ROBODOC is a product having been used for thousands of procedures concerning hip and knee surgery. 1994 opened the way to robot- assisted laparoscopic ...
Surgical robots at TIMC: where we are and where we go. Jocelyne Troccaz1, Philippe Cinquin1, Peter Berkelman1, Adriana VilchisGonzales1, and Eric Boidard1 1
TIMC Laboratory, Institut de l’Ingénierie de l’Information de Santé, Faculté de Médecine, Domaine de la Merci, 38706 La Tronche cedex, France
Abstract. Grenoble was one of the first places worldwide where a robot was used in clinical routine. In 1989, the first robot-assisted procedure on a patient took place in the neurosurgery department of the Grenoble University Hospital. An industrial robot modified to satisfy to the clinical constraints was used as a positioning device for guiding minimally invasively the surgical tool to a planned target. Based on that experience and on our knowledge of the clinical domain, we progressively re-directed our activity towards the design of specific robotic devices answering important issues among which safety, interactivity and clinical suitability. In this paper, we describe in more details our view of what surgical robotics is and should be and we illustrate our approach by the description and discussion of research in progress at TIMC: the PADyC arm, a passive device constraining the surgeon’s motions in function of a pre-planned surgical protocol - the TER system a non rigid and portable robot for tele-echography - and LER a portable endoscope holder. We will discuss the specificities of medical robotics. Finally, we will draw the perspectives that we foresee for this domain.
1
Introduction
Surgical robotics mainly consists in using a robot in an operating room (OR) as a more or less automated instrument holder. The objectives may be: to make a procedure geometrically more precise [17], to recover dexterity through articulated instruments in key-hole surgery [9], to move heavy instruments [20] or sensors [3], to allow remote interventions [15] or intra-body procedures [5], etc. Those image-guided systems evolved from the use of industrial robots to specifically developed devices. In this introduction, we report a short history of surgical robotics with major milestones; those milestones refer only to systems used with real patients. Then, we describe the different classes of surgical robots. 1.1 Milestones A robot was used for the first time to assist a surgical intervention on a patient in 1985 by Kwoh and colleagues [12] in the Memorial Hospital of Los Angeles. A Unimation PUMA 200 held a laser for neurosurgical interventions. The clinical
experiments stopped after a series of 22 patients because the robot was considered as not suited to a clinical application. From 1989, a modified industrial robot has been used for stereotactic neurosurgery guidance in Grenoble (TIMC and University Hospital [13]); after the robot re-design, this system has turned into an industrial product named Neuromate1. In Grenoble, the system is still in use after thousands of interventions. In 1991, the first Trans Urethral Resection of the Prostate (TURP) was performed on a patient with a PUMA 560 robot which motion was constrained by a mechanical frame [6]. This was the first robotassisted procedure for soft tissues. Further versions of this system including a motorized frame were designed and commercialized in the nineties/twenties. 1992 made surgical robotics popular to a larger audience when ROBODOC1 [17] initially developed by the IBM Yorktown Heights Research Center and based on a IBM scara robot was used for a series of 10 patients for bone machining in total hip arthroplasty. ROBODOC is a product having been used for thousands of procedures concerning hip and knee surgery. 1994 opened the way to robotassisted laparoscopic procedures. AESOP2, a scara robot with 4 active and 2 passive dofs, controlled by the surgeon was used to move the endoscope [18]. This system was the first surgical robot which obtained the FDA agreement for market introduction. In 1998, the DaVinci3 system [9] allowed to perform in Paris the first robot-assisted fully endoscopic coronary artery bypass grafting in cardiac surgery [14]. DaVinci is a master-slave system which slave robot integrates intra-body degrees of freedom (DOFs). In 2001, the ZEUS2 robot was used for a transcontinental cholecystectomy procedure [15]. The surgeon, located in NewYork (USA), tele-operated the robot for an intervention on a patient located in Strasbourg (France). Those milestones do not represent the exhaustive history of surgical robotics (see http://www.surgetics.org for a more complete description); as can be seen they concern a majority of commercial products. This will be discussed in section 3. However, they illustrate the evolution and demand of clinical applications. Neurosurgery was obviously the first domain where key-hole surgery was developed and high precision was required. In the early nineties, orthopedics became one of the fields where procedure optimization became very important both because the average life duration increased and because elderly people got physically active much longer. Because orthopedics is close in some sense to industrial machining of parts, robotics had a good opportunity to get developed in this domain. The success is not the expected one. This will also be discussed in section 3. Finally, laparoscopic procedures raised many problems related to the reduction of vision and dexterity of the surgeon whilst at the same time the number of such procedures increased drastically because of their advantages in terms of reduction of pain and recovery time. Thus, this has been the third domain where robots were significantly pushed forward. 1
Neuromate and ROBODOC are trade marks from Integrated Surgical Systems Inc. AESOP and ZEUS are trade marks from Computer Motion Inc. 3 DaVinci is a trade mark from Intuitive Surgical Devices Inc. 2
1.2
System classification
