brandt1, A. Zimolong1, L. Carrat2, K. Radermacher1, S. Lavallée3, G. Rau1. 1Helmholtz-Institute for Biomedical Engineering, RWTH Aachen, Aachen, Germany, 2TIMC-IMAG Lab, University Joseph Fourier, La Tronche Cedex, France, 3PRAXIM, Grenoble, France. Objectives: The general objective of the CRIGOS project is the development of a compact surgical robot system for image guided orthopaedic surgery. Utilizing a modular architecture, the goal is to provide a robotic system which could serve as a general tool in CAOS and can be applied to a variety of orthopaedic applications. Background: The initial concept of CRIGOS and results of a first feasibility study were presented at CAOS’96 and other major meetings as well. It comprises a small and lightweight surgical robot which is guided by image data with an emphasis on calibrated x-ray fluoroscopy. Since 1998 the CRIGOS system is being developed on a European level with participation of two research centers, three industrial companies, and five clinics, all with great expertise in the area of Computer Assisted Surgery. Methods: To bring down the system’s complexity and to optimize the efficiency and effectiveness in handling this special surgical tool, a user-centered approach to system design was chosen. 15 different orthopaedic interventions proposed by medical experts were closely analyzed and ranked with regard to medical relevance, technical feasibility and cost/benefit ratio. For the seven highestranked interventions, different scenarios using the CRIGOS system were derived. Similar tasks in these applications were defined as “task macros” making it possible to easily set up new scenarios for other orthopaedic interventions. Using this methodology the requirements and interfaces (mechanical, electrical, information flow) of all functional components including technical equipment and human resources could be identified. This allows not only the realization of these modules at different locations throughout Europe, but also their reapplication in other CAS systems. Results: By the definition of task-macros, a modular system architecture could be created which allows for easy integration of other surgical interventions as well as the distributed development of the system. Based on this analysis, major improvements on x-ray-based surgical planning and an improved version of the compact robot could be realized in the framework of our research. First results concerning the integration of the robotic system into the OR will be presented. 214 Abstracts from CAOS ’99 ANIMATION OF IN VITRO BIOMECHANICAL TESTS P.A. Cripton1,2, Y. Bourquin1, M. Sati1, T.E. Orr1, G.A. Dumas2, L.-P. Nolte1,2. 1M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland, 2Department of Mechanical Engineering, Queen’s University, Kingston, Canada. Objectives: To develop a method to graphically replay in vitro biomechanical tests to allow intuitive interpretation of kinematic and force data independent of the viewer’s area of expertise. Background: In the interdisciplinary world of orthopaedic biomechanics it is often difficult to communicate complex kinematic, force and strain concepts so that they are understandable by all members of the interacting disciplines, such as surgeons, engineers, biologists etc. Design/Methods: Three 3.0 mm aluminum spheres were glued to each of the two vertebrae from a C2-3 segment of a human cervical spine. Before the biomechanical test, CT scans were made of the specimen (slice thickness 5 1.0 mm and slice spacing 5 1.5 mm). The specimens were subjected to a variety of biomechanical loads (such as moments, shear, and compression forces). During the biomechanical testing, a disc pressure sensor and bone-mounted strain gauges were used to estimate the force flow through the specimen. Kinematics were measured using an optoelectronic motion analysis system. The locations of the aluminum spheres were digitized using a space pointer and the motion analysis system. The CT images were reconstructed and the digitized aluminum sphere locations were used to match the CT and biomechanical test data. The biomechanical tests were “replayed” by animating the reconstructed CT models in accordance with the recorded experimental kinematics. Disc pressure and bone strain indicators were also placed in the animation image. Results and Conclusions: The animated test replays allowed sensitive and intuitive analysis of the biomechanical data. This technique improves the ability of experts from disparate backgrounds to interpret and discuss this type of biomechanical data. PC-BASED 3D SOFTWARE FOR PREOPERATIVE PLANNING IN RECONSTUCTIVE FOOT SURGERY C. Dahlen, H. Zwipp. Department of Trauma and Reconstructive Surgery, Universitätsklinikum Carl Gustav Carus der Technischen Universität Dresden, Dresden, Germany. Reconstructive foot surgery after complex trauma requires a detailed preoperative depiction of the injury to allow an exact planning of the corrective operation. Severe three-dimensional malalignment cannot be sufficiently assessed by either an extended series of standard x-rays or a two-dimensional CT or MRI. Only a real model (i.e., a stereolithography) enabled the surgeon to understand the complete pathology and to plan the corrective procedures. To facilitate the preoperative planning a 3D software was developed, which is able to visualize a 3D voxel data model with a standard low-cost PC in nearly real-time, and can be handled by every interested surgeon. CT or MRI data are downloaded by an adapted or a standard Dicom-III interface. After segmentation, the 3D-model may be transluminated, enlarged, and viewed from all directions. Any 3D-cut and 2D reconstructions in real-time are possible, as well as a stereoscopic 3D view with LCD shutterglasses and animations. Several subsegmented parts of the model can be moved continuously in all 6 degrees of freedom or can be placed using a special matching strategy. All manipulations are registered and a planning report is printed. The software allows simulation of various reductions or osteotomies, not only in the foot but in all parts of the skeleton. These procedures are directed either by anatomical guidelines and landmarks or by the mirrored image of the intact opposite side. Bony defects are depicted and measured as a volume, implants will be placed virtually and selected preoperatively. From the voxel data, conventional x-rays may be generated which can be compared later with intraoperative x-rays. The volume data may be used for external applications like stereolithography, production of individual prostheses, or intraoperative navigation systems. Further applications of the software are teaching and research. Abstracts from CAOS ’99 RECONSTRUCTION OF LARGE BONE DEFECTS OF THE NEUROCRANIUM WITH CUSTOM MADE IMPLANTS MANUFACTURED BY RAPID PROTOTYPING E. Euler1, S.M. Heining1, M. Buchloh2, D. Kotsianos3. 1Chirurgische Klinik, Klinkum Innenstadt der LMU München, Munich, Germany, 2BMW AG, Munich, Germany, 3Radiologische Klinik, Klinkum Innenstadt der LMU München, Munich, Germany. Objectives: Manufacturing of individually matched, mechanically stable, and biocompatible implants for the reconstruction of large bone defects of the neurocranium; Reduction of the operation time and avoidance of toxic and thermal influences on the brain surface by pre-production; Cost reduction by use of customary bone cement. Background: Osteoclastic trepanations due to major brain injuries often result in large bone defects of the neurocranium. The removed bone can be stored in frozen condition and replaced some months later after decline of the brain swelling. However, in some cases, the bone transplant is disintegrated, due to the biological (desmal bone) and mechanical (lack of load) situation. As an alternative, artificial material, e.g., bone cement, can be used to cover the defect. The manufacturing of the artificial bone lid during the operation has the following disadvantages: prolonged operation time; possible thermal or toxic damage done to the brain surface at contact during the modeling; mis-fit; and asymmetry to the contralateral side, therefore giving unsatisfactory cosmetic results. We describe a method that avoids these disadvantages. Design/Method: The basis of the presented method is a preoperative model of the skull in 1:1 scale, manufactured by means of Rapid Prototyping using the CCT-data of the patients skull. The skull defect is formed out of wax in line with the contour of the skull, then a silicone mold of this spare part is produced. The mold can be sterilized and used intraoperatively to produce an exactfitting lid from bone cement. Results: First clinical experiences show the following advantages: Clear reduction of the operation time due to pre-production of the implant. Contact of the bone cement with the brain surface during fitting and polymerization of the bone cement is avoided and thus biocompatibility is increased. Cosmetic results are excellent, and symmetry of the skull can be achieved. Conclusions: The reconstruction of bone defects of the neurocranium with custom-made implants by means of Rapid Prototyping represents a technically safe, biologically gentle, and functionally very good method. 