evaluation of a telerobotic system to assist surgeons in microsurgery

Biomedical Paper

Evaluation of a Telerobotic System to AssistSurgeons in Microsurgery

Hari Das, SC.D., Haya Zak, Ph.D., Jason Johnson, B.S., John Crouch, B.S., andDon Frambach, M.D.

Jet Propulsion Laboratory, California Institute of Technology, Pasadena (H.D., H.Z., J.J.), Universityof Southern California, School of Medicine, Los Angeles, (J.C.), and Miramar Eye Specialists, Ventura,

(D.F.), California, USA

ABSTRACT A tool was developed that assists surgeons in manipulating surgical instruments moreprecisely than is possible manually. The tool is a telemanipulator that scales down the surgeon’s handmotion and filters tremor in the motion. The signals measured from the surgeon’s hand aretransformed and used to drive a six-degrees-of-freedom robot to position the surgical instrumentmounted on its tip. A pilot study comparing the performance of the telemanipulator system againstmanual instrument positioning was conducted at the University of Southern California School ofMedicine. The results show that a telerobotic tool can improve the performance of a microsurgeonby increasing the precision with which he can position surgical instruments, but this is achieved at thecost of increased time in performing the task. We believe that this technology will extend thecapabilities of microsurgeons and allow more surgeons to perform highly skilled procedures cur-rently performed only by the best surgeons. It will also enable performance of new surgical proce-dures that are beyond the capabilities of even the most skilled surgeons. Comp Aid Surg 4:15–25(1999). ©1999 Wiley-Liss, Inc.

Key words: telemanipulator, position scaling, assistance for microsurgery, performance evaluation

INTRODUCTIONSurgical operations on the eye, ear, brain, nerves, andblood vessels require precise positioning of surgicalinstruments because of the minute size of features tobe manipulated in the surgical field. Surgeons oftenuse a microscope to help them see these featureswhile manipulating the instruments with their hands.Much effort is spent teaching microsurgeons tech-niques to reduce the tremor in their hands to positioninstruments with great precision and manipulate themsmoothly. In this article we report results from eval-uation experiments conducted on a telerobotic system

that was developed to assist surgeons in overcomingthese limits in their manual dexterity.

Telerobots have traditionally been developedfor use in hazardous environments and for spaceexploration.2 Their use as tools to help surgeons inmicrosurgery is a relatively recent concept.5,18 TheRobot Assisted MicroSurgery (RAMS) sys-tem4,13,14evaluated in this article allows a surgeonto command motions for a surgical instrument us-ing an input device that precisely measures hishand motions in six degrees of freedom. The mea-

Received May 19, 1998; accepted March 15, 1999.

Address correspondence/reprint requests to: Hari Das, 4800 Oak Grove Dr. MS 198-219, Pasadena, CA 91109. E-mail:[email protected].

Computer Aided Surgery 4:15–25 (1999)

©1999 Wiley-Liss, Inc.

surements are read into a computational subsystemwhere they are transformed then used to drive asix-degrees-of-freedom robot that holds a surgicalinstrument. The surgical instrument thus replicatesin some fashion the motions entered by the surgeonat the input device in real time.

Advantages of using this method to controlthe surgical instrument are:

● The surgeon’s hand motions can be scaleddown then used to drive the robot holding thesurgical instrument, thus positioning the in-struments more precisely than is possiblemanually.

● It is possible to remove unwanted motions inthe surgeon’s hands—for example, eliminat-ing tremor by filtering the signals before usingthem to drive the surgical instrument.

● Forces sensed at the surgical instrument canbe amplified, then haptically presented to thesurgeon’s hands.

● Advanced computer control of routine func-tions, or operator- and computer-shared controlof the telerobotic system based on sensed inter-action data or moving within modeled geomet-rical constraints, can improve the performanceof a surgeon by reducing the mental or physicaleffort needed to perform difficult and complexprocedures. Examples of this include limitingthe maximum force applied at the surgical in-strument, automatically tracking a moving sur-face, and limiting motion of the surgical instru-ment to avoid sensitive tissue.

