1451

Computer and82. Computer and Robot-AssistedMedical Intervention

Jocelyne Troccaz

Medical robotics includes assistive devices used bythe physician in order to make his/her diagnosticor therapeutic practice easier and more efficient.This chapter focuses on such systems. It introducesthe general field of computer-assisted medicalinterventions, its aims, and its different compo-nents, and describes the place of robots in thiscontext. The evolution in terms of general de-sign and control paradigms in the development ofmedical robots are presented and issues specific tothat application domain are discussed. A view ofexisting systems, ongoing developments, and fu-ture trends is given. A case study is detailed. Othertypes of robotic help in the medical environmentexist (such as for assisting a handicapped person,for rehabilitation of a patient or for replacementof some damaged/suppressed limbs or organs) butare outside the scope of this chapter.

82.1 Clinical Context and Objectives............... 1451

82.2 Computer-Assisted Medical Intervention 145282.2.1 CAMI Major Components ................ 145282.2.2 Added Value of a Robot ................ 1453

82.3 Main Periodsof Medical Robot Development .............. 145482.3.1 The Era of Automation (1985–1995) 145482.3.2 The Era of Interactive Devices

(1990–2005) ................................ 145582.3.3 The Era of Small and Light

Dedicated Devices(2000 to Present Day) .................... 1455

82.4 Evolution of Control Schemes................. 1458

82.5 The Cyberknife System: A Case Study ...... 1459

82.6 Specific Issues in Medical Robotics ......... 1461

82.7 Systems Used in Clinical Practice ............ 1462

82.8 Conclusions and Emerging Trends .......... 1463

82.9 Medical Glossary ................................... 1463

References .................................................. 1464

82.1 Clinical Context and ObjectivesInformatics and technology have dramatically trans-formed clinical practice over the last decades. Thistechnically oriented evolution has run parallel to otherspecific evolutions of medicine:

• Diagnostic and therapy procedures tend to be lessand less invasive for the patient, aiming at reduc-ing pain, postoperative complications, hospital stay,and recovery time. Minimal invasiveness results insmaller targets reached through narrow access (nat-ural or not) with no direct sensing (vision, touch)and limited degrees of freedom.• More and more data are handled for each patient(e.g., images, signals) in order to prepare and mon-

itor the medical action, and these multimodal datahave to be shared by several participating actors.• As in many other domains, quality control getsincreasingly important and quantitative indicatorshave to be made available.• Traceability becomes mandatory, especially regard-ing the ever-increasing number of legal cases.Traceability is also of primary importance in costmanagement.

These evolutions make the medical action increasinglycomplex for the clinician, at both the technical andorganizational levels. Computer-assisted medical inter-vention may contribute a lot to those clinical objectives.

PartH

82

1452 Part H Automation in Medical and Healthcare Systems

They may provide quantitative and rational collabora-tive access to patient information, fusion of multimodal

data, and their exploitation for planning and executionof medical actions.

82.2 Computer-Assisted Medical Intervention

The development of this area from the early 1980s re-sults from converging evolutions in medicine, physics,materials, electronics, informatics, robotics, etc. Thisfield and related subfields are given several almostsynonymous names: computer-assisted medical in-tervention (the most general), augmented surgery,computer-assisted surgery, image-guided surgery, med-ical robotics, surgical navigation, etc. We will use theterm computer-assisted medical intervention (CAMI)herein for this domain.

We define CAMI as aiming to provide tools thatallow the clinician to use multimodal data in a ratio-nal and quantitative way in order to plan, simulate,and accurately and safely execute mini-invasive medicalinterventions. Medical interventions include both diag-nostic and therapeutic actions. Therapy may involvesurgery, radiotherapy, local injection of drugs, interven-tional radiology, etc. (Some of these medical terms areexplained in a glossary located at the end of this chap-ter.)

82.2.1 CAMI Major Components

CAMI [82.1] may be described as a perception–decision–action loop, as presented in Fig. 82.1. Theperception phase includes data acquisition and pro-cessing, the development of specific sensors, and their

Virtual patient

Planning

Simulation

Perception

Action

Real world Virtual world

DecisionGuidingsystems

Biomechanicalmodels

Atlases

Statistical models

Fig. 82.1 CAMI methodology

calibration. Data may be acquired pre-, intra- or postop-eratively. Images may provide anatomical informationor functional one; data provided by the sensors may beone-dimensional (1-D), two-dimensional (2-D), three-dimensional (3-D, sparse or dense) or four-dimensional(4-D, i.e., varying with time). Each imaging sensorbrings its specific type of information and multimodal-ity is in general necessary. The need to assist theintervention in a quantitative and accurate way requiresthe calibration of sensors enabling both the transforma-tion from image coordinates to spatial coordinates andthe correction of possible image distortions. Specificsensors such as position sensors (also called local-izers) and surface sensors have been integrated intoCAMI components. Localizers give access to positionand orientation of objects (instruments, other sensors,anatomical structures such as bones). Surface sensorsgive access to the external surface of an object (an or-gan, for instance). All those information are potentiallyuseful to plan and control the execution of a medicalaction.

The main objective of the decision stage is to buildan integrated numerical model of the patient and, insome cases, a model of the action. As mentioned pre-viously, many types of information may be useful: dataprovided by imaging sensors, for instance, but alsoa priori medical knowledge (for instance, statisticaldata about organ shapes or occurrence of pathologies,biomechanical models of the limbs, etc.). One very im-portant stage is data fusion, also called registration,which corresponds to the action of representing all theinformation in a single reference frame. Registration,and in particular medical image registration [82.2], hasbeen a very active domain for three decades. From thisintegrated model, the medical action can be planned;for instance, one may have to determine the type, size,position, and orientation of a knee prosthesis that pro-vides the best alignment of hip, knee, and ankle joints;or one may decide which number, shape, intensity, po-sition, and orientation of radiation beams would allowto radiate a tumor of a given shape, with a given dose,whilst sparing organs at risks. Planning may involvehighly interactive tools where the clinician navigatesin the data and specifies the selected strategy. It mayalso include optimization tools when the medical goal

PartH

82.2

Computer and Robot-Assisted Medical Intervention 82.2 Computer-Assisted Medical Intervention 1453

can be specified as an optimization problem (radiother-apy planning is an excellent example). In some cases,planning may be very difficult and simulators could pro-vide help for the computation of the clinical outcomeof a selected action; for instance, when a bone frag-ment has to be moved in the patient’s face, it may bevery useful for the surgeon to foresee the functionaland aesthetic consequences of this gesture on soft tis-sues [82.3]. Registration and planning are included inthe decision stage.

Action requires accurate execution of the plannedintervention. Often, a registration stage is necessaryto transfer the planned action to the interventionalconditions; for instance, the action is planned frompreoperative data, e.g., magnetic resonance imaging(MRI), and must be registered to the real patient in in-traoperative conditions. Two main types of assistanceexist: navigational aids and robots. In the first case, theaction is monitored using suitable sensors such as lo-calizers; information is rendered to the clinician aboutthe planned and executed actions. The clinician makesuse of this information to control his/her action, suchsystems are called passive. The first surgical navigatorshave been used for neurosurgery [82.4, 5]. The actioncan also be performed more of less autonomously bya robot. The first application of a robot in CAMI tookplace around 1985; the application field was also neu-rosurgery. The medical robot needs to be connected topatient data and models in order to be able to transferthe planned intervention to the robot coordinates; thisis the robot registration problem. The medical robot istherefore always an image-guided robot.

Simulators may also be developed for training pur-poses. The advantages are the abilities to providefrequent and rare medical cases, to gain realistic ex-perience with lower stress than with real patients, andto quantitatively evaluate the practitioner. This may fa-cilitate the acquisition of new medical skills, with newtechniques and/or tools. This can also be considereda very valuable component of CAMI.