215 TUMOR DEFECTS OF THE PELVIS AND THE ACETABULUM - VALUE OF CUSTOM-MADE PROSTHESES E. Euler, A. Betz, L. Schweiberer. Chirurgische Klinik, Klinkum Innenstadt der LMU München, Munich, Germany. Objectives: To show the indications and the technical possibilities of pelvic bone replacement with standard implants and custommade prostheses. Background, Methods and Results: The reconstruction of bone defects of the pelvic ring after tumor resection has different aims: Ventrally, stability is secondary. Herniations can be avoided by reconstruction of the pelvic ring and soft tissue wall with plates, allogenic flaps (i.e., gore-tex) and allogenic partial prostheses (i.e., poliacetal). The reconstruction of these defects is biomechanically unimportant. Bone defects of the central dorsal pelvic ring require stabilization by osteosynthesis material, with or without interposition of bone or bone cement. The resection of unilateral bone tumors of the dorsal pelvic ring without stabilization results in fibrous pseudarthrosis with acceptable functional result. Resection of the acetabular region without any reconstruction results in a “flail hip”. Leg length can be corrected by a “saddle prosthesis”. Arthroplastic measures after resection of the acetabular region provide functionally better results than strictly resectional methods. After type II resection, an anatomic-like reconstruction of the hip joint can be achieved with custom-made prostheses. Custom-made metal prostheses offer nearly unlimited possibilities of skeletal reconstruction. An example of pelvic ring replacement with integrated acetabulum is shown. Such reconstructions are risky and expensive. A preoperative analysis of the benefit-risk-ratio is therefore essential. Conclusions: Custom-made prostheses are indicated in cases of primary tumors or solitary metastases, if radical resection can be achieved, and if the functional restoration of the hip joint is essential. 216 Abstracts from CAOS ’99 SEMI-AUTOMATIC REGISTRATION TECHNIQUE FOR IMAGE-GUIDED PLACEMENT OF ILIOSACRAL SCREWS N. Glossop1, R. Hu2, G. Dix2, Y. Behairy2. 15116 Bissonnet St., Suite 324, Bellaire, TX 77401-4007 USA, 2Foothills Hospital, Division of Orthopaedics and Neurosurgery Surgery, University of Calgary, Calgary, Canada. Introduction: A method for semi-automatically registering rigid anatomy has been developed and demonstrated for the placement of iliosacral screws. The technique makes use of unique properties of the new passive “Polaris” tracking system (Northern Digital, Waterloo, Canada). This position sensor tracks the position of passive retroreflective spheres attached to a rigid tracking frame. Materials and Methods: A special plastic passive tracking frame containing several embedded radio-opaque fiducials was used to dynamically reference the pelvis. The plastic frame was fixed to the pelvis with a screw prior to the scan. Following scanning, the data was downloaded to a commercial IGS workstation, the “SCOUT” (SNN, Mississauga, Canada). Precise knowledge of the fiducials’ position gained from a priori knowledge of their relationship with the coordinate system of the dynamic tracking frame enabled the registration to be made automatically. A tracked drill guide was then used to position wires in the iliosacral joint. Results: Both screws were placed accurately into the pelvis and avoided critical structures including the cauda equina, L5 and S1 nerve roots. By relieving the surgeon of the need to register, the procedure was conducted quickly and efficiently. The technique also has the potential of being more accurate than current techniques because no error is introduced by the tool-optical tracker during registration. Conclusions: The new technique is a fast and accurate method of registration whenever a rigid body can be affixed to the anatomy prior to scanning. Although we have demonstrated it for sacro-iliac screw fixation, this same methodology can be extended to include almost any IGS procedure, including cranial registration, long bone registration and spine registration. COMPARING ELECTRO-OPTICAL AND ELECTROMAGNETIC GUIDANCE IN THE PREPARATION OF MINIMAL ACCESS SURGICAL TRAINING ENVIRONMENTS S. Grange1, T. Bunker2, J. Cooper3, S. Waldhausen4. 1,2Princess Elizabeth Orthopaedic Centre, Exeter, United Kindom, 1/3Department of Computer Science, University of Exeter, Exeter, United Kingdom, 4Forschungszentrum Karlsruhe GmbH, Karlsruhe, Germany. Introduction: High quality visual images for virtual environment generation were prepared using two comparative spatial coordinate referencing systems, identifying the position of individual frames in the recorded video database. Methods: A cadaveric shoulder arthroscopy database was recorded and replayed using an ATM Network. An electro-optical (OptotrakTM) spatial reference system is compared with an electromagnetic (Flock of BirdsTM) system, for position tracking of an arthroscope in the virtual environment. Results: 1. Electro-optical equipment is used safely in the operating theater environment 2. Electromagnetic devices are affected by ferromagnetic environments 3. Electromagnetic devices are inexpensive, and thus more suitable for a training environment 4. It was necessary to register the two systems relative to each other using “public domain” algorithms. The precision of the electro-optical equipment exceeds the current specification of the video world database matrix granularity and the amplitude of physiological oscillation of the manual operator. Discussion: The dataset may be integrated into a simulated training environment which may then be accessed in a non-critical training environment where it is cost-effective to use electro-magnetic position-sensing devices. Interchangeability between such systems is based upon the mathematical analysis available from our website – www.dcs.ex.ac.eu/evw. Conclusions: The electro-optical system allows accurate registration of the frame recording position and playback, however, high capital cost may prohibit widespread use for training. Developing mathematical correlation between the two systems integrates their specific use in different environments. Abstracts from CAOS ’99 217 A VERSATILE IMPINGEMENT DETECTION ALGORITHM EVALUATION OF IMAGE TRANSFER IN TRAUMA CARE Q. Hu, U. Langlotz, J. Lawrence, F. Langlotz. M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland. B. Ilango1, V. Shah, J.F. Haines, J.B. Day, A.P. Maskell, P.R. Kay. Departments of Orthopaedic Surgery, Trafford General Hospital, Manchester, Booth Hall Hospital, Manchester, Blackburn Royal Infirmary, Lancashire, Preston Acute Hospitals, Lancashire, and1The University Department of Orthopaedic Surgery, Manchester, United Kingdom. Objectives: Developing a versatile impingement algorithm suitable for medical applications. Background: Previously-published research on impingement (collision) detection methods is not suitable due to two factors: polyhedral approximation of essentially non-linear anatomic models, and non-availability of geometric and temporal coherence between initial and final states of objects. Design/Methods: Suppose we have two objects, Obj1 and Obj2, and Obj2 moves with respect to Obj1. The developed algorithm enables us to obtain the surface representation of Obj1 and Obj2 from their anatomic CT-scan data, to build