Several classifications were proposed to describe the existing systems. The one we use is based on the type of interaction the surgeon has with the system. Passive systems display information to the surgeon so that he can compare his gesture with a pre-planned strategy and correct it if needed. Non robotic so called navigation systems provide this functionality (see [16] for instance). Active systems autonomously realize a sub-procedure under the supervision of the surgeon. An example is ROBODOC. Semi-active systems necessitate the combined action of the surgeon. For instance a mechanical guide positioned by the robot is used by the surgeon for a linear action (e.g. Neuromate). Those systems basically constrain the surgeon action. More recently, this class has been enlarged to include programmable devices which provide the same type of interaction. We renamed them synergistic systems. Finally, tele-operated master-slave systems allow the execution of orders transmitted by the surgeon to a slave robot. Scaling and filtering of movements are possible. The distance between the master and the slave is quite variable. Most often a few meters separate the master from the slave.
2 Work in progress in TIMC 2.1 Development philosophy The first version of the first robotic system developed in TIMC – a semi-active system for neurosurgical procedures [13] – was based on the modification of an existing industrial robot (from the AID French Company4). Maximal joint velocities were reduced. Reduction ratios were increased to make the robot irreversible. When the robot was in position, its power was turned off to forbid any motion during the surgical gesture. Special attention was paid to safety for trajectory generation and collision avoidance. Indeed, the application was rather critical since the robot end-effector had to come very close to the head of the patient and because the instrument had to be positioned inside the brain, and potentially very close to anatomical structures at risk which damage could have dramatic consequences for the patient. This application made us familiar with all the technical problems related to such an application (calibration issues, registration to the images, OR constraints, etc.). It also made us conscious of the very specific nature of surgical robotics as compared to industrial automation robotics. This led us to envision the development of a robot enabling interaction with the surgeon whilst being more generic than a semi-active robot for the very numerous procedures where automation is not wished. This resulted in the design 4
This company does no longer exist.
and development of the PADyC5. In a second time, we also decided to develop light robots which integration in the OR could be made easier. TER (a robot for tele-echography) and LER (a robotic endoscope holder) were designed and realized. 2.2 PADyC Principles of synergy The basic principle of synergistic devices is the following. Both the surgeon and the synergistic device hold the tool, apply forces to it and to each other, and impart motions. Under computer control, the synergistic device may allow the surgeon to have control of some degrees of freedom while the device controls the others. The system filters the motions proposed by the surgeon to keep only those which are compatible with the surgical plan. For instance, during the pre-planning stage, the orthopaedic surgeon selects the cutting plane for machining the patient bone before placing the knee prosthesis element. In such a case, the synergistic system guarantees that the motions of the cutting tool are strictly limited to the preplanned plane while the surgeon is in charge of the selection of the motions within the plane. Such a system takes the best advantage of (1) the robot with its precision and its computer-based model of the surgical action, and (2) the surgeon and his knowledge, know-how, sensing capabilities and ability to react to unexpected or non-modelled events. Synergistic systems have been implemented using several types of technologies. Acrobot (Active Constraint Robot), an implicit force controlled robot, is used for knee surgery [10]. The motors and the operator are considered at the same level in the control loop. Whereas this system is based on an active mechanism, Cobot [4] and PADyC are based on passive joints. Cobot (Cooperative robot) is based on non-holonomic elements allowing the coupling of DOFs. For instance, for a planar 3 DOF (2 translations and 1 rotation) Cobot, x,y and α parameters are defined such that tgα=vy/vx where vx and vy are the Cartesian velocities in the plane of motion. This Cobot can naturally constrain the motion to a pre-defined trajectory by controlling α. PADyC general design PADYC principles were introduced in [22]. The actuation of PADyC comes exclusively from a human operator. This choice of a passive device although it may have some drawbacks was aimed at providing intrinsic safety. In order to filter the tool motions proposed by the surgeon moving the end-effector, each joint of PADyC is equipped with two freewheels and a clutching mechanism. A freewheel provides the basic function of unidirectional motion. For instance, as 5
Passive Arm with Dynamic Constraints
shown in figure 1, let us consider that the internal part of the freewheel is fixed (ωi+ = 0). If the operator imparts motion to the external part of the freewheel with the velocity ωuser in the positive direction, the motion will be blocked while the motion is free in the negative direction. If a motor is now associated to the internal part of the freewheel and rotates with the velocity ωi+, then both directions of motion are allowed but ωuser is bounded by ωi+ in the positive direction. For one joint, each of the two freewheels is associated with one clutching motor rotating in a single direction. The velocities ωi- and ωi+ of the two motors associated to the joint Ji are computer-controlled. The operator moves the other part of the freewheel at velocity ωuser. Depending on the values of the relative velocities (ωi+,- ωuser), the motion is blocked or authorized6. Based on this mechanism, one can guarantee that at each instant the resulting velocity ωi of the joint Ji is bounded by +,ω i− ≤ ω i ≤ ω i+ . By a suitable choice of the ωi values, called the dynamic constraints, the computer is able to control the instantaneous joint velocities (direction and intensity) and to guarantee that a given task is correctly executed. Such a robot may be seen as a prevention system keeping only and facilitating displacements of the tool which are authorized by the surgical plan.