Based on numerical data, tracking of objects (in-struments, sensors, and anatomical structures), andpositioning of tools with navigation or robotized aids,those CAMI procedures are fully traceable.

82.2.2 Added Value of a Robot

Automation is generally not a primary goal of medicalrobotics, where the interaction with a clinical opera-

tor has to be considered with a very special attention.Indeed, most often medical robots are not intended toreplace the operator but rather to assist him/her wherehis/her capabilities are limited. In general CAMI sys-tems are considered only as sophisticated tools in thehands of the clinician.

In medicine, like in many other application areas,the advantages of the robot are its precision, ability torepeat a task endlessly, potential connection to com-puterized data and sensors, capability to operate inhostile environments [biological or nuclear contamina-tions, war or catastrophe areas, space (orbital station) orundersea (submarine), etc.] where clinician presence orabilities may be limited and humans may need medicalcare.

Navigational aids have already demonstrated theirclinical added value in various specialties (neuro-surgery, orthopaedics in particular) and their integrationin the clinical environment is generally easier than fora robot. Safety issues are also more limited. Moreovernavigation systems are very often more cost effective.These are the reasons why it is very important to usea robot only for clinical applications where it can of-fer functionalities that the navigation system cannot;potential specific robot abilities are:

• To realize complex geometric tasks (for instance, tomachine a 3-D bone cavity)• To handle heavy tools (e.g., radiation apparatus) orsensors (e.g., an intraoperative surgical microscope)• To provide a third hand to the clinician• To be remotely controllable and to offer scaling ca-pabilities in terms of transmitted motions or forces• To filter undesired movements (such as physiologi-cal shaking in teleoperation)• To be force-controllable down to very small forcescales• To execute high-resolution high-accuracy motions(for microsurgery)• To track moving organs and synchronize to externalevents based on some signals• To be introduced into the patient for intrabody ac-tions.

Another aspect that will be further discussed is the ab-solute necessity to demonstrate the clinical added valueof the system, i. e., to prove that it brings a clear clin-ical benefit at some level (for the patient, hospital, orhealthcare system).

PartH

82.2

1454 Part H Automation in Medical and Healthcare Systems

82.3 Main Periods of Medical Robot Development

Early medical applications of robotics are characterizedby transferring to this domain accurate and automatedtool positioning capabilities of robots originally devel-oped in industrial applications.

82.3.1 The Era of Automation (1985–1995)

As mentioned previously the first surgical robots wereintroduced in neurosurgery. Neurosurgery already hada very long tradition of minimal invasiveness and use ofcomputerized 3-D imaging data; indeed, the first com-puted tomography (CT) scans in the early 1970s wereperformed for brain imaging. Stereotactic neurosurgeryis a particular type of minimally invasive procedurewhich consists of blindly introducing a linear toolinto the brain through a keyhole (3 mm diameter) forbiopsies, removal of cysts or haematomas, placementof stimulation (Parkinson disease) or measurement(epilepsy) electrodes, etc. Conventionally, stereotacticneurosurgery is performed with the help of a stereotac-tic frame whose functions are to immobilize the patientskull, register preoperative data to the intraoperativesituation (the frame is installed on the patient’s head be-fore the preoperative examination), and finally to offermechanical guidance for tool insertion. Except for thefirst function, a robot can advantageously replace theframe: it is easily connected to imaging data, is less in-vasive, and may offer a larger range of trajectories fortool positioning. Anthropomorphic robots associatedto stereotactic frames or robotized stereotactic frameshave been developed and clinically evaluated in theearly 1980s (for instance [82.6, 7]). Reference [82.7]describes the accurate positioning of a guiding toolwith respect to preoperative data in a stereotactic neu-

a) b)

Fig. 82.2 (a) Grenoble robot for neurosurgery (TIMC laboratory and Grenoble University Hospital) and (b) the Neuro-mate industrial version

rosurgery application using a Puma 260 robot. Thecited paper reports a series of 22 patients. While [82.8]presents a CAMI system integrating a modified indus-trial robot (reduced speed, nonbackdrivable joints) fora similar application. The first patient (Fig. 82.2a) wasoperated on with this system in 1989 and since thenhundreds of patients have been treated with this tech-nology. This system was the academic precursor of theNeuromate product (Fig. 82.2b) with which thousandsof patients were treated. These systems are called semi-active because the robot is only a guiding device andthe gesture (drilling the skull and inserting the needle)is still performed by the surgeon through a mechanicalguide positioned by the robot.

Reference [82.9] proposed an automated view ofthe whole process of stereotactic neurosurgery: a robotequipped with different types of tools in a tool-feedereffector was installed in a CT room; after imaging dataand planning, the robot performed the drilling of thebone and the placement of the surgical tool. CT enabledrepeated image control. The robot was specifically de-signed for this application. The first two patients wereoperated in 1993. Eight cases of biopsies are reportedin [82.9]. As far as we know, this system has not beenextensively used in clinical routine.

In orthopaedics, the placement of prostheses re-quires the preparation of the bones; for instance a cavityhas to be drilled before inserting the femoral compo-nent of a hip prosthesis; similarly, planar cuts have tobe realized on the tibia and femur extremities in orderto place knee prosthesis components. Thus, such stagesof the interventions are very close to machining a me-chanical part: a 3-D shape has to be accurately realizedby sawing or milling with given position and orienta-

PartH

82.3

Computer and Robot-Assisted Medical Intervention 82.3 Main Periods of Medical Robot Development 1455

tion. This is why the idea of using a robot came verynaturally for such tasks. Robodoc [82.10, 11], used forcavity preparation in total hip arthroplasty, was devel-oped from 1986 first in a laboratory setup. Then, from1989 to 1991, 29 dogs were operated with the helpof the system. The first ten patients were included ina Food and Drug Administration (FDA)-approved clin-ical research trial between 1991 and 1993; very largeseries of patients for comparison of traditional interven-tions with Robodoc-assisted ones were included from1995 to the early 2000s. Thousands of patients benefitedfrom the use of Robodoc. Robodoc is active in the sensethat a part of the surgical action (machining the bone) isperformed autonomously by the robot under surgeon su-pervision. Several other systems were developed basedon a similar approach.

The underlying idea in this first period was that sur-gical subtasks such as accurate positioning of a tool ormachining, based on numerical data, could be trans-ferred to a robot and automated to a certain extent.

82.3.2 The Era of Interactive Devices(1990–2005)

Whilst the automation era focused on rigid and non-deformable anatomical structures, the second era ischaracterized by the development of more interactivecontrol schemes for complex tasks, in particular for in-terventions on soft tissues.

Indeed, in the two previous examples (Sect. 82.3.1),the anatomical structure of interest is rigid and nonde-formable; in the case of stereotactic neurosurgery, thebrain is accessed through a small hole. The trajectorythrough the hole is simple (a linear tool is inserted toa given target). Brain motion and deformation can beneglected and the skull is immobilized. As regards or-thopaedics, bones are not deformable and can be fixedto external fixtures to avoid any motion. Obviously thereare much more clinical situations where the proceduresare more complex and concern soft, mobile, and de-formable tissues. For such applications automation isoften still out of reach or may not be the preferred so-lution, for instance, when the expertise of the clinicianis so complex that it cannot totally be transferred to therobot.