Development of practical systems for assist-ing microsurgeons is a growing field of researchand commercial interest. Microrobotic systems de-veloped for biomedical applications have a richvariety of innovations. We list here some of theresearch recently conducted in this area. Hunter etal.9 described a parallel-link design of a five-de-grees-of-freedom telerobotic system and its associ-ated controls and virtual environment that was ca-pable of extremely precise positioning. Amagnetically levitated telerobot developed by Sal-cudean et al.12 used a macro positioning robot witha mini master–slave telerobot mounted on its tip forprecise microsurgery. Dario et al.6 described a teth-ered microrobotic capsule with miniature robotarms designed for traveling up the colon for use incolonscopy. Hannaford et al.8 reported a voicecoil–actuated manipulator design for long-range te-leoperation with micrometer precision control, asdemonstrated in protein crystal growth experi-

ments. A two-fingered manipulation system used tograsp objects 2mm in size was developed by Tani-kawa et al.16 Much effort is spent developing ro-botic systems for surgery, but relatively little efforthas been devoted to quantifying their benefits incontrolled experiments. Qualitative assessment ofbenefits of prototypes in practical applications pro-vides an independent measure of the potential ofnew technology. Although there is increasing ac-tivity in the field of research and development oftelemanipulation systems for microsurgery, we areunaware of other human performance experimentsbeing conducted with randomized variables andmultiple subjects to compare a microsurgical tel-erobotics system with manual performance. Thework reported in this article is an attempt to vali-date the utility of the RAMS research prototype byperforming a pilot experiment comparing the per-formance of the RAMS system with a purely man-ual method of precision instrument positioning. Wehave thereby attempted to determine the benefits ofa telerobotic tool for precise positioning of instru-ments.

Many well-designed experiments on humanperformance in manual control have been reportedin recent years. The experimental procedure re-ported in this article is similar to procedures othershave used in evaluating human performance. In onestudy, a comparison of the effect of different con-trol modes on operator performance was con-ducted. In the experiment, multiple subjects used ateleoperation system designed for space applica-tions to perform a peg-in-hole insertion task.7 Sim-ilar evaluations were used in a study comparingtelerobotic control of a laparoscopic camera usingthe AESOP system with manual control of thecamera.10 Tendick et al.17 studied imaging systemsin a knot-tying task. They concluded, based onanalysis of variance (ANOVA), that direct viewingwas superior in performance but comparable intime taken to complete tasks when compared withthree other videoscopic system modalities. Adel-stein and Ellis1 studied the effect on human spatialsituation awareness of the addition of roll control,in addition to pan and tilt control, when viewing aremote or computer-synthesized scene. Their re-sults from ANOVA indicated that their subjectsperformed equally well with and without the rollcontrol.

The RAMS system is a telerobot with me-chanical robot arms interacting with the environ-ment and the surgeon’s hands under the control ofcomputers. Elements of the RAMS system are me-chanical, electronics, real-time control, and user

16 Das et al: Telerobotic System for Microsurgery

interface software subsystems. These are brieflydescribed in the next section. An experiment con-ducted on the RAMS system at the University ofSouthern California School of Medicine is de-scribed in the subsequent section, and the finalsection of the article reports results from the testsand concludes with a discussion of the current stateof the work and future prospects.

TELEMANIPULATOR SYSTEMDESCRIPTIONA drawing of the components of the RAMS systemis shown in Figure 1. The surgeon holds the handleof the master input device as he would a surgicalinstrument. The master device is used to commandmotions for the slave-held instrument. Hand mo-tions are read into the real-time computing system,where they are processed then used to drive theslave robot.

A graphical user interface (GUI) imple-mented on a workstation (and on a laptop personalcomputer for field testing) is used to configure theparameters of the system—for example, setting ra-tios for position scaling. The components of theengineering system can be classified into four sub-

systems: mechanical, electronics, servo-control,and configuration and user-interface software. Theblock diagram in Figure 2 illustrates the interac-tions between the subsystems.

The mechanical subsystem consists of themaster and slave robots. The master device is ableto sense its handle position in six degrees of free-dom with an accuracy of 30mm. The slave robotcan position a surgical instrument mounted at its tipwith an accuracy of 12mm in six degrees of free-dom. The workspace of the slave robot is a hemi-sphere with a diameter of 30 cm, while the masterhas a 53 5 3 5-cm cubic workspace. The elec-tronics subsystem consists of motor amplifiers, asafety electronics module, and its associated relaysto monitor faults in the system and act on them.These elements ensure that a number of potentialerror conditions are handled quickly and gracefully.The servo-control subsystem consists of both hard-ware and software. The hardware is composed ofthe DSP-based PMAC servo-control boards to con-trol the joints of the master and slave arms and aMotorola MVME-167 (MC68040 processor-based)board installed in a VME chassis. The software onthe servo-control boards positions the joints of the

Fig. 1. Components of the RAMS system.