a look-up table based on the spatial extents of Obj1, to calculate the spatial indices for all the transformed surface points of Obj2 according to a unique linear transform that maps 3D coordinates to a spatial index, and to search through the look-up table to locate impingement. Results: For a phantom consisting of a base and a shaft, the overall accuracy error was less than 1.5 mm. For a patient data set with left pelvis and femur (surface points were 102, 885 and 629, 964 respectively), initialization time was 175 seconds, and the impingement for any relative movement could be reliably detected within 85 milliseconds. Conclusions: We have developed a versatile impingement detection algorithm, based upon look-up table and linear transform. The method is general purpose in the sense that objects can be of any shape and it can be extended to many objects in the scene. Background: Telemedicine as a communication tool is found to be useful in providing improved quality healthcare, but is expensive, requires specialist expertise and set-up, and is not accessible to the majority of clinicians. The internet and digital imaging are widely available and relatively cheap technology which may be used to enhance the quality and effectiveness of health care. We investigated a simple method of capturing radiographs and clinical images by a digital camera in a casualty department and transferring the images by e-mail through the internet to make a diagnosis and plan treatment for trauma patients. Methods: We studied 53 consecutive patients, who ultimately required admission to hospital for the treatment of their injuries. The radiographs were digitized using a digital camera and x-ray light box in the A&E department. Clinical photographs of patients’ injuries were also taken. All images were transmitted by e-mail through the internet to five orthopaedic surgeons experienced in trauma care. Results: All images were successfully transmitted. Images were received within three minutes of transmission. In all cases, an accurate diagnosis could be made and treatment plan suggested by viewing the images alone. The images and the radiographs matched closely in 96% of cases. There was minor variation between the observers in the scoring of images and x-rays but the diagnosis and management plan were consistent for each individual consultant. Conclusions: We found that digitization and transfer of images can be done successfully in a casualty department by using a digital camera, computer and e-mail. The process is simple, transmission is easy and quick, and transmitted images are easily retrieved and seen within a few minutes. This method supplements the alreadyexisting telephone communication amongst clinicians, which involves discussion of patients condition and/or the x-ray findings. We feel that images used in conjunction with telephone conversation can be of good educational value for the trainees. This method can be particularly useful when dealing with complex trauma, where a second opinion is required from a relevant specialist. The image quality in this study was good enough to make a diagnosis and treatment plan in all the cases, and we feel that this method may be used in other branches of medicine, surgery, pathology etc., to achieve better care for patients by providing access to specialist advice from any part of the world. 218 Abstracts from CAOS ’99 COMPUTER-ASSISTED MICROSURGERY: INTRODUCTION OF A NEW ROBOTIC ARM PEDICLE SCREW PLACEMENT USING COMPUTER ASSISTED TECHNIQUE B.D. Krapohl1, J.E. Zins2, M. Siemionow2. 1Orthopedic Center, Kassel, Germany. 2Department of Plastic and Reconstructive Surgery, The Cleveland Clinic Foundation, Cleveland, OH, USA. P. Merloz1, J. Tonetti1, A. Eid1, F. Fontanel1, S. Lavallée3, J. Troccaz2, P. Cinquin2, L. Pittet2. 1Centre Hospitalier Universitaire, Grenoble, France, 2TIMC-IMAG Lab, University Joseph Fourier, La Tronche Cedex, France, 3PRAXIM, Grenoble, France. Objectives: We evaluated a new RAMS (robot assisted microsurgery) workstation designed by the Jet Propulsion Laboratory, NASA, Pasadena, California, USA, in different microsurgical procedures. Methods: The robotic arm (slave robot) housed the instrument, a microsurgical forceps. The arm comprised five joints, each allowing motion in two directions. The robotic arm was operated by the surgeon with a joystick (master robot). The robotic arm could be adjusted by setting parameters for scale of motion and tremor filter. The RAMS workstation was evaluated in the following microsurgical tasks: 1. 2. 3. 4. 5. 6. 7. Removal of a foreign body. Removal of an intravascular thrombus Intravascular injection Insertion of an intravascular catheter Small vessel dissection Ligation of small vessel side branches Microsurgical anastomosis. Results: In tasks 1 to 4, the robot functioned as an operating as well as an assisting instrument. In tasks 5 to 7, the robot could merely be used as an assisting tool because of its poor rotational function. Advantages of the system were its precise functioning void of any tremor, which was important especially when tissue or instruments needed to be held for a longer time period, as well as its ability to replace an assisting person to some extent. Deficiencies of the system were its 15-minute warm-up, the spatial conflict within the operative field, and the poor rotation of the robotic tip. Conclusions: RAMS workstation could replace the surgeon’s second hand, or it could function as his third hand with precision and absence of tremor. Aim of this study: The aim of this study is to improve the reliability of pedicle screwing. Transpedicle screw insertion may cause three types of complication: neurologic, vascular and mechanical. Previous studies of surgical procedures have shown a significant rate of incorrect placement of the screw ranging from 10% to 40%. Material and methods: A new technique that combines preoperative computed tomography (CT) imaging with intraoperative passive navigation has been used to perform 66 pedicle screwings in the thoraco-lumbal region. In the same period, 66 pedicle screwings were performed manually in the same region and on the same vertebral levels. Surgery was followed in all cases by postoperative radiographs and computed tomography examination, on which measurements of screw position relative to pedicle position could be done. Results: Comparison between the two groups showed that six screws in 66 vertebra (9%) had incorrect placement with computer assisted technique, whereas 30 screws in 66 vertebra (45%) had incorrect placement with manual insertion (P,0.0001). The intraoperative accuracy provided by the computer after registration was better than 1 mm. Discussion: These good results are similar to the results obtained by Foley and Nolte. The cortex penetration observed with the computer assisted technique were not imputed to computer failures. Errors in acquiring data by the surgeon in the pre- and peroperative steps may explain the six incorrectly-placed screws. Conclusions: This clinical experience shows that the accuracy and the reliability of this computer assisted technique are good. Abstracts from CAOS ’99 3-D ULTRASOUND USED FOR THE VISUALIZATION OF THE NEWBORN’S ENTIRE HIP JOINT H.M. Overhoff1, D. Lazovic2, U. von Jan1. 1Department for Computer Science, University of Hildesheim, Hildesheim, Germany, 2 Orthopedic Department, Medical School Hannover, Hannover, Germany. Objective: To improve the diagnostic value of an existing ultrasound (US) screening method by application of 3-D US. Background: The developmental dysplasia of the hip (DDH) is the most frequent congenital skeletal malformation. Graf established a screening method for DDH in Central Europe based on indicators derived from interactively determined landmarks in a defined single plane US scan. The use of Graf’s method reduced undetected DDH drastically, but the method has its drawbacks. It demands high skills from the examiner, the femoral head is not clearly visible and the indicators do not reflect the spatial joint structure. Methods: The hip joints of 13 newborns (aged 2-18 weeks) were scanned by free-hand sweeps using a conventional US equipment completed with a self-developed 3-D US imaging system. By automatic image segmentation the cartilaginous and ossified joint structures and the landmarks of the 2-D US diagnostic procedure were detected. Results: The femoral head and acetabulum were visualized and color-coded on a computer screen. The automatic segmentation to achieve this visualization was highly correct (mean absolute error 2 pixels (0.24 mm)). Conclusions: The visualization of the spatial relationship of the acetabulum and the femoral head is possible by means of 3-D US. Further work should address the development of an indicator quantifying the spatial aspects (concentricity, head coverage) of DDH. 219 3-D SOLID REAL RAPID PROTOTYPING MODELS IN ORTHOPAEDIC SURGERY: CLASSIC AND NEW INDICATIONS P. Peters1, W. Mittelmeier2, R. Gradinger2, W. Plötz1. 