In contrast to earlier automated robot control,robotic development in the mid 1990s was charac-terized by more direct operator control. In particular,efforts were applied towards teleoperation; this formof robotic application was traditionally used in the nu-clear industry. In this situation, the surgeon is totally

in control of the surgical tool through a master–slave apparatus. The distance between the masterand slave components is highly variable: althoughthe main use is for very close teleoperation (in thesame room), some very long distance (thousands ofkilometers) experiments have taken place. The function-transferring movements from the master to the slavemay involve scaling down (forces and/or motions) andfiltering. The main clinical applications are in en-doscopic surgery where instruments and optics areintroduced into the patient’s body through small inci-sions. Those entry points limit the instruments’ possiblemotions [four degrees of freedom (DOFs) insteadof six] and the surgeon has to operate under videocontrol. In a first stage, the motion of the opticalsystem – the endoscope – alone was robotized: in-stead of requiring one assistant to move the endoscopefor hand–eye coordination, the robot is controlled bythe surgeon himself/herself by different means: voicecontrol (AESOP [82.12]), head movements (Endoas-sist [82.13]) or high-level image processing software(Sect. 82.4). More recently the displacement of in-struments has also transferred to robots, potentiallyoffering extra DOFs; this is the case of the DaVincimulti-arm system that offers intrabody DOFs of the in-struments [82.14].

From the mid 1990s synergistic devices [82.15] alsonamed hands-on robots [82.16] were proposed. The ra-tionale for such systems is that the clinical applicationis generally so complex that it can only be partially em-bedded in numerical models and data. Therefore relyingboth on the clinician for his/her very high skills, ca-pacity of judgement, intelligent perception, and on thecomputerized robot for its quantitative knowledge ofthe planning, accuracy, and sensors is potentially veryfruitful. In the PADyC [82.17, 18], Acrobot [82.16, 19],and Makoplasty [82.20] systems, the surgical tool is at-tached to the robot effector and the surgeon holds it. Themotions proposed by the human operator are filtered bythe robot in order to keep only the part of the motionswhich is compatible with the surgical plan; for instance,the tool has to keep in a plane or in a given region.Different technologies implement this principle of con-strained motions: clutchable freewheels, backdrivablemotors, controlled brakes, etc.

82.3.3 The Era of Small and Light DedicatedDevices (2000 to Present Day)

This era tends toward a miniaturization of robots upto the stage where they can be attached to a body

PartH

82.3

1456 Part H Automation in Medical and Healthcare Systems

a) b) c) d)

Fig. 82.3a–d On-body and on-bone robots at Grenoble: (a) TER slave robot for tele-ultrasound examination (TIMC,Grenoble University Hospital); (b) LER-Viky endoscope holder (TIMC, EndoControl-Medical Company, Grenoble Uni-versity Hospital and Paris La Pitié Salpétrière Hospital); (c) LPR MRI- and CT-compatible robot for image-guidedpunctures (TIMC, Grenoble University Hospital); (d) Praxiteles robot for total knee arthroplasty (PRAXIM, TIMC,Brest University Hospital)

part, inserted into the patient’s body or integrated intomechatronic surgical instruments.

In systems presented in Sects. 82.3.1 and 82.3.2,the robot mechanical architectures were inspired bytraditional industrial robotics: general-purpose anthro-pomorphic six-DOF (or more) arms are used. Systemsare very versatile and can adapt to a variety of tasks.However, this generality has a cost: these robots arealso very often quite cumbersome and are attached tothe floor, bed or ceiling. Their working space may bequite large and not easily manageable in an operatingroom (OR) that offers limited space and in which peo-ple (patient and staff) are very close. This is why newgenerations of robots have been specifically designedand developed for limited application areas.

This new generation of systems has produced robotsso small that they can be supported by the patient bodyor attached to one of his/her bones. Because of theirsize these robots are more easily integrated into theOR and are designed to be very well suited to a spe-cific task. These two factors offer potential advantagesover traditional systems by providing a safer operat-ing environment and cost effectiveness. The GrenobleTechniques for Biomedical Engineering and Complex-ity Management (TIMC) laboratory was a pioneer insmall-scale robotics, and Fig. 82.3 illustrates four suchpioneering works. TER [82.21] (Fig. 82.3a), a robotfor tele-ultrasonic examination, has been clinically val-idated for remote examination of abdominal aorticaneurysm and for emergency care of abdominal trau-mas; LER [82.22], a light endoscope holder, has beenvalidated on pigs and corpses, and VIKY, the corre-sponding industrial product (Fig. 82.3b) is EuropeanCommunity (EC)-marked and has been used on pa-tients; LPR [82.23] (Fig. 82.3c), a CT/MRI-compatible

robot for punctures, has been validated for CT on pigsand for MRI on volunteers (but without puncture);Praxiteles [82.24] (Fig. 82.3d), a robot for total kneearthroplasty, has been validated on corpses and recentlyentered multicentric clinical evaluation on patients.

Following the same philosophy, the Mars system(now called Mazor [82.25], a product) is presented indetail in Chap. 78. It is being clinically evaluated on pa-tients for spine surgery. This list of body-supported orbone-mounted devices is not exhaustive.

Other systems are sufficiently small to be com-pletely introduced into the body of the patient. Somegroups focus on locomotion issues; for instance, severalversions of an inchworm-type compact robot that movesinside the intestine have been proposed [82.26]. Theaim is to replace the long, quite rigid, and painful en-doscope traditionally used in colonoscopy; experimentson pigs have been carried out successfully. With a rathersimilar locomotion principle, a robot that moves on thesurface of the heart has been developed [82.27]; exper-iments on live animals have been carried out. A robotrolling on soft tissue organs in the abdomen has alsobeen described [82.28]; this robot carries a camera andexperiments on animal organs have been carried out.Finally, a smaller and simpler device actuated from out-side of the body using external magnetic fields has beenproposed [82.29]; such a robot would be injected intothe eye, for instance, for drug delivery to retina vessels.

Other research groups aim at giving extra DOFs totraditional instruments: the domain of active cathetersthat can adapt actively to vessel curvature has been in-vestigated for several years [82.30, 31]. Recently, someof these systems have reached the market (Table 82.1).Finally, the field of articulated tools for endoscopicsurgery is also very active; the aim is to develop surgical

PartH

82.3

Computer and Robot-Assisted Medical Intervention 82.3 Main Periods of Medical Robot Development 1457

Table 82.1 Industrial medical robots

System name Clinical specialty Type Last distributing Estimated number Status

company of installed systems

Neuromate Stereotactic Semiactive Schaerer Mayfield 15 < nb < 20 Unknown

neurosurgery (mechanical guide)

PathFinder Stereotactic Semiactive Prosurgics Ltd. Unknown Unknown

neurosurgery (mechanical guide)

Surgiscope Microsurgery Surgical ISIS 15 < nb < 20 No longer

microscope holder distributed

MKM Microsurgery Surgical Carl Zeiss Unknown No longer

microscope holder distributed

Robodoc Orthopaedics Automated ISS Inc. 70 < nb < 80 No longer

(knee, hip) machining of bones distributed

in the USA

and in

Europe

Caspar Orthopaedics Automated URS ortho 50 < nb < 60 No longer

(knee, hip) machining of bones distributed

Cyberknife Radiotherapy Positioning and Accuray Inc. 134* < nb Growing

motion of the

radiation device

DaVinci Endoscopic procedures Endoscope and Intuitive 875** < nb Growing

(cardiac, digestive, instrument holder Surgical Inc.

gynecologic, urology, etc.) – intrabody DOFs

Zeus Endoscopic procedures Endoscope and Intuitive Unknown No longer

instrument holder Surgical Inc. distributed

Aesop Endoscopic procedures Endoscope holder Intuitive 800 < nb < 1000 No longer

Surgical Inc. distributed

EndoAssist Endoscopic procedures Endoscope holder Prosurgics Ltd. Unknown Unknown

Naviot Endoscopic procedures Endoscope holder Hitachi Unknown Unknown

Lapman Endoscopic procedures Endoscope holder Medsys Unknown Unknown

Viky Endoscopic procedures On-body endoscope Endocontrol Probably < 10 Emerging

holder Medical

Acrobot Orthopaedics Hands-on robot Acrobot Ltd. Unknown Emerging

Sculptor (knee, spine, etc.)