Das et al: Telerobotic System for Microsurgery 17

slave arm and determines the torque at the joints ofthe master arm based on signals from the high-levelconfiguration software subsystem. Software on theMVME-167 board interfaces signals from the high-level configuration software subsystem to theservo-control subsystem. The configuration controland user-interface software subsystem reads inputsfrom a GUI and sets the appropriate configurationparameters on the RAMS system, computes thekinematics and high-level control functions, anddetermines the signals with which to drive themaster and slave arms. The engineering design ofthe RAMS system has been reported with greaterdetail by Charles et al.4 A photograph of the RAMSsystem is shown in Figure 3.

We use a fairly simple and conventional con-trol method to achieve master-to-slave control. Themaster robot forward kinematics algorithm com-putes the master handle position and orientationfrom master joint positions read in from the servocontrol cards through shared memory. The handleincremental motion is computed then filtered(based on the GUI filter bandwidth setting). It isthen scaled (based on a GUI parameter setting) andtransformed to a slave base-referenced coordinateframe. The incremental motion is added to the slavetip position and orientation (computed from mea-sured slave joint positions by the slave forwardkinematics algorithm). The new desired slave tipposition and orientation is fed to an inverse kine-matics routine that computes the correspondingjoint positions. The result is used to command the

new positions of the joints of the slave throughshared memory and the servo control boards. Al-though the RAMS system is currently capable ofreflecting forces measured from the slave robot tipback to the master handle, force feedback was not

Fig. 2. Subsystems of the RAMS system.

Fig. 3. Photograph of a prototype of the RAMS system.


used in the experiment because it had not beenimplemented at the time.

EVALUATION EXPERIMENTAND RESULTSIn this section, we report a pilot study conducted atthe Doheny Eye Institute at the University ofSouthern California (USC) School of Medicine andat the Jet Propulsion Laboratory (JPL) to determinethe advantages of the RAMS system over manualperformance in a simple probe-positioning task. Dr.M. Siemionow, a hand surgeon at the ClevelandClinic Foundation, evaluated an early prototype ofthe RAMS system in 1997. A single subject (amicrosurgery resident) performed a comparison ofmanual versus telerobotic operation in seven mi-crosurgical procedures. Dr. Siemionow reportedthat in five of seven procedures tested, the RAMSprototype performed better than manual tech-niques.15 In contrast to the Cleveland Clinic study,a randomized variables study with multiple sub-jects was performed to obtain a statistical evalua-tion of the performance of the RAMS system. Thegoal of the experiment was to compare the abilityof human subjects to position a probe precisely in amicroscopic field using the RAMS system withperformance of the same task while manually hold-ing a similar probe. Rather than use complex mi-crosurgical procedures as tasks to be performed inour experiment, as was done in the ClevelandClinic study, we chose a simple probe-positioningtask. The reasons for this were:

● Performance is measured by a blind data col-lection system, so objective measures of thedifference in performance could be imple-mented.

● The aspect of performance tested (ability toposition precisely) is well defined. Humanmanual dexterity is a complex motor skill andwe wanted to limit ourselves to simple aspectsof it.

● The simplicity of the task enabled multiplerepetitions to be performed by each subject inorder to obtain a statistical measure of thedifference between manual and teleroboticperformance of the task.

● The conduct of the experiment was automatedand controlled by the data collection system,thereby simplifying the experimental proce-dures, and the simple procedures make it easyto enforce identical experimental conditionsfor the subjects within the randomized vari-ables.

● A robust task board was designed and fabri-cated to withstand repeated abuse duringtraining and experimentation. It would havebeen difficult to keep a more complex taskboard from deteriorating during the experi-ment. Repetitions of the tests were easily per-formed because of the minimal setup timerequired.