1Klinik für Orthopädie, Medizinische Universität zu Lübeck, Lübeck, Germany, 2Klinik für Orthopädie und Sportorthopädie der TU München, Munich, Germany. Introduction/Background: After extensive experiences with Rapid Prototyping in tumor surgery, primary and revision surgery of THA, and in triple osteotomies of the pelvis, this technique is now also used for planning and simulating pelvic surgery in Computer Assisted Orthopaedic Surgery. Materials and Methods: From 4/93 to 12/98, 140 solid 3-D models on a 1:1 scale were produced using 3-D-CT data. 107 3-D pelvic models were produced for our own hospital. Indications were hip dysplasia (59 patients), pelvic tumors (35 patients), severe degenerative changes (8 patients), and acetabular loosening in THA (5 patients). Results: In pelvic tumor surgery with internal hemipelvectomy and hemipelvic endoprosthesis, solid 3-D models are standard practice allowing for exact preoperative planning and precise intraoperative orientation, as well as for a custom-made endoprosthesis. In the case of loosening of acetabular components of THA with large bone deficiences, an exact reconstruction using custommade implants (cranial cup) is possible. In triple pelvic osteotomies, preoperative planning defining the placement of the acetabulum and precise intraoperative osteotomies are feasible. Experimental studies with navigation systems using 3-D real pelvic models show the feasibility of computer assisted osteotomies and a more exact positioning of osteosynthesis material. Conclusions: 3-D solid pelvis models on a 1:1 scale manufactured using 3-D-CT data are a key to success in difficult surgery of the pelvis. They are now being successfully used in experimental CAOS studies. New techniques are introducing preoperative planning and simulation in the virtual 3-D world. 220 Abstracts from CAOS ’99 DOUBLE BUNDLE ANATOMIC ACLRECONSTRUCTION WITH CAS APPLICATION ACCURACY OF LINEAR STEREOTACTIC TOOLS A.B. Wymenga1, J. Luites1, L. Blankevoort2, R. vd Venne2, M. Sati3 and H.-U. Stäubli4. 1Department of Orthopedic Surgery, St. Maartenskliniek, Nijmegen, The Netherlands, 2Orthopedic Research Laboratory, University of Nijmegen, The Netherlands, 3 M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland, 4Tiefenauspital, Bern, Switzerland. C. Amstutz, L.-P. Nolte. M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland. Introduction: The use of CAS to improve accurate femoral tunnel placement during arthroscopic double bundle ACL-reconstruction requires a 3D template of the femoral AMB- and PLB-insertions of the anterior cruciate ligament (ACL). Background: In computer aided surgical navigation systems, all tracked components such as the surgical object and the stereotactic tools are assumed to be rigid bodies. However, during surgical action, some of the components may undergo considerable deformation. For slender linear tools, this may result in significant application errors. Design/Methods: 28 femoral AMB- and PLB-insertion sites were 3D digitized, as were the surface of the notch and the cartilage border of the lateral condyle. The notch surface was approximated by a cylinder, fitted on the surface data. The projected cartilage border-data determined the cylinder location. The calculated AMBand PLB-insertion centers were also projected on the cylinder. All 28 cylinders were scaled and averaged to a mean template. A curve was fitted through all cartilage border-data, resulting in a function describing the mean border. The mean position with regard to the top of the cartilage border of all 28 AMB- and PLB-centers was calculated. Results: The averaged notch surface can be modeled by a cylinder with a mean radius of 10.5 mm. The function which describes the cartilage border fits with an explained variance of 75%. The mean position of the AMB center is at 11 mm anterior and 7 mm proximal from the top of the cartilage border; the PLB center is at 7 mm and 4 mm. Conclusion: With CAS and the developed template, the anatomic AMB- and PLB-center of the individual joint can be localized within 4mm. Objectives: To provide quantitative estimates for the range of application error during the use of slender linear tools such as drill bits in surgical navigation systems. Design/Methods: Different surgical drill bits have been chosen (Synthes product line), providing a representative spectrum of slender linear instruments. A custom setup consisting of a device holding a plastic bone block and a commercial stereotactic drilling machine (Medivision, Oberdorf, CH) was used. Components were tracked at a frame rate of up to 350 Hz. A mathematical model provided the relations between drill-bit parameters, such as diameter and length, and the associated deformation. Results: Setup and frame rate are appropriate to estimate deformation. Deflections typically range from 5 to 10 mm, but may be up to 20 mm. As is expected, the extent of deformation is strongly dependent on the user’s specific way of applying the drill. From the mathematical model follows the upper deflection limit for an arbitrary drill-bit geometry. Conclusions: Deformation of slender stereotactic instruments may significantly reduce the desired application accuracy. However, proper understanding of how to reduce excessive forces and moments during the drilling process, as well as the use of additional cannulated guides, in particular for extremely thin and long tools, may avoid possible surgical complications. Abstracts from CAOS ’99 SESSION VIII: SIDE ASPECTS AND NEW AREAS OF CAOS INTRAOPERATIVE 3-D RECONSTRUCTION AND GEOMETRIC MODELING OF THE SPINE USING BIPLANE X-RAYS 1 1,2 1 1 F. Cheriet , C.-É. Aubin , S. Delorme , H. Labelle , J. Dansereau1,2; J.A. de Guise1,3. 1Sainte-Justine University Hospital, Montréal, Canada, 2Department of Mechanical Engineering, École Polytechnique, Montréal, Canada, 3École de Technologie Supérieure, Montréal, Canada. Objectives: To develop intraoperative 3D reconstruction and geometric modeling techniques to evaluate surgical correction of scoliotic deformities based on only two x-rays. Background: A CT scan is generally used to build geometric models for intraoperative tracking, although it involves high radiation when used for long spinal segments. Traditional radiographic reconstruction methods based on the Direct Linear Transformation technique (DLT) are not suited for intraoperative use because they require a large calibration object that completely encloses the patient trunk. Design/Methods: During the surgical procedure, a sterile 16 x 18 cm aluminum calibration object with 12 markers is put on the patient before taking the postero-anterior and lateral x-rays. The markers are digitized on each radiograph, and the DLT technique is used to compute an initial estimate of the geometrical parameters of the radiographic setup (source-to-film and source-to-calibrationobject distances, etc.). An optimal estimate of the geometrical parameters is computed from the non-linear minimization of the mean square distance between the observed and analytical projections of the markers on the pair of radiographs. Six points per vertebra are reconstructed and used as control points to deform generic detailed vertebrae (using geometrical transformations) and to build a geometric model of the spine. Results and Discussion: A first in vitro validation was done on a non-pathologic dry cadaveric human spine. The accuracy of the 3D reconstruction was similar to the one obtained from the DLT technique (2-3 mm). Accuracy of the geometric model was about 3 mm. An in vivo evaluation made on a scoliotic patient showed that this technique gives better results than the DLT technique in a real surgical environment, with half the RMS reconstruction error on the calibration object. We believe that the proposed approach will be useful to improve the evaluation accuracy of the 3D correction obtained during the surgical procedure. 