PIGalileo Orthopaedics (knee) On-bone semiactive PLUS Orthopedics Unknown Unknown

CAS AG

Praxiteles Orthopaedics (knee) On-bone semiactive Praxim Probably < 10 Emerging

Makoplasty Orthopaedics (knee) Hands-on robot Mako Inc. Probably < 10 Emerging

Mazor Orthopaedics (spine, etc.) On-bone semiactive Mazor Surgical 10 < nb < 20 Emerging

Technologies

Estele Radiology Telerobotics Robosoft Probably < 10 Emerging

(ultrasounds)

PartH

82.3

1458 Part H Automation in Medical and Healthcare Systems

Table 82.1 (cont.)

System name Clinical specialty Type Last distributing Estimated number Status

company of installed systems

Sensei Interventional Teleoperated Hansen Medical Unknown Emerging

radiology (cardiology) robotic catheter for

heart mapping (intra-

cardiac) with force

feedback

CorPath Interventional Teleoperated Corindus On evaluation – Emerging

radiology (cardiology) catheter not yet distributed

(intracoronary)

* Number of installed systems in early May 2008 (source Accuray Inc.)

** Number of installed systems in early May 2008 (source Intuitive Surgical Inc.)

tools equipped with intrabody DOFs to recover full mo-bility of the tools with respect to the organs. The specificarea of intrabody devices raises very challenging issuesrelated to biocompatibility, safety, power supply, anddata transmission. Smart pills such as the M2A [82.32]

or Norika [82.33] ones, which are swallowed by a pa-tient and enable visualization of the gastrointestinaltrack, do not yet integrate active devices but can be seenas precursors of future highly integrated mechatronicintrabody devices.

82.4 Evolution of Control Schemes

Evolution from bone applications to soft tissues alsoresulted in the evolution of control schemes providinga more active role for the surgeon or aiming at a real-time perception–decision–action control loop.

A large number of the oldest systems integratea single-shot perception–decision–action process; forinstance a planned trajectory is selected from CT dataand transferred to the intraoperative conditions afterregistration of preoperative to intraoperative data. Toguarantee that the plan is still valid intraoperatively, theanatomical structure of interest must not move. Thisapproach has been largely used for neurosurgery and or-thopaedics surgery; in both cases the structure is fixedusing a stereotactic frame or external fixtures. The op-erator is often limited to the role of a supervisor whenthe robot moves the tool.

More recently (Sect. 82.3.2) the operator has beengiven a more active role in the execution of the task.In the case of comanipulation the robot and the oper-ator participate simultaneously to the motions of thetool. In the Acrobot system [82.19] the proportional–integral–derivation (PID) coefficients of the control laware given by functions that depend on the position of thetool with respect to a region of allowed motions, whichcorresponds for instance to the bone to be removed for

prosthesis implantation: inside the region, the motionsproposed by the operator and detected by a force sensorare transmitted to the tool without significant modifica-tion; in an intermediate region the motions are more orless transmitted depending on their direction; motionsoutside the planned region are strictly forbidden. TheSteady Hand [82.34] works similarly: the operator ma-nipulates a handle and proposed motions are detected.In contrast to the Acrobot, the Steady Hand does notselect permitted directions of motions but rather scalesdown motions and forces for microsurgery or biologyapplications. In the case of PADyC the principle isslightly different [82.17, 18]; PADyC is a passive armfor which each joint is equipped with two freewheelsthan can be independently clutched or unclutched usinga motor; the motor is velocity controlled, and this veloc-ity determines the range of allowed motion of each jointin each direction at each instant. This range of motionis computed from the representation of the task (po-sition to reach, trajectory to follow or region to keepinside for instance) and from the current position ofthe robot. No force sensor is necessary. For teleoper-ation, the operator interacts with a master device andthe slave robot reproduces the motions proposed by theoperator. Transfer functions may enable scaling and fil-

PartH

82.4

Computer and Robot-Assisted Medical Intervention 82.5 The Cyberknife System: A Case Study 1459

tering of disturbing shaking motions like in the DaVinci.Teleoperation can integrate a force feedback to the oper-ator like in the Sensei system for endocardial robotizedcatheter control.

Few systems integrate hybrid force–position controlof the robot. One application case is external ultrasoundexamination, which requires constant contact of the ul-trasonic probe on the body of the patient. Thus severalrobotic systems for ultrasound examination (remote orautomated) have been developed with a hybrid controlscheme [82.35, 36]. They generally combine positioncontrol in the main direction of motion with force con-trol in order to maintain contact during probe motion.Dermarob [82.37], developed for skin sampling for skingrafts in burnt patients, follows a similar hybrid controlscheme.

Some of the systems developed for orthopaedics andneurosurgery offer some basic tracking abilities usinga localizer and markers attached to rigid structures of in-terest. This is, for instance, the case of Caspar, which isused for ligamentoplasty in knee surgery, and the case ofa frameless version of the Neuromate. Systems such asMazor or Praxiteles which are mounted on the anatom-ical structure of interest move with it and thereforesuppress the tracking problem. Recently other systems

have been developed with evolved abilities for trackingof soft tissues. Instead of having a fixed target, the robothas to track it in real time, in general from informationextracted from imaging data. Typical examples are re-lated to organ motions induced by cardiac or respiratorycycles. In those cases, a real-time perception–decision–action loop must be developed. The Cyberknife system(Sect. 82.5) developed for radiotherapy [82.38, 39] de-termines the current position of a tumor moving withpatient respiration by using external markers localizedin real time [82.40, 41]. Alternative approaches con-sist of coupling motion in the image information tomotion of the robot at a low level through visual ser-voing. Pioneer work was described in [82.42]; fromhigh-speed camera data, the motion of relevant fidu-cials on the heart surface was computed and fed back toa slave robot for teleoperation, giving the operator thefeeling of a stable target. More recently several groupshave applied visual servoing to image-guided minimallyinvasive actions. Video images are used for robotizedendoscopy [82.43,44]; ultrasonic images have also beenintroduced [82.45, 46] for cardiac and vascular applica-tions. Directly coupling motion detection in the imagesto motion of the joints requires very careful robustnessanalysis.

82.5 The Cyberknife System: A Case Study

This system is described in more detail to make the in-trinsic complexity of a robotic CAMI system a bit morevisible. This section is also intended to introduce the po-tentially long road from an initial paradigm and a firstprototype to a clinically used product. Specific issues re-lating to medical robotics and explaining this long roadare discussed in Sect. 82.6.

The Cyberknife system, distributed by Accuray Inc.,has been developed in the context of radiotherapy. Usu-ally, when treating a patient with conventional linearaccelerators, the patient is positioned onto a couch thathas four main DOFs (three translations and one rota-tion), and the linear accelerator allows orientation of theradiation beam using two more rotational DOFs. Theconvergence point of the radiation beams is termed theisocenter. A reference frame Riso centered on this pointis associated with the complete radiation system. TheDOFs enable positioning of the tumor on the isocen-ter and orientation of the radiation beams properly withrespect to the tumor in order to execute the plannedtreatment (Sect. 82.2.1). In practice, due to the complex-

ity of positioning the patient and orienting the beamswith such machines, treatments generally consist ofrather simple ballistics with rather few radiation beams.The Cyberknife concept proposes to install the radiationsource on a six-DOF robot in order to avoid repeatedcombined motion of the couch and linear accelerator.This enables the execution of complex treatments withhundreds of beams distributed around the tumor withouthaving to move the patient. Using such a large numberof small beams enables sculpting of the dose distribu-tion to the tumor shape with high accuracy.