+
ωi+ ωuser Fig. 1. PADYC mechanical principle.
Task description A surgical task is described as a sequence of elementary task constraints expressed using a set of programming modes. Free mode: a mode where the motion is free. No constrain applies. The robot behaves as a tracking device. Position mode: a mode where a target position has to be reached. At each instant, the only authorized motions are those which reduce the distance to the
6
See the interactive demo on http://www.surgetics.org
goal. Such a mode is useful to go to an approach position or to position an object (sensor, instrument, prosthesis element). Trajectory mode: a mode where a trajectory has to be followed with a given accuracy (for instance for a puncture). Positive (resp. negative) region mode: a mode where the instrument has to keep inside (resp. outside) a pre-defined region. The negative region mode may be used to guarantee that organs at risks are not damaged. Specific modes (conical, linear, planar, etc.). Those modes are optimized trajectory or region modes. One difficulty is to translate those task constraints (generally expressed in the Cartesian space in the patient reference frame) into dynamic constraints (expressed in the joint space of the robot). This problem is similar to the very well known problem of C-space obstacle determination for robot path planning. Details of the algorithmic approach can be found in [23]. Status of the project One-DOF, two-DOF and three-DOF laboratory prototypes allowed the mechanical design, the development of the algorithmic tools and the validation of the interaction approach [7]. Based on these successful results, it was decided to build a 6 DOFs PADyC for a real clinical application (see figure 2). Computer-aided puncture of the heart was selected; the system was designed, programmed and experimentally evaluated.
Fig. 2. PADYC 6DOF prototype (left: real prototype – right: simulation of cardiac application).
Although good results were obtained – algorithms were generalized to 6D, positions were accurately reached and system transparency was rather good – we could not envision the application of this robot in the OR due to an unacceptable elasticity of the structure [19] that could result in potential dangerous deviations of the tool under unexpected external force application. The large size and weight of the robot was one reason for that unwanted elasticity. Therefore, we decided to investigate the development of a mini-PADyC with cable transmission. A
prototype has been designed and work is in progress. The mini-PADyC is intended to be positioned directly on the patient body. 2.2 Portable robots Most surgical robots are rigid structures placed on the floor or attached to the bed or to the ceiling of the OR. Very recently, research projects were launched to develop light robots positioned directly on the patient body. This has several advantages. One of them is related to the lack of space in the OR. Another one is that the workspace being much smaller the robot motion and related safety can be controlled in a better way. Finally, because the robot moves with the patient, the problem of target motion (either due to unexpected events or to natural conditions such as breathing) can be tackled much more easily7. [21] presents a miniature parallel robot attached to the spineous process of the vertebra for insertion of screws in the vertebral body of a patient. [5] describes a robotic endoscope for colonoscopy and plans to make this system cheap and disposable. Those systems are a very interesting illustration of this evolution of medical robotics. In a similar spirit, we developed the TER system for robotic tele-echography and LER an endoscope holder as described in the coming sections. TER Among many types of medical equipment, ultrasound diagnostic systems are widely used because of their convenience and safety. Performing an ultrasound examination involves good eye-hand coordination and the ability to integrate the acquired information over time and space; the physician has to be able to mentally build 3D information from both the 2D images and the gesture information and to put a diagnosis from this information. Some of these specialized skills may lack in some healthcare centres or for emergency situations. Tele-consultation is therefore an interesting alternative to conventional care. Development of a high performance remote diagnostic system, which enables an expert operator at the hospital to examine a patient at home, in an emergency vehicle or in a remote clinic, may have a very significant added value. Several robot-based echography projects have been launched worldwide (see [1] for instance). The tele-operated TER system [24] allows the expert physician to move by hand a virtual probe in a natural and unconstrained way and safely reproduces this motion on the distant robotic site where the patient is. The physician located in the master site moves the virtual probe placed on a haptic device to control the real echographic probe placed on the slave robot. Bi-directional connections (through LAN, ADSL, ISDN or very high rate networks) are used for data transfer between the master and slave sites. One originality of TER lies in its slave robot 7
Let us remind that for instance in ROBODOC, the bone of the patient is rigidly attached to a frame to guarantee that no motion may occur during the procedure.