221 “MORPHING” THE SPINE: SEGMENTATION AND TRANSFORMATION OF PREOPERATIVELY-ACQUIRED IMAGE DATA SETS K.T. Foley, R.S. Teichman, K.R. Smith. Image-Guided Surgery Research Center, Memphis, TN, USA. Objectives: To investigate the feasibility of updating a threedimensional image data set to reflect intersegmental spinal motion occurring between the time of image acquisition and surgery. Background: Current techniques for image-guided spinal surgery do not allow for accurate simultaneous registration of more than one spinal segment. Design/Methods: Atlantoaxial CT scans were obtained in fresh cadavers using 1 mm slice thickness and 14 cm FOV. The scan data were then segmented into individual C1 and C2 vertebrae in an automated fashion through mapping from a digitized, annotated cervical spine atlas, utilizing a process that has previously been used for brain mapping. The C1 and C2 segments were reduced and immobilized with sublaminar cables, then individually registered utilizing a contour-mapping algorithm. Based upon segmentation of the preoperative data set as well as segmental registration data from surgical space, the image space data set was transformed. Results: The transformed image space data set display rendered C1 and C2 in their “new” relative positions, reflecting their relationship in surgical space, post-reduction. Optimal transarticular screw pathways could then be plotted and tracked in real-time through both C1 and C2. Conclusions: A segmented spinal image data set can be transformed based upon intraoperative registration, allowing computerized display of the “new” spatial relationships of the vertebrae at the time of surgery. We conclude that it is feasible to update a preoperatively acquired image data set to reflect changes in rigid body spatial relationships brought about by patient positioning and/or surgical manipulation. 222 Abstracts from CAOS ’99 ERGONOMIC EVALUATION OF SYSTEMS FOR COMPUTER ASSISTED SURGERY APPLICATION-SPECIFIC ACCURACY REQUIREMENTS FOR IMAGE-GUIDED SPINAL SURGERY A. Zimolong, K. Radermacher, C. Hachmöller, G. Rau. Helmholtz Institute for Biomedical Engineering, RWTH Aachen, Aachen, Germany. Y.R. Rampersaud, D.A. Simon, K.T. Foley. Image-Guided Surgery Research Center, Memphis, TN, USA. Objectives: In order to assure an efficient, effective, and safe handling of systems for computer assisted surgery, it is necessary to assess the ergonomic design. Our portfolio of tools and methods for evaluation of ergonomic criteria adapted from human factors research to CAOS is presented, together with first evaluation results. Background: “Ease-of-use” is often claimed to be a quality criterion for CAOS systems, but is hardly ever proven. In contrast, human factors research provides sound criteria and measures for ergonomic design of usable systems. Adapting ISO9241 to CAOS, usability may be derived as the extent to which the system can be used by the surgeon to achieve his surgical goals, such as repositioning of the acetabulum or fixation of pedicle screws according to the surgical planning, with effectiveness, efficiency, and satisfaction. Methods: In order to measure the usability of a planning system for individual templates and for a commercial system for computer assisted spine surgery, we selected a twofold approach: First, utilizing a software tool for standardized evaluation, compliance with established standards was measured. Second, physicians and medical students were asked to perform a given task while being monitored by an eye-tracking system coupled with a video camera for event-logging. Questionnaires and interviews to capture user characteristics and satisfaction completed this approach. Results: By performing the guideline-based evaluation method, flaws and inconsistencies as well as non-conformance with standards could be detected and led to a list of recommendations for re-design of the human-computer interface. User-based evaluation of these recommendations could prove on a case base that handling of certain aspects of the system was improved. Results from the second evaluation step confirmed previous findings, and revealed further conceptional problems of the users at some stages of the planning process. However, it could be demonstrated that first-time users are able to plan individual templates for surgery without training and help within 15 minutes. Conclusions: From these results it could be shown that the selected approach is able to detect systematically flaws and inconsistencies of the system under evaluation. Because of these specific findings, recommendations to tackle these shortfalls can be worked out. Objectives: To derive theoretical translational and rotational accuracy requirements for image-guided (IG) spinal screw placement using existing morphometric data. Background: Previous studies have examined the clinical accuracy of particular IG systems, underlying causes of inaccuracy in these systems, and methods for quantifying this inaccuracy. However, application-specific accuracy requirements have not been adequately addressed. Design/Methods: We developed a geometric model relating spinal anatomy to accuracy requirements for IG surgery. Modeling the pedicle as a cylinder, we determined the accuracy required to avoid pedicle perforation at its narrowest point using clinically-relevant screw diameters. Results: An inverse relationship was found between allowable translational and rotational errors for safe screw placement. As anticipated, accuracy requirements were greatest at levels where the requisite screw diameter approximated the dimensions of the pedicle. These requirements were highest for T5, followed in descending order by T4, T7, T6, T3, T12, L1, T8, T11, C4, L2, C3, T10, C5, T2, T9, C6, L3, C2, T1, C7, L4 and L5. Maximum allowable error ranged from 0.0 mm/0.0° at T5 to 3.8 mm/22.7° at L5. Conclusions: This study demonstrates that extremely high accuracies are required to safely place pedicle screws at certain levels of the spine. In some cases, these accuracies are beyond the capabilities of existing IG systems. We hypothesize that other factors, such as the mechanical constraints imposed by the pedicle wall and the surgeon’s tactile feedback, are critical for achieving the improved clinical accuracy of image guidance systems demonstrated in the literature. Abstracts from CAOS ’99 “MATCHING ERROR” – WHAT IS THIS NUMBER? R. Bächler1,2, H. Bunke2. 1M.E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland, 2Institute for Computer Science and Applied Mathematics, University of Bern, Bern, Switzerland. The process of establishing the mathematical correspondence between different image modalities of the patient or between the patient and his images is termed “registration”. Often, registration is also called “matching”, as a matching function is used to calculate the transformation that represents the mathematical correspondence between modalities. In computer aided surgery (CAS), the surgeon should get some sort of feedback on the quality of the registration, so that he knows, e.g., the accuracy of the tool representation on the computer screen with respect to the tool location in situ. Most CAS systems provide this feedback as a number that represents a matching “error” or registration accuracy. This number often just indicates the quality of the numerical calculation of the transformation, and it is very difficult to correlate this number to the accuracy of the system in order for the surgery to be performed. As the computer has no a priori knowledge of the “true” transformation, the performance of the matching algorithm depends heavily on the quality of the input data. If the input data is not very well defined, it might become difficult for the matching algorithm to calculate a reasonable transformation. Different approaches exist that aim at giving the surgeon a better understanding of what the computer actually calculated, and how he might improve the accuracy. One approach is to indicate a volumetric target error, i.e., to give an upper boundary for the deviation between the actual tool location in the target area and its representation on the screen. Another approach is to guide the surgeon through the data acquisition process for the registration, so that the probability of a poor registration is minimized. However, there is still much work to be done before a CAS system can comprehensively tell the surgeon how accurately and reliably the registration has been achieved. 