The first version of the Cyberknife [82.38] was in-stalled at Stanford Medical Center in February 1994,and the first patient was treated with the system onJune 8, 1994. The current version of the system inte-grates a Kuka robot (Fig. 82.4a). In early May 2008,the Accuray Company reported 134 installed systemsworldwide and 40 000 patients treated, mostly for brain,spine, lung, prostate, liver or pancreas tumors.

Regarding patient data, preoperative CT is tradition-ally used for treatment planning; other modalities can

PartH

82.5

1460 Part H Automation in Medical and Healthcare Systems

Calibration

Intra-treatmentPre-operative

Calibration

Initial registrationon markers

b)

a)

Internal markers External markers

Tracking

Model

Rx-ray(1)Rloc

Riso

Rrobot

RCT

Rx-ray(2)

Calibration

Diagnosticx-ray source

Treatmentcouch

Robot manualcontrol pendant

Imagedetectors

Linear accelerator

Robotic arm

Fig. 82.4a,b The Cyberknife system: (a) the treatment room setup,(b) functional description of tracking functionalities

be fused to CT data such as MRI or positron emissiontomography (PET) for planning refinement. Intraoper-ative x-ray imaging produced by two x-ray systemsenables initial patient positioning and participates intracking. The system is calibrated, which means thatthe spatial relationships between the different referenceframes Riso, Rrobot, and Rx-ray (associated, respectively,with the isocenter, robot, and x-ray devices) is deter-mined using specific procedures and calibration objects.Such calibration is necessary to transform intratreat-ment imaging information into robot positions withrespect to the radiation system. The transfer of planninginformation from RCT to Riso is possible due to imageregistration procedures described in more detail in thefollowing sections.

In the first version of the system, dedicated toneuro applications, after initial setup, patient motionwas detected and corrected before each dose deliv-

ery from a single beam. Thanks to the CT exam, twosynthetic x-ray images called digitally reconstructed ra-diographs (DRRs) are computed. They correspond towhat would be seen by the two x-rays machines inthe treatment room for a perfect position/orientationof the patient with respect to planning data. Tworeal images are acquired just before dose deliveryand automatically compared with the DRRs; this re-sults in the computation of the patient’s shift from theideal position/orientation. For small errors, the robotposition/orientation with respect to the patient is auto-matically corrected by combining this information withcalibration data. For larger discrepancies, a replanningphase is necessary to avoid any collision during robotrepositioning.

A later version of the system enables real-timetracking of organs that move with patient respira-tion (lung, liver, and kidney, for instance) [82.40]. Asregards patient respiration during treatment, several ap-proaches are used in conventional radiotherapy: the lessaccurate method consists of enlarging the targeted zoneto account for tumor motion; irradiating healthy tissuesis more acceptable than missing cancerous cells. Thisenlarged zone can be computed by combining infor-mation from two scanners acquired at the end of theexhale and inhale phases. A second approach consists ofsynchronizing dose delivery to a given stage of the res-piratory cycle (end of inhalation, for instance) but thisstage has to be reasonably repeatable and detected re-liably. In a more sophisticated approach, such as thatdeveloped in the Synchrony version of Cyberknife, thesystem tracks the organ motion and the robot followsthis motion during dose delivery in order to executethe planned treatment properly [82.41]. Since it is notpossible to obtain tracking information from x-ray im-ages – as such organs may not be visible on radiographsand continuous imaging would result in overirradia-tion of the patient – a localizer is introduced into thetreatment room. This localizer, associated to Rloc andcalibrated with respect to Riso, is used to track pas-sive markers placed on the patient’s chest; it deliversthe marker positions about 20 times per second. X-rayimages can be acquired about every 10 s. Because themotion of internal organs is different from chest exter-nal motion internal radiopaque markers are implantedclose to the tumor before the preoperative CT examina-tion. The relationship between the tumor position andthe internal markers is determined using the CT data;the internal markers enable initial patient setup by reg-istration of CT data to x-ray images. The relationshipbetween internal markers visible on x-ray images and

PartH

82.5

Computer and Robot-Assisted Medical Intervention 82.6 Specific Issues in Medical Robotics 1461

external markers tracked by the localizer is learnt ina preliminary stage where both data are acquired duringseveral breathing cycles. During dose delivery, betweentwo successive x-ray acquisitions, the position of in-ternal markers is interpolated from the position of theexternal markers and from the model; each new acqui-sition enriches the model by adding a new couple ofsynchronous positions of internal and external mark-ers. Finally, the position of the tumor can be deducedfrom the internal markers and fed back to the robotfor tracking and accurate dose delivery in spite of or-gan motion. The different components of the trackingfunction are illustrated in Fig. 82.4b. Current work dealswith noninvasive alternatives to the implanted internalmarkers.

As can be seen in this example, a lot of hardwareand software components are associated with the robotand many stages are involved in real-time determina-tion of the robot position with respect to the patient. Toprovide the required accuracy for precise dose delivery,

errors have to be reduced as far as possible for each ofthese elements; this is especially demanding. Moreover,because the robot develops very large forces and torquesin the immediate vicinity of the patient, reliability androbustness are mandatory. Unfortunately, as for manyother commercialized systems, due to intellectual prop-erty issues few data are available concerning detailedindustrial developments and technical testing.

The Cyberknife approach supposes the ability to in-stall a linear accelerator as end-effector of the robot;the Cyberknife robot typically carries 6 MeV accelera-tors which are quite compact and light compared withother radiation devices. When higher power is necessaryor when radiation beams are produced by synchrotronsand/or cyclotrons – for proton therapy, for instance –the radiation source cannot be positioned and orientedby a robot. Alternative systems [82.47] have been pro-posed to position the patient robotically relative to theradiation beam by means of parallel robotized seats orrobotized couches.

82.6 Specific Issues in Medical Robotics

As has been briefly introduced in the previous sections,medical robotics raises specific issues at the technical,clinical, and organizational levels in a very intricateway. This section discusses those issues in more detail.

Firstly this type of robot has to be used in a humanenvironment. Because these robotic systems generallyrequire close collaboration with a clinician, specificman–machine interface questions have to be resolved.In the case of surgery, the operator cannot interact easilywith classical man–machine interfaces because he/shehas to work under sterile conditions. This motivates thedevelopment of specific interfaces such as voice controlor foot pedals. In the case of comanipulation, the part ofthe robot that is held by the clinician must be made ster-ile before each intervention. Because the robot is used inclose proximity to human beings – at least the patient –safety issues are mandatory. Safety has to be demon-strated both at the hardware and software levels. Variousapproaches are possible [82.48]. The choice of specificarchitectures of robots or comanipulation control modesmay solve part of the problem. Norms and regulationsare intended to guarantee that the system is safe. How-ever, it would certainly be interesting to develop suchsystems using specific design methodologies introducedin critical applications (aeronautics, nuclear plants, etc.)to really anticipate the behavior of such a complex sys-

tem and avoid any misuse. This direction is still to beexplored.

The clinical use of the robot also requires elec-tromagnetic compatibility with the environment, inparticular in the OR. When the robot is used in spe-cific environments such as inside a CT or MRI imagingsensor, the robot must not disturb image acquisition orcorrupt the data. This may significantly constrain de-sign choices and selected materials: for instance in theMRI environment no ferromagnetic parts should be in-tegrated into the robot. Sterile cleaning of the robot isalso an issue. Bone-mounted robots, which come intovery close contact to the patient, must be completelycleanable. This is the case, for instance, for the Prax-iteles and LER-VIKY systems (Sect. 82.3.3), which areautoclavable. In the case of larger robots, their end-effector may be introduced inside sterile disposableplastic bags. However, the robot connection with thetool that is in contact with the patient has to be cleanedin a sterile way. Finally, the system and the robot mustbe designed such that the robotic procedure can beeasily and rapidly converted into the conventional in-tervention, at any time, in case of problems. This mayalso have important design consequences.