DescriptionThe experiment conducted was to have human sub-jects perform a probe tip-positioning task. Twooptions were available for performing the task:manual positioning and telerobotic positioning. Thesetup, named manual positioning, is shown in Fig-ure 4. A microscopic task board was designed withtargets, each 0.003 inches in diameter and 0.025inches apart, arranged in a rectangular 43 6 arrayon a grid, as shown in Figure 5. The task for thesubjects was to touch the targets in a specifiedsequence, as shown by the gray arrows, using ahand-held probe without touching the background.In the telerobotic positioning mode, the subjectsperformed the identical task but used the telerobotinput device controlling an identical probe held bythe slave robot.

Three groups of subjects were selected.Group 1 consisted of seven second-year medicalstudents at the USC School of Medicine, group 2consisted of seven ophthalmology surgeons at theDoheny Eye Institute, and group 3 was composedof nine robotics engineers at JPL. Training and theactual experiment were conducted in one continu-ous session. A preexperiment test showed that sub-jects required more training with the robotic systemthan with the hand-held probe mode to reach a

Fig. 4. Data collection for probe positioning task shownin manual configuration.


leveling off of their performance improvement. Be-cause of the limited availability of the medicalstudents and surgeons, the training time for thesetwo groups of subjects was limited to performingeight repetitions of the task with the teleroboticsystem and two repetitions with the hand-heldprobe. The complete session for a subject lasted anaverage of about 30 min. The tests with groups 1and 2 were conducted at the Doheny Eye Institute.The tests on the JPL robotics engineers were con-ducted at JPL, and their training consisted of per-forming 18 repetitions of the task with the telero-botic system (i.e., twice the training received by themedical students and ophthalmology surgeons) andtwo repetitions with the hand-held probe. The com-plete session for a subject lasted an average ofabout 60 min.

Data collected were the number of errors (i.e.,contacts between the probe and the background)and the time taken to complete the task. The datawere collected by the data acquisition system im-plemented on a laptop computer. The experimenterinitialized the data acquisition system at the start ofeach task. The subject started the clock on thetiming of the task and the collection of data bytouching the first target with the tip of the probe.The subject then proceeded by touching the nexttarget with the probe tip upon hearing an audiblebeep indicating that he had successfully touchedthe current target. The process was repeated foreach of the 10 targets. Errors were logged as the

accumulation of contacts at the data acquisitionsample rate between the probe and the background.The task was completed by successfully touchingthe targets in sequence until the final target wastouched. This stopped the clock from timing thetask and recording accumulating errors in perform-ing the task.

Tremor filtering parameters on the teleroboticsystem were varied during the experiment to deter-mine whether alternatives in those parameters af-fected the performance of the subjects when usingthe telerobotic system. The tremor filter setting waschosen to limit the hand motion to 5 or 30 Hz. The30-Hz setting was chosen to pass through to theinstrument tip the full bandwidth of the subject’shand motion. The 5-Hz setting was conservativelychosen to eliminate tremor. In the human arm,tremor is reported to be on the order of 6–12 Hz.11

In the braced hand configuration used in this ex-periment, tremor has been reported to be in the 7-to 12-Hz range.3 Subjects were also asked to pro-vide feedback on a questionnaire at the end of theirexperimental session regarding their impressions ofthe experiment and their performance in the differ-ent modes of operation offered.

A typical session with a subject at the DohenyEye Institute (and at JPL) went as follows:

● The procedure to be followed in conductingthe experiment was verbally described to thesubject.

Fig. 5. Rectangular array of targets for probe-positioning task and closeup of target.


● The task was performed once using the telero-botic system as a trial run to familiarize thesubject with the procedure.

● Data were collected during two trials (three atJPL) of manually performing the task.

● Training was provided in two trials (three atJPL) while manually positioning the probe.

● A training procedure for the use of the RAMSsystem followed. The tremor filter settingswere randomly set to either 5 or 30 Hz for thenext eight trials (18 at JPL)

● The experiment was run with data collectionfor analysis. Two tests (four at JPL) were runwith one setting of the tremor filter, then an-other two tests (four at JPL) were run with theother setting of the tremor filter.

● The subject filled out a questionnaire on thetest.

Data Analysis and ResultsMeans and standard deviations for the respectivegroups for normalized time taken to complete thetest are shown in Figure 6. Normalized errors in-curred during the test comparing telerobotic tomanual performance are