223 IS THERE A CONTAMINATION RISK FOR THE SURGICAL TEAM THROUGH ROBODOC’S HIGHSPEED BONE CUTTER? M. Nogler1, C. Wimmer1, C. Lass-Flörl2, E. Mayr1, C. Bach1, M. Krismer1. 1Department of Orthopaedic Surgery, University of Innsbruck, Innsbruck, Austria, 2Department of Hygiene, University of Innsbruck, Innsbruck, Austria. Purpose: During the cutting of the femoral cavity in the ROBODOC procedure, we observed an aerosol cloud of irrigation fluid, blood and tissue debris. If this cloud is contaminated with bacterial and viral vectors it can be a potential source of contamination for the operative team. The aim of this study was to detect the extension of this aerosol cloud. Methods: Not being supplied with a sufficient irrigation system (until Summer 1998), the necessary irrigation for the ROBODOC® procedure is performed by a continuous flow through an iv-needle ( 1.5 mm, 18G). Additional fluid is added with a syringe towards the cutter’s top. On top of the femur, the cutter moves through a basin formed by the bone which is filled with blood and irrigation fluid. We tested a flat cutter (P/N 100923) and a ball cutter (P/N 100869) in four standard situations. For macroscopical detection we used artificially colored Nigrosin-solution. In a second experimental setting, we exposed the cutter to a fluid contaminated with staphylococcus aureus and detected the bacterial contamination using standard cultures. Results: We detected the aerosol cloud in an area of 6.0 x 3.6 m. Extension, concentration, and shape of the covered area varied depending on the irrigation situation. Cutting in a fluid basin was found to be the most extreme setting for both cutter types, with the highest vector concentration occurring in the cutter’s turning direction. A difference between the ball and flat cutter was only seen in the irrigation situation with the iv-needle. The vector room concentration in all settings was 1.6 x 104 CFU/ml. Conclusions: ROBODOCS’s high speed cutter produces an aerosol cloud in an area in which all members of the surgical team, sterile and non-sterile, are affected. We conclude that a potential contamination risk for the surgical team exists. Water-resistant clothing and face protection is therefore necessary for everybody in the operating room. 224 Abstracts from CAOS ’99 USE OF ULTRASOUND TO DETERMINE LEG GEOMETRY ENDOSCOPIC NAVIGATION AND REAL-TIME IMAGE PROCESSING P. Keppler1, W. Strecker1, F. Gebhard1, M. Simnacher2, L. Kinzl1, L. Claes2. 1Department of Traumatology, Hand and Reconstructive Surgery, University of Ulm, Ulm, Germany, 2Institute of Orthpaedic Research and Biomechanics, University of Ulm, Ulm, Germany. W. Konen1, M. Scholz2, S. Tombrock1. 1Zentrum für Neuroinformatik GmbH, Bochum, Bochum, Germany, 2Neurochirurgische Universitätsklinik, Ruhr-Universität Bochum, Bochum, Germany. Objectives: Presentation of a new computer aided three-dimensional ultrasound system to measure the torsion, length and axis of the lower extremities in one sitting. Background: Post-traumatic deformities of the lower extremities are not uncommon. Today, the Gold Standard for determining the torsion and length of the lower extremities is the CT, which suffers from projection errors and radiological hazard. Design/Methods: We integrated a conventional ultrasound diagnostics machine (SIEMENS Sonoline 400) with an ultrasound navigation system (ZEBRIS CM 6/60) and a personal computer. This combination enables us to determine the position of the 5 MHz linear transducer in 3-D space. Through custom software running under Windows NT, we can save the ultrasound scan and the three-dimensional position of the transducer simultaneously. The bony surfaces of the femur and tibia are the anatomical landmarks which define the reference planes, the torsion, length and axis. With the computer mouse we mark the necessary points of interest. Because the reference planes are defined on the bony contours of the femur and tibia, the system is independent of the position of the patient. Markers fixed on the lower leg monitor patient motion during the investigation. Results: We examined 50 adults with the ultrasound system who had a normal leg geometry as determined by CT in the last 4 years. Mean difference between the two methods in torsional measurement was 3 degrees (max. 8 degrees); in length, 2 mm (max. 7 mm); and in leg axis, 2 degrees (max. 3 degrees). The inter- and intraobserver error of experienced ultrasound investigators was 1/-2 degrees (sd) in torsion and 1/-2mm (sd) in length. Conclusions: This new ultrasound method is an accurate method to measure the leg geometry in one sitting without radiological hazard or projection errors. It is therefore recommended for screening investigations and preoperative planning. The motion-tracking sensors on the lower leg largely negate movement effects, making the system especially appropriate for children. A new visual navigation support system (VNS) for endoscopic interventions allows extraction of 3D-information from endoscopic video data and superimposition of 3D-information onto such live video sequences. Materials and Methods: The endoscope is coupled to a position measurement system and a video camera as components of a calibrated system. We can measure and display anatomical landmarks of the patient as viewed from the current position of the camera. Another module, the so-called virtual map, allows the storage of endoscopic images (2-4 images per second) in map form and - in the case where the direct view is lost (bleeding) - the retrieval of virtual images, thus helping the surgeon to maintain a visual impression of the operating field. A special version of the virtual map module with red detector allows identification of the source of bleeding in “red-out” situations using previously-stored images. This makes an intervention (coagulation) possible for the first time in such situations. Results: The navigational part of the VNS has been used so far in 10 endoscopic interventions with good success. The coagulation without direct view was successfully tested in animal experiments (175 bleeding situations), resulting in 89.1% successful bleeding stoppages. Conclusion: An endoscopic navigation system has been developed which offers the surgeon various modules by directly accessing the image information. These new modules could lead to a revolutionary advance in image-guided surgery. As a spin-off, the VNS can be used as a comfortable video archiving and retrieval system for high-quality endoscopic images, and as a system for training by navigating through endoscopic images from previous operations. Abstracts from CAOS ’99 225 CAOS: CHALLENGES FOR THE OPERATING ROOM OF THE 21ST CENTURY SESSION IX: COMPUTER ASSISTED KNEE ARTHROPLASTY AND OSTEOTOMY R.D. Bucholz. J. Bakewell Section of Image Guided Surgery. Division of Neurosurgery, Department of Surgery, St. Louis University School of Medicine, St. Louis, MO, USA. VIRTUAL FEMUR CREATED FROM ULTRASOUND IMAGE VOLUMES FOR PLANNING OF TOTAL KNEE ARTHROPLASTY Objectives: To delineate the problems preventing the widespread adoption of computer assisted orthopedic surgery within the next decade. M. Liebing1, D. Lazovic2, H.M. Overhoff1. 