As mentioned previously, introducing a new med-ical device requires demonstrating the added value

PartH

82.6

1462 Part H Automation in Medical and Healthcare Systems

over existing techniques from a clinical perspective;for instance, using the robot could enable perform-ing less-invasive interventions, resulting in shorter stayof the patient in the hospital. In the same way, usinga robot to hold the endoscope may save one assistantwho can be transferred where his/her skills are usedin a much better way. However, evaluating such orga-nizational benefits obviously requires very precise costand resource management: all necessary human or ma-terial resources participating in the intervention, fromdiagnosis to long-term follow-up, have to be taken intoaccount. On a more medical level, using a robot tomachine bone for prosthesis placement may result inlonger lifetime of the prosthesis and/or fewer joint dis-junctions (for the hip, for instance). In a similar wayusing a complex radiotherapy treatment delivered us-ing a robot may allow dose escalation and may resultboth in better control of a tumor and fewer complica-tions. In general, it is not easy to predict how accurateand sophisticated a robotic procedure should be to makea significant difference on the clinical level. To provethis may also be very challenging. The evaluation ofclinical benefits may require long-term trials involv-ing several centers and many patients. Those clinicaltrials have to be conducted in accordance with the ethi-cal standards and regulations of the country concerned;those standards and regulations may be very strict, but

they vary from country to country. Finally, added valuemay also be evaluated in terms of commercial advantagefor a hospital that may attract more patients when hightechnology is used for painless, minimally invasive pro-cedures. This added value should be significant enoughto compensate for drawbacks related to the introductionof a robot such as the increase of procedure duration,which is often observed even when the learning pe-riod is finished. (See Chap. 77 for additional content onautomation in hospitals and healthcare.)

Cost is obviously another issue; indeed, severalof the distributed systems are quite expensive. If weconsider some of the systems (for instance, Robodoc,Caspar, DaVinci, and Cyberknife) listed in Table 82.1,costs average in the range US $1 000 000–2 000 000.Aesop cost around US $100 000. Frequently a mainte-nance cost of 10% per year has to be added, and someof these systems generate an extra cost per intervention(for instance, about US $1000–2000 for the DaVincisystem). This may be quite a heavy cost for hospitalsand clinics. Moreover, depending on healthcare fundingmodels in different countries, some of those costs maynot be affordable by health insurances. The higher theinvestment, the more significant the added value has tobe to justify the expense. More recently developed sys-tems (smaller and simpler robots, disposable devices)are likely to propose more affordable solutions.

82.7 Systems Used in Clinical Practice

For 20 years many medical robotic systems have beendeveloped in laboratories and evaluated to a certainextent. Evaluation is twofold: at the technical level itconsists of characterizing accuracy, reliability, robust-ness, etc. This stage may be realized on laboratorysetups using phantoms that mimic, more or less real-istically, the concerned part of the body. At the clinicallevel, experiments with corpses or animals enable a firstapproach to a more realistic evaluation of clinical fea-sibility and performance. Finally, a study on series ofpatients is always necessary to fully evaluate the sys-tem and its clinical added value. Relatively few medicalrobots have undergone the whole evaluation process,reached the market, and gone into wide clinical use.There are indeed two major challenges: how to turna laboratory prototype into a certified product, and howto make this product an industrial success. The rea-sons for this still limited diffusion of medical roboticsin the clinical world certainly come from the spe-cific constraints of medical robotics discussed above

and probably from the questionable added value of therobot in a number of cases. The complexity of clini-cal evaluation, certification, and marketing also makesthe process very long and expensive; for instance, inorthopaedics, demonstrating the advantage of robotsover competing techniques may take more than 10 yearssince the stability and life duration of prostheses cannotbe demonstrated any earlier. At the same time, to eval-uate them it is necessary to install robots in hospitals,sometimes at the company’s expense. Convincing a hos-pital to buy such expensive devices before any medicalevidence of their added value is available is particularlychallenging.

Table 82.1 attempts to list as largely as possiblethe industrial systems that are, or have been, signifi-cantly clinically used in routine and emerging products;this table is intended to give a flavor of the clini-cal spread of the technique. Numbers of systems areestimates established in early 2008. As can be seenseveral systems are no longer distributed: this de-

PartH

82.7

Computer and Robot-Assisted Medical Intervention 82.9 Medical Glossary 1463

serves further comments. Surgiscope and MKM, whichare both surgical microscope holders, probably faceda limited added value versus cost ratio. As regardsRobodoc, this system had not yet demonstrated a clini-cal benefit when misuse of the robot resulted in manyclinical complications and legal cases in Germany;this ended up with the removal of the robot fromthe US and European markets. Caspar, which offeredfunctions very similar to those of Robodoc, probablysuffered from the same unproven added value issueand from the failure of Robodoc. Aesop and Zeusare no longer distributed due to intellectual propertyconflicts. Aesop was, however, very successful in clin-ics and could certainly be considered as an industrialsuccess.

The success of DaVinci is probably due to its abilityto offer intrabody DOFs for endoscopic surgery; laparo-scopic radical prostatectomy is one of the main clinicalvectors of the large dissemination of the DaVinci sys-

tem. As regards Cyberknife, the ability to performcomplex dose distributions with tens of small radiationbeams and the capacity to radiate the tumor accuratelyduring patient respiration are probably the keys to itssuccess. Moreover, conventional radiation apparatus areexpensive devices and cost issues may be less critical inthe case of radiation therapy than for surgery.

Table 82.1 also shows that emerging systems areoften based on quite different design philosophy:small dedicated systems with more interactive controlschemes. As can be seen, favorite applications are inendoscopy and knee arthroplasty, where the market islarge (6 000 000 laparoscopic surgeries per year world-wide, and 600 000 knee prostheses per year worldwide).

Many other systems are somewhere in between theacademic and industrial worlds and are not included inthis table. Several of those forthcoming devices concernthe introduction of needles into the body (for biopsies,punctures, brachytherapy, etc.).

82.8 Conclusions and Emerging Trends

In this chapter, we have introduced the main motiva-tions for computer-assisted medical interventions andpresented the place of medical robots in this gen-eral paradigm. Different generations of robots havebeen proposed and evaluated worldwide, ranging fromsystems inspired from industrial automation to morespecific clinical robots. In parallel to those evolutionsin terms of robot architecture and application areas, wehave shown that the control modes also evolved in dif-ferent directions: giving a larger place to the operatorthrough real cooperation or closing the control loopwith real-time imaging data for tracking mobile and de-formable targets. The future is probably in the mergingof such type of controls; for instance, the robot mighthandle synchronization with a moving organ while theoperator would control the fine motions with respect tothe stabilized target.

As regards industrial products resulting from thisdomain, the evolution has been very similar. The

future will tell if those new design choices willresult in a much larger spread of medical robots.Reference [82.49] reports 39 000 service robots forprofessional use installed worldwide up to the end of2006, among which 9% (about 3500 devices) wouldbe medical robots. However, this report does not tellprecisely what is included in this category (for ex-ample, integration of haptic devices, integration ofmechanical localizers or others). From the numbersmentioned in Table 82.1, our estimate of installed robotsin 2008 would be closer to 2500, which is proba-bly less than half the number of installed navigationworkstations. Increasing this ratio requires careful se-lection of applications with significant added valueand the development of user-friendly cost-effectivesystems. Intrabody highly integrated mechatronic de-vices potentially open the range of applications ina dramatic way. This domain has still to be largelyexplored.