architecture. It is a lightweight, uncoupled non-rigid robot placed on the patient body (see figure 3). The robot design naturally ensures a continuous contact between the echographic probe and the body of the patient which is necessary for correct echographic images acquisition.
Fig. 3. The TER slave robot. (left: close-up on the orientation structure – right: experimental evaluation).
The TER robot includes two independent structures – one parallel (2 DOFs), one serial (4 DOFs) – having two independent groups of actuators. Straps and cables connected to actuators are used to position and orient the echographic probe. This provides natural compliance at some expense of accuracy. However in this application, the required precision is not very high and the physician “closes the loop” from the information he gets from the echographic images; thus, robot intrinsic accuracy is no longer an issue. Experiments have been successfully performed on anatomical phantoms and on voluntary persons. The clinical validation of TER between Brest and Grenoble (1100km) for patients suffering from aortic abdominal aneurysms is in progress. LER In conventional minimally invasive surgery, surgeons operate with long, thin instruments through “keyhole” incisions approximately 10mm in diameter in the abdomen of the patient. An endoscope, a thin optical tube which is inserted through one of the incisions and connected to an external video camera, is used to visualize the internal organ structures and instrument tips. The endoscope video camera image is displayed on a monitor during surgery. Since the surgeon generally has both hands occupied with surgical instruments, an assistant is necessary simply to hold the endoscope steady in a desired position. For practical use in an OR environment, the endoscope manipulator must be unobtrusive, safe, simple to set up and use, and easily sterilizable. Several robot systems such as [18] have previously been developed for laparoscopic endoscope manipulation during surgery. Commercially available robotic surgical endoscope manipulators are included in the DaVinci, Aesop and Zeus. These manipulators are typically elements of large, heavy, complex and expensive systems and resemble conventional industrial robot manipulators. To make the OR integration easier, we
have taken a simple, lightweight, low cost approach instead [3]. The positioning mechanism is fixed to the endoscope and strapped to the patient at the incision location, so no rigid base is necessary and the manipulator moves with the patient during breathing, repositioning by the surgeon, motions of other instruments, or any other displacement of the abdomen wall (see figure 4). LER has three DOFs (pan, tilt and zoom) controllable via voice, joystick, tracking commands. The first versions of LER relied on cable actuation. They have been replaced in the current version by small sterilizable brushless D.C. motors “on board”. LER weight is 625g. Experiments on cadavers were very successful. We have recently applied to the medical ethical committee for authorization of LER clinical validation which is the next stage.
Fig. 4. Experimental evaluation of LER on a corpse.