1Department for Computer Science, University of Hildesheim, Hildesheim, Germany, 2 Orthopedic Department, Medical School Hannover, Hannover, Germany. Background: It is now common for neurosurgeons to routinely employ computer assistance and navigational systems for all cranial procedures, even in small community hospitals. However, only a few orthopedic surgeons use computer assistance at all, much less routinely. The needs of an orthopedic surgeon are quite different than those of a neurosurgeon, and therefore it is not reasonable to expect that systems developed for neurosurgeons will be appropriate for orthopedic procedures. To address how this technology can be migrated to orthopedics, the components of a standard image guided procedure consisting of preoperative imaging, image registration, intraoperative tracking, visualization, system control, effectors/instruments, and intraoperative imaging will be considered. Suggestions for improvements and changes in each component will be discussed that would make it more likely that these systems will be used in even routine orthopedic procedures. Conclusions: Navigational systems have much to offer orthopedic surgeons in terms of improvement in safety, decreased invasiveness, and, eventually, time saving for orthopedic interventions. However, basic technological improvements, which represent opportunities for engineers working in this area, will be required, along with changes in the operating room infrastructure, before computer assistance will become a part of routine orthopedic care. Objective: To investigate the feasibility of a determination of size and position of a femoral prosthesis by interactive planning based on virtual distal femur models, which are generated from 3-D ultrasound (US) image volumes and automatic image processing. Background: Incidence of Total Knee Arthroplasty (TKA) is increasing. The main problem is the correct placement of the femoral prosthesis. Pre-operative planning and intra-operative realization relies on lower-limb X-rays and extra- or intramedullary devices. This may lead to malplacements. Methods: We generated virtual models of the distal femur for 10 knee joints. In a free-hand sweep, we recorded 3-D US image volumes, which typically took 90 sec. The bony structures were detected automatically in the US images, and then appeared as curvilinear lines. Spatially arranged, these lines formed femur models, which were used for the placement of a femoral prosthesis in an interactive computer-based 3-D planning. Results: The automatic image segmentation took about 45 minutes. Its analysis showed correct detection of the bone surface (mean distance error 4 pixels (0.48 mm), 75%-quartile 9 pixels (1.08 mm)). The interactive planning took 5-10 minutes. Conclusions: The results demonstrate that 3-D US can provide data on position and size of bony structures. Bone models created by analysis of 3-D US image volumes are feasible for simulation of TKA implantations. 226 Abstracts from CAOS ’99 COMPUTER ASSISTED KNEE ARTHROPLASTY: PRELIMINARY CLINICAL RESULTS WITH THE ORTHOPILOT SYSTEM A NEW APPROACH IN INDIVIDUAL KNEE JOINT REPLACEMENT USING SUBTRACTION PROCESSING OF THREE-DIMENSIONAL CT-SCAN DATA F. Picard1,2, D. Saragaglia1, E. Montbarbon1,3, C. Chaussard1,3, F. Leitner4, O. Raoult4. 1Hopital Sud, Grenoble, France, 2Northwestern University, Chicago, IL, USA, 3Hopital de Chambery, 4Praxim, Grenoble, France. W. Thoma1, C. Schuster2, R. Wuttge3, L. Zichner2, L. Schuster3. 1 Orthopädische Praxis Thoma-Götz, Frankfurt, Germany, 2Orthopädische Universitätsklinik Frankfurt am Main, Frankfurt, Germany, 3Radiologisches Institut Wuttge-Hannig-Sindelar, Munich, Germany. Introduction: We tested this system on a series of 30 patients with primary osteoarthritis of the knee (15 patients were operated with a classical procedure, SEARCH AESCULAP Knee, and 15 patients with the ORTHOPILOT procedure). The trial was a prospective randomized parallel study carried out in Grenoble from January 13 to December 1, 1998. Patients: 30 patients between 55 and 89 years of age (mean 69). 14 right knees and 16 left knees were included in the study after checking eligibility. Method: The same surgeon (Saragaglia) operated all the patients. Two independent surgeons followed up the patients 6 weeks after surgery. Results: Radiological outcomes: (long leg X-ray) Femorotibial angle: min 3 degrees varus, max 3 degrees valgus, Usual technique (UT) 10 patients (66.6%), Computer Technique (CT) 15 patients (100%). Clinical outcomes: Duration of surgical procedure: UT average 74 min (55-100 min), CT average 101.6 min (80-130 min). Duration of tourniquet: UT average 73.3 min (60-90 min), CT average 92.3 min (75-120 min). Post operative bleeding: UT average 414.6 ml (100-890), CT average 490 ml (100- 790); 4 patients in the UT group had post-operative bleeding in excess of 790 ml, the highest level in the CT group. Complications: UT 3 cases with complications (2 TE and 1 stiffness), CT no complications. Conclusions: These results answered our requirements for accuracy. Obviously, the series was too limited to draw final conclusions. However, these results for computer assisted knee arthroplasty seem very encouraging. Objectives: Bone destruction caused by tumor, or septic or aseptic loosening of endoprosthesis may cause problems which cannot be sufficiently solved by means of conventional implants. In addition, the morphological differences in the anatomic shape of the knee joints may require a custom-made endoprosthesis that takes into consideration the individual kinematics of the joint. Therefore, the implant should restore as precisely as possible the former morphological appearance of the bone in order to make sure that the sequence of movements in the joint is identical to that which existed before bone destruction. Background: With the aid of a new software package, an image analyzing process has been developed which allows subtraction of the CT-volume data of the affected region with the aim of receiving the 3-dimensional form of the required implant. Method: In an experiment, we produced by means of a spiral CT a 3-dimensional data-set of a knee-joint specimen. Then, corresponding to a medial and femoropatellar arthrosis, bone was removed by a reamer, taking into consideration the individual consistency of the bone and the importance of cortical support for the future implant. After this, a second CT was performed. The second image series was subtracted from the first one, and the result verified by 3-dimensional-imaging. Small defects in the surface were eliminated by calculating reference points; bigger chondral or bone lesions were remodeled by using the reciprocal data for the contralateral knee joint. The modified data-set was fed into a milling-machine, and a three-piece (polyethylene, cobalt chrome) prosthesis produced. Results: The implant showed very good press-fit characteristics in relation to the surface of the individual knee joint. Under the condition of functional testing there was no need for further fixation. The movement of the implant-bearing knee corresponded absolutely to the physiological movements. This method produces a custom-made knee-prosthesis for an individual and minimizes bone resection, optimizes the congruence of the contact areas of bone and implant, and reduces alignment difficulties in surgery. Such a knee-prosthesis reproduces precisely the physiological and individual movement of the knee, so a longer survival is to be expected. Abstracts from CAOS ’99 227 PLANNING AND GUIDANCE OF TIBIAL OSTEOTOMIES: EARLY CLINICAL RESULTS A FLUOROSCOPY-BASED SYSTEM FOR COMPUTER ASSISTED HIGH TIBIAL OSTEOTOMY R.E. Ellis1,2, J.F. Rudan2, M.M. Harrison2. 