82.9 Medical Glossary

• Aneurysm: a local hernia on a blood vessel po-tentially resulting in vessel rupture and internalhaemorrhage

• Autoclavable: being cleanable in the autoclave(pressured vapor sterilization with temperaturesgreater than 100 ◦C)

PartH

82.9

1464 Part H Automation in Medical and Healthcare Systems

• Arthroplasty: plastic surgery of the joints involved,for instance, in joint replacement (e.g., hip, knee)• Biopsy: the action of taking samples of a tissue witha needle for further analysis• Brachytherapy: the introduction of radioactive seedsinto an organ for tumor destruction• Catheter: a flexible tube introduced into the body,typically blood vessels, for instance, to inject drugsor to place dilatation devices. Generally, the end ofthe catheter can be slightly curved by the physicianfrom the outside of the body• Colonoscopy: endoscopic examination of the colon• Computed tomography (CT): 3-D imaging fromx-ray acquisition, enabling good visualization ofbony structures and air cavities• Digitally reconstructed radiograph (DRR): a syn-thetic x-ray projection image computed from a CTvolume being given the position of a virtual sourceand virtual image plane• Endoscopic surgery: surgery involving a minimalaccess to the body through natural cavities or inci-sions and visualization of the internal organs usingrigid or flexible optical sensors (the endoscope)• EC marking: the certification from the EuropeanCommunity necessary for marketing a products ofany type in the Economic European Community; itinsures that the product complies with the Europeanregulations in terms of safety, health, environment,etc.• FDA: Food and Drug Administration; the US ad-ministration in charge of controlling the safety andefficacy of health-related products(http://www.fda.gov)• Interventional radiology: a category of therapeuticor diagnostic procedures executed under imagingcontrol

• Magnetic resonance imaging (MRI): 3-D imagingfrom magnetic resonance of hydrogen protons inthe body, enabling detailed visualization of soft tis-sues• Neurosurgery: brain or spine surgery• Operating room (OR): also called operating theater• Orthopaedics: a surgical specialty dealing withskeleton bones and joints• Positron emission tomography (PET): a functionalimaging modality in which a radioactive marker isassociated to a metabolically active molecule andinjected into the body of the patient; the metabolicactivity is traceable thanks to the radioactive markerand can be reconstructed in 3-D thanks to to-mography techniques similar to the one used forCT• Puncture: the action of inserting a linear tool (a nee-dle or an electrode for instance) into the body• Preoperative, intraoperative, postoperative: before,during, after the intervention• Prostatectomy: surgical removal of the prostate incase of cancer; laparoscopic prostatectomy is theminimally invasive version of this surgery• Radiopaque: visible on x-ray images• Radiotherapy: the destruction of pathologic tissues(mostly tumors) by ionizing particles• Stereotactic neurosurgery: a minimally invasive ac-cess to the brain requiring very accurate localizationof intracranial structures• Stereotactic frame: a mechanical device for perfectimmobilization of a patient’s skull, also used fortransferring the surgical plan and for guiding thesurgical tool• Ultrasonic examination: an imaging modality (2-Dor 3-D) based on the propagation of ultrasound inthe body; it visualizes tissue interfaces.

References

82.1 P. Cinquin, E. Bainville, C. Barbe, E. Bit-tar, V. Bouchard, L. Bricault, G. Champleboux,M. Chenin, L. Chevalier, Y. Delnondedieu, L. Des-bat, V. Dessenne, A. Hamadeh, D. Henry, N. Laieb,S. Lavallée, J.M. Lefebvre, F. Leitner, Y. Men-guy, F. Padieu, O. Peria, A. Poyet, M. Promayon,S. Rouault, P. Sautot, J. Troccaz, P. Vassal: Computerassisted medical interventions at TIMC laboratory:passive and semi-active aids, IEEE Eng. Med. Biol.Mag., special issue Robot. Surg. 14(3), 254–263(1995)

82.2 D.L.G. Hill, P.G. Batchelor, M. Holden, D.J. Hawkes:Medical image registration, Phys. Med. Biol. 46,R1–R45 (2001)

82.3 M. Chabanas, V. Luboz, Y. Payan: Patientspecific finite element model of the facesoft tissues for computer-assisted maxillofa-cial surgery, Med. Image Anal. 7(2), 131–151(2003)

82.4 E. Watanabe, T. Watanabe, S. Manaka, Y. Mayanagi,K. Takakura: Three-dimensional digitizer (neuro-navigator): A new equipment for CT guided

PartH

82

Computer and Robot-Assisted Medical Intervention References 1465

stereotactic neurosurgery, Surg. Neurol. 27, 543–547 (1987)

82.5 H.F. Reinhardt: Neuronavigation: a ten year re-view. In: Computer-Integrated Surgery, ed. byR.H. Taylor, R. Mosges, S. Lavallée (MIT Press, Cam-bridge 1996) pp. 329–341

82.6 P.J. Kelly, B.A. Kall, S.S. Goerss, F. Earnest:Computer-assisted stereotaxic laser resection ofintra-axial brain neoplasms, J. Neurosurg. 64(3),427–439 (1986)

82.7 Y.S. Kwoh, J. Hou, E.A. Jonckheere, S. Hayati: A ro-bot with improved absolute positioning accuracyfor CT guidedstereotactic brain surgery, IEEE Trans.Biomed. Eng. 35(2), 153–160 (1988)

82.8 S. Lavallée, J. Troccaz, L. Gaborit, P. Cinquin,A.L. Benabid, D. Hoffmann: Image guided oper-ating robot: a clinical application in stereotacticneurosurgery, Proc. IEEE Conf. Robot. Autom. (Nice1992) pp. 618–624

82.9 D. Glauser, H. Fankhauser, M. Epitaux, J.L. Helfti,A. Jaccottet: Neurosurgical robot Minerva: first re-sults and current developments, J. Image-Guid.Surg. 1(5), 266–272 (1995)

82.10 H.A. Paul, W.L. Bargar, B. Mittlestadt, B. Musits,R.H. Taylor, P. Kazanzides, J. Zuhars, B. Williamson,W. Hanson: Development of a surgical robot forcementless total hip arthroplasty, Clin. Orthopaed.Relat. Res. 285, 57–66 (1992)

82.11 W.L. Bargar, A. Bauer, M. Borner: Primary and re-vision total hip replacement using the Robodoc(R)system, Clin. Orthopaed. Relat. Res. 354, 82–91(1998)

82.12 J. Sackier, Y. Wang: Robotically assisted laparo-scopic surgery: from concept to development, Surg.Endosc. 8, 63–66 (1994)

82.13 S. Aiono, J.M. Gilbert, B. Soin, P.A. Finlay, A. Gor-dan: Controlled trial of the introduction of a roboticcamera assistant (EndoAssist) for laparoscopiccholecystectomy, Surg. Endosc. 16(9), 1267–1270(2002)

82.14 G.S. Guthart, J.K. Salisbury: The Intuitive telesur-gery system: overview and application, Proc. IEEERobot. Autom. Conf. (San Francisco 2000)

82.15 J. Troccaz, M. Peshkin, B.L. Davies: Guidingsystems: introducing synergistic devices and dis-cussing the different approaches, Med. Image Anal.2(2), 101–119 (1998)

82.16 M. Jakopec, F. Rodriguez y Baena, S.J. Harris,J. Cobb, B.L. Davies: The hands-on orthopaedicrobot Acrobot: early clinical trials of total knee re-placement surgery, IEEE Trans. Robot. Autom. 19(5),902–911 (2003)

82.17 J. Troccaz, Y. Delnondedieu: Semi-active guid-ing systems in surgery. A two-DOF prototype ofthe passive arm with dynamic constraints (PADyC),Mechatronics 6(4), 399–421 (1996)