3
Discussion and conclusion
3.1 From laboratory experiments to the OR Whilst surgical robotics is a more and more active field (cf. the IEEE TRA special issue on medical robotics, 2003:19(5)), few systems really entered the OR. As could be seen in section 1.1, many of the most famous systems are not the most innovative ones in terms of robotics but they were operational systems, often based on rather well-known technology, offering guarantees concerning reliability and safety and answering a clinical problem. Let us remind that surgical robotics involves several development stages; as in many domains, the robot has to be developed to fit the application requirements and evaluated. The evaluation ranges from laboratory technical experiments quantifying accuracy, robustness, etc. to clinical experiments through a
continuum: with phantoms8, isolated organs, cadavers, animals, healthy volunteers (when the application makes it possible) and finally with patients. Clinical evaluation means that the system developers have to prove in close collaboration with clinicians from several hospitals, if possible from several countries, that the developed system has some clinical added-value and does not result in some extra-complications, accidents, excessive blood-loss or operating time, etc. as compared to conventional reference techniques. This clinical evaluation with patient follow-up for several months up to several years (in orthopedics for instance) is very strictly regulated and requires application to ethical committees (CCPPRB9 in France). In [2] the authors report such a development from the very first research ideas of ROBODOC to provisional clinical conclusions and to still open questions about the system clinical added-value. As a direct consequence of this long and complex process, surgical robotics is a domain where industrial partnership is mandatory. Quite surprisingly, this may be a barrier to innovation. Moreover, some laboratories may have a difficult access to clinical partners which limits their ability to go through those different stages. The second reason for a rather limited use of robots in surgery is, from our point of view, related to the real difficulty of introducing a robot in a surgical OR for everyday clinical practice. This is due to the machine complexity and size, and to its generally limited user-friendliness. Price is also most likely a limiting factor. For many clinical applications, passive systems which are simpler and cheaper that robotic ones are perfect tools: this also contributed to keep robots out of the ORs. 3.2 Open issues In this paper, we have shortly presented new interaction paradigms proposed by synergistic systems. A lot has still to be done to prove their practical efficacy. This sub-domain certainly requires more coordinated research with the haptic domain; this also necessitates working more closely to researchers of man/machine interface design. We also described new technological developments aiming at making the robot a small “intelligent” instrument. This is a challenge for technologists and relationships with micro-technology development are certainly a key for success. The very open issues that we see in terms of innovative research in surgical robotics concerns robots having to operate autonomously on mobile and deformable soft tissues. Few such robots already exist. One can cite DERMAROB [8] a hybrid force-position controlled robot for automated sampling of the skin. In this case, the problem is to be able to regularly cut skin samples on a deformable, 8
A phantom is an artificial object reproducing as accurately as possible an organ or a part of the body. 9 Comité de Consultation pour la Protection des Personnes dans la Recherche Biomédicale.
un-modeled surface with variable mechanical properties; solving such a problem is a first step towards the general problem we mentioned below. Indeed, in general, robot-aided interventions include a planning stage most often performed before the intervention from pre-operative data, a registration stage allowing to transfer the planning to the intervention conditions and an execution phase. If the target is mobile – located for instance in the heart or the kidney – a dynamic acquisition and registration of medical data should take place in real time, all along the intervention, such that an autonomous robot could follow the motion of the target or synchronize its action to this motion. This is a very challenging issue. Closing the loop on external data such as physiological signals or medical images imposes to be able to process reliably and in real time the acquired data. Results have been published about visual servoing of an endoscope holder [11] for instrument tracking and distance control to the organs. [20] also describes how a robot for radiotherapy could follow the organ motion. But here also, the treated problems are a little simpler since the robot moves the endoscope (resp. the radiotherapy treatment machine) at some distance of the organs (resp. the body). In general, safety is a very critical issue. Robust real-time image processing is far from being solved because most of these images are particularly reluctant to automated processing. We think that those problems – real time data processing, automated robot control from these data, reliability and safety analysis – will be key issues of the coming decade.
References 1. Abolmaesumi, P., Salcudean, S.E., Wen-Hong Zhu, Sirouspour, M.R., DiMaio, S.P. (2002). Image-guided control of a robot for medical ultrasound. In IEEE Trans. On Rob. And Automation, 18(1):11-23. 2. Bargar, W.L., Bauer, A., Borner, M. (1998) Primary and Revision Total Hip Replacement Using the Robodoc System. In Clinical Orthopaedics, (354):82-91, September 1998 3. Berkelman, P., Cinquin, P., Boidard, E., Troccaz, J., Letoublon, C., Ayoubi, J.-M.(2003) Design, Control, and Testing of a Novel Compact Laparascopic Endoscope Manipulator. In Journal of Systems and Control Engineering, 217(14):329-341 4. Colgate, J.E., Peshkin, M., Moore, C.. (1996) Passive robots and haptic displays based on nonholonomic elements. In Proceedings of the IEEE International Conference on Robotics and Automation 5. Dario, P., Carrozza, M.C., Pietrabissa, A. (1999) Development and in vitro testing of a miniature robotic system for computer-assisted colonoscopy. In Computer Aided Surgery, 4(1):1-14 6. Davies, B.L., et al. (1995) A Robotic Procedure for Transurethral Resection of the Prostate. In Proceedings of the 2nd Int. Symposium on Medical Robotics and ComputerAssisted Surgery, pp 264-271, Wiley 7. Delnondedieu, Y. (1997) Un robot à sécurité passive en réponse aux problèmes de sécurité et d’ergonomie en robotique médicale. Ph.D. Thesis, Institut National Polytechnique de Grenoble, TIMC laboratory, Grenoble, in french.
8. Dombre, E., Duchemin, G., Poignet, P., Pierrot, F. (2003) DERMAROB: a Safe Robot for Reconstructive Surgery. In IEEE Trans. On Rob. And Aut. 19(5). 9. Guthart, G.S., Salisbury, J.K. (2000) The IntuitiveTM telesurgery system: overview and application. In Proceedings of IEEE Robotics and Automation Conference, San Francisco, Avril 2000 10. Jakopec, M., Harris, S.J., Rodriguez y Baena, F., Gomes, P., Cobb, J., Davies, B.L. (2001) The first clinical application of a “hands-on” robotic knee surgery system. In Computer Aided Surgery, 6(6):329-339 11. Krupa, A., Gangloff, J., Doignon, C., de Mathelin, M., Morel, G., Leroy, J., Soler, L., Marescaux, J. (2003) Autonomous Retrieval And 3-D Positioning of Surgical Instruments In Robotized Laparoscopic Surgery Using Visual Servoing, In IEEE Trans. On Rob. And Aut. 19(5). 12. Kwoh, Y.S., Hou, J., Jonckheere, E.A. et al (1988) A robot with improved absolute positioning accuracy for CT guided stereotactic neurosurgery. In IEEE Trans. Biomed. Eng. Vol35, pp153-151. 13. Lavallée, Troccaz, J., Gaborit, L., Cinquin, P., Benabid, A.L., Hoffmann , D. (1992) Image guided robot: a clinical application in stereotactic neurosurgery. In Proceedings of the IEEE Int. Conf. on Robotics and Automation, pp618-625. 14. Loulmet, D. (1999) First endoscopic coronary artery bypass grafting using computerassisted instruments. In J. Thoracic Cardiovasc. Surg. 118(1):4-10. 15. Marescaux, J., Leroy, J., Rubino, F., Smith, M., Vix, M., Simone, M., Mutter, D. (2002) Transcontinental robot-assisted remote telesurgery: feasibility and potential applications. Ann Surg. 2002 Apr;235(4):487-492. 16. Mosges, R. et al. (1989) Computer assisted surgery. an innovative surgical technique in clinical routine. In Proceedings of Computer Assisted Radiology, CAR 89, pp413-415, H.U. Lemke, editor, Springer-Verlag. 17. Paul, H., Bargar, W., Mittlestadt, B., Musit, B., Taylor, R.H., Kazanzides, P., Williamson B., Hanson, W. (1992) Development of a surgical robot for cementless total hip arthroplasty. In Clinical Orthopaedics and related research, (285):57-66, December 1992 18. Sackier, J.M., Wang, Y. (1995) Robotically assisted laparoscopic surgery: from concept to development. In Computer Integrated Surgery: Technology and Clinical Applications. Taylor, Lavallée, Burdea, Mosges, Eds., pp577-580, MIT Press 19. Schneider, O., Troccaz, J. A Six Degree of Freedom Passive Arm with Dynamic Constraints (PADyC) for Cardiac Surgery Application: Preliminary Experiments. (2001) In Computer-Aided Surgery, special issue on medical robotics, 6(6):340-351 20. Schweikard, A., Glosser, G., Bodduluri, M., Murphy, M.J., Adler, J.R. (2000) Robotic motion compensation for respiratory movement during radiosurgery. Comput Aided Surg. 2000;5(4):263-77. 21. Shoham, M., Burman, M., Zehavi, E., Joskowicz, L., Batkilin, E., Kunicher, Y. (2003) Bone-Mounted Miniature Robot for Surgical Procedures:Concept and Clinical Applications. In IEEE Trans. On Rob. And Automation, 19(5) 22. Troccaz, J., Lavallée, S., Hellion, E. (1993) PADYC : A Passive Arm with Dynamic Constraints. In Proc. of ICAR'93, (Int. Conf. on Advanced Robotics), pp361-366. 23. Troccaz, J., Delnondedieu, Y. (1998) Synergistic robots for surgery : an algorithmic view of the approach. In Proceedings of the Workshop on the Algorthmic Foundations of Robotics, WAFR'98, Pankaj, Agarwal, Kavraki, Mason Eds., A.K.Peters. 24. Vilchis, A., Troccaz, J., Cinquin, P., Masuda, K., Pellissier, F. (2003) A new robot architecture for tele-echography. In IEEE Trans. On Rob. And Automation, 19(5)