1Computing and Information Science, Queen’s University, Kingston, Canada, 2Surgery, Queen’s University, Kingston, Canada. P.F. La Palombara, S. Martelli, C. Casadei, M. Marcacci. Biomechanics Laboratory, Istituti Ortopedici Rizzoli, Bologna, Italy. Objectives: To design and evaluate a CT-based surgical system for planning and guiding high tibial osteotomy. Background: The literature reports high variation between the radiographically-determined tibial shaft/tibial plateau angle in corrective high tibial closing-wedge osteotomy and the preoperatively planned TPTS angle. It was hypothesized that a CT-based system for planning the correction in three dimensions, coupled with an intraoperative guidance system, could reduce such error. Design/Methods: A three-dimensional model of each patient’s tibia was extracted from the preoperative CT scan by isosurface calculation. The surgeon preoperatively determined the tibial shaft and tibial plateau, the location of the proximal resection plane, the orientation of the wedge to be removed, and the total correction angle. The planning system performed a “virtual surgery”, allowing the surgeon to inspect a prediction of the reduced osteotomy to verify that the medial bone cortex and proximal fragment were appropriate for hinge formation and step-staple support, respectively. Registration was performed either by fiducial means or by robust surface registration. Intraoperative guidance included visualization of the CT slices, superposition of the planned resection planes and a virtual surgical instrument, and numerical guidance of the surgical instrument to the resection planes. An in vitro study compared traditional technique with the computer technique, and an in vivo pilot study on patients was initiated. Results: The in vitro study demonstrated that the maximum correction error and the standard deviation of error were reduced by half (p,0.05). The in vivo pilot study suggested that similar accuracy is achievable in a clinical setting (maximum TPTS error 1.5, n55). After initial acclimation to the system, registration and drilling of the Kirschner wires used in the modified Coventry technique were performed in 6 min, with no intraoperative fluoroscopy required. Conclusions: The participating surgeons found that the planning system gave a greater appreciation of the individual patient’s anatomy, and was valuable in determining the three-dimensional effect of the intended osteotomy. Intraoperative guidance was easy to follow. The system demonstrated that preoperative planning, including a virtual surgery to predict the outcome, could be safely and effectively used in a clinical setting. Objectives: The aim of the proposed approach to computer assisted HTO is to ensure a higher degree of consistency between pre-operative planning and intra-operative execution, increase the global accuracy standard, and reduce the patient exposure to Xrays, while limiting the complexity of the man-machine interface and the technology-induced cost overhead. Background: HTO is a common surgical option for patients suffering from severe knee arthritis. Correction of the joint alignment is attained by ablating a wedge of bone from the proximal third of the tibia. Although osteotomy rates are steadily decreasing, HTO is still commonly felt to be the most appropriate surgical treatment for younger patients. A major drawback of the traditional surgical approaches is the difficulty in obtaining the planned correction. Inaccuracy in joint re-alignment can seriously compromise the long-term result of the intervention. Design/Methods: Our system includes a plexiglas cage equipped with metallic markers used for calibration of intra-operative fluoroscopic images. Two pairs of orthogonally oriented scans are acquired intra-surgically. A simple intra-operative planning phase allows the surgeon to define the correct pose of the guides used for bone resection on the calibrated images. Data are then transferred to a robot, the reference system of which has been previously registered to the image-based coordinate system. The robot is used as a pointer, which indicates to the surgeon how to accurately insert the guides. As the patient’s leg is held rigidly coupled to the bed/robot during the intervention, no additional fluoroscopic scans are needed. Results/Conclusions: The system is currently at an early stage of development. The work done so far includes the definition of the system and user requirements, the construction of a cage prototype, the implementation of an image calibration protocol, preliminary trials on calibration accuracy, and the creation of a provisional planning interface. Although further extensive investigation and testing are mandatory before expressing a conclusive assessment of the system effectiveness, our feeling is that the technique is promising and adaptable to various computer assisted surgical applications. 228 Abstracts from CAOS ’99 REGISTRATION OF THE TIBIA IN ROBOT-ASSISTED TOTAL KNEE ARTHROPLASTY BY MEANS OF A FORCE CONTROLLED ROBOT K. Denis1, G. Van Ham1, J. Vander Sloten1, R. Van Audekercke1, G. Van der Perre1, J. De Schutter2, J. Bellemans3. 1Division of Biomechanics and Engineering Design, K.U.Leuven, Heverlee, Belgium, 2Division of Production Techniques, Machine Design and Automation, K.U.Leuven, Heverlee, Belgium, 3Division of Orthopaedics, K.U.Leuven, Heverlee, Belgium. Objectives: Developing a strategy for robot-assistance in Total Knee Arthroplasty (TKA) to improve the quality of the bone cuts. Background: Cementless TKA requires a high surface flatness of the cuts to maximize the chances for bone ingrowth and accurate machining of the bone surfaces to obtain proper joint kinematics. Design/Methods: To register the tibia, an intramedullary rod is inserted and its orientation is measured by the robot. After registration, the surgeon moves the robot-arm under force control and the robot constrains the motion to the predefined cutting plane. Monitoring the milling forces provides the surgeon with additional information about the local bone quality. Results: The achieved surface flatness is 0.1-0.2 mm, which is significantly better than in conventional surgery. The accuracy of the registration is 0.9 degrees in the frontal plane and 1.2 degrees in the sagittal plane. An exponential relationship between milling force and local bone density is established. The preparation of the tibia is demonstrated on a cadaveric leg. A similar procedure is now being applied for the femur. Conclusions: This work shows that co-operation between the surgeon and a robot is able to improve the quality of the bone cuts in TKA. A SPECIAL-PURPOSE ROBOTIC SYSTEM FOR KNEE SURGERY B. Davies1, S. Harris1, M. Jakopec1, A. Fan1, J. Cobb2. 1Mechatronics in Medicine Group, Imperial College, London, United Kingdom, 2Orthopaedic Department, Middlesex Hospital, London, United Kingdom. Objectives: To develop a special-purpose robotic system (called ACROBOT) specifically for knee replacement surgery, to utilize the Imperial College concept of “Active Constraint” control. The system includes a 3-D model using CT scans, a simple pre-operative planner, the robot system, and an interoperative display and controller for the Surgeon. Background: As part of the UK Link initiative, over the last 2 years the Department of Health has funded Imperial College to research into systems for robotic knee surgery, in conjunction with Mr. Cobb of Middlesex Hospital, Howmedica plc, and Armstrong Healthcare Ltd. Methods: The robotic system comprises a Passive Gross-Positioning Robot, which can be locked off at the appropriate location and orientation when in the region where cuts are to be taken. The Passive robot in turn carries a small, precision robot that can be used to cut the bones very accurately, in order to precisely align and fit prosthetic knee components. The robot can be driven directly by the surgeon in an “Active Constraint” mode, which allows the Surgeon to hold a force-controlled handle and move the robot within regions defined by the Computer Program. In this way the robot can be used for safety, to prevent cutting prohibited regions, and also to provide accuracy in cutting 3D complex shapes, whilst the surgeon uses his/her inherent judgement and sensing capability to directly control the robot. The robot also has an “automatic” mode. A simple PC-based planning system is used to position models of the appropriate prosthesis on reconstructed CT scans of the knee, and to plan the procedure. These plans are then used to automatically generate the robot regions of constraint and for the cutting routines. Fiducial markers are used in the CT scans to allow registration of the robot to the patient and to the pre-operative plans. The patient’s leg is clamped in a fixture and the limb treated as a fixed object with safety-monitoring for unwanted motion. Results: The low cost modeling and planning system is easy to use and provides the surgeon with a valuable aid to choosing the most appropriate size and position of the prosthesis. The plans are automatically passed to the robot for implementation. Specially adapted fiducial screws are used to register the robot to the patient. The sequence of cuts is implemented to an overall accuracy of around 1.5 mm. Phantom and cadaveric trials are in progress.