82.18 O. Schneider, J. Troccaz: A six degree of freedompassive arm with dynamic constraints (PADyC) for

cardiac surgery application: preliminary experi-ments, Comput. Aided Surg., special issue Med.Robot. 6(6), 340–351 (2001)

82.19 S.C. Ho, R.D. Hibberd, B.L. Davies: Robot assistedknee surgery, IEEE Eng. Med. Biol. Mag. 14(2), 292–300 (1995)

82.20 Mako, www.makosurgical.com82.21 A. Vilchis, J. Troccaz, P. Cinquin, K. Masuda,

F. Pellissier: A new robot architecture for tele-echography, IEEE Trans. Robot. Autom., specialissue Med. Robot. 19(5), 922–926 (2003)

82.22 P. Berkelman, P. Cinquin, E. Boidard, J. Troccaz,C. Letoublon, J.-M. Ayoubi: Design, control, andtesting of a novel compact laparascopic endoscopemanipulator, J. Syst. Control Eng. 217(14), 329–341(2003)

82.23 E. Taillant, J. Avila-Vilchis, C. Allegrini, I. Bricault,P. Cinquin: CT and MR compatible light puncturerobot: architectural design and first experiments.In: Medical Image Computing and Computer As-sisted Interventions, LNCS, Vol. 2, ed. by G. Goos,J. Hartmanis, J. van Leeuwen (Springer, Berlin Hei-delberg 2004) pp. 145–152

82.24 C. Plaskos, P. Cinquin, S. Lavallée, A. Hodgson:Praxiteles: a miniature bone-mounted robot forminimal access total knee arthroplasty, Int. J. Med.Robot. Comput. Aided Surg. 1(4), 67–79 (2006)

82.25 M. Shoham, M. Burman, E. Zehavi, L. Joskowicz,E. Batkilin, Y. Kunicher: Bone-mounted miniaturerobot for surgical procedures: concept and clini-cal applications, IEEE Trans. Robot. Autom., specialissue Med. Robot. 19(5), 893–901 (2003)

82.26 B. Kim, Y. Jeong, H. Lim, J. Park, A. Menciassi,P. Dario: Functional colonoscope robot system,Proc. IEEE Int. Conf. Robot. Autom. (2003)

82.27 N.A. Patronik, C.N. Riviere, S. El Qarra, M.A.Zenati: The HeartLander: A Novel Epicardial Crawl-ing Robot for Myocardial Injections Int. Congr. Ser.(CARS’2005), Vol. 1281 (Elsevier, Amsterdam 2005)pp. 735–739

82.28 M.A. Rentschler, S.M. Farritor, K.D. Iagnemma:Mechanical design of robotic in vivo wheeled mo-bility, J. Mech. Des. 129, 1037–1045 (2007)

82.29 K.B. Yesin, K. Vollmers, B.J. Nelson: Modeling andcontrol of untethered biomicrorobots in a fluidicenvironment using electromagnetic fields, Int. J.Robot. Res. 21(5/6), 527–536 (2006)

82.30 K. Ikuta, M. Tsukamoto, S. Hirose: Shape memoryalloy servo actuator with electric resistance feed-back and application to active endoscope, Proc.IEEE Int. Conf. Robot. Autom. (Philadelphia 1988)pp. 427–430

82.31 T. Fukuda, S. Guo, K. Kosuge, F. Arai, M. Negoro,K. Nakabayashi: Micro active catheter system withmulti degrees of freedom, Proc. IEEE Int. Conf.Robot. Autom. (San Diego 1994) pp. 2290–2295

82.32 www.givenimaging.com82.33 www.rfnorika.com

PartH

82

1466 Part H Automation in Medical and Healthcare Systems

82.34 R. Kumar, P. Berkelman, P. Gupta, A. Barnes,P.S. Jensen, L.L. Whitcomb, R.H. Taylor: Prelim-inary experiments in cooperative human/robotforce control for robot assisted microsurgical ma-nipulation, Proc. IEEE Int. Conf. Robot. Autom. (SanFrancisco 2000) pp. 610–617

82.35 W.H. Zhu, S.E. Salcudean, S. Bachmann, P. Abol-maesumi: Motion/force/image control of a diag-nostic ultrasound robot, Proc. IEEE Int. Conf. Robot.Autom. (San Francisco 2000) pp. 1580–1586

82.36 E. Dombre, X. Thérond, E. Dégoulange, F. Pierrot:Robot-Assisted detection of atheromatous plaquesin arteries, Proc. IARP Workshop Med. Robot. (Vi-enna 1996) pp. 133–140

82.37 E. Dombre, G. Duchemin, P. Poignet, F. Pierrot.Dermarob: A safe robot for reconstructive surgery,IEEE Trans. Robot. Autom., special issue Med.Robot. 19(5), 876–884 (2003)

82.38 A. Schweikard, M. Bodduluri, J.M. Adler: Plan-ning for camara-guided robotic radiosurgery, IEEETrans. Robot. Autom. 14(6), 951–962 (1998)

82.39 M. Bodduluri, J.M. McCarthy: X-ray guided roboticradiosurgery for solid tumors, Proc. IEEE/ASME Int.Conf. Adv. Intell. Mechatron. (Como 2001) pp. 1065–1069

82.40 A. Schweikard, G. Glosser, M. Bodduluri, M.J. Mur-phy, J.R. Adler: Robotic motion compensation forrespiratory movement during radiosurgery, Com-put. Aided Surg., special issue Plan. Image Guid.Radiat. Ther. 5(4), 263–277 (2000)

82.41 A. Schweikard, H. Shiomi, J. Adler: Respirationtracking in radiosurgery without fiducials, Int. J.Med. Robot. Comput. Assist. Surg. 1(2), 19–27 (2005)

82.42 Y. Nakamura, K. Kishi, H. Kawakami: Heart beatsynchronization for robotic cardiac surgery, Proc.IEEE Int. Conf. Robot. Autom. (Seoul 2001) pp. 2014–2019

82.43 A. Krupa, J. Gangloff, C. Doignon, M.F. de Mathelin,G. Morel, J. Leroy, L. Soler, J. Marescaux: Au-tonomous 3-D positioning of surgical instrumentsin robotized laparoscopic surgery using visual ser-voing, IEEE Trans. Robot. Autom., special issueMed. Robot. 19(5), 842–853 (2003)

82.44 S. Voros, J.A. Long, P. Cinquin: Automatic detec-tion of instruments in laparoscopic images: a firststep towards high-level command of robotic en-doscopic holders, Int. J. Robot. Res. 26(11-12),1173–1190 (2007)

82.45 M.A. Vitrani, G. Morel, T. Ortmaier: Automatic guid-ance of a surgical instrument with ultrasoundbased visual servoing, Proc. IEEE Int. Conf. Robot.Autom. (Barcelona 2005) pp. 508–513

82.46 P. Abolmaesumi, S.E. Salcudean, W.H. Zhu,M.R. Sirouspour, S.P. DiMaio: Image-guided con-trol of a robot for medical ultrasound, IEEE Trans.Robot. Autom. 18(1), 11–23 (2002)

82.47 S. Pinault, G. Morel, M. Auger, R. Ferrand, G. Mabit:Using an external registration system for daily pa-tient repositioning in protontherapy, Proc. IEEE/RSJInt. Conf. Intell. Robot. Syst. (IROS’07) (San Diego2007) pp. 4289–4294

82.48 B.L. Davies: A discussion of safety issues for med-ical robots. In: Computer-Integrated Surgery, ed.by R.H. Taylor, R. Mosges, S. Lavallée (MIT Press,Cambridge 1996) pp. 287–296

82.49 www.worldrobotics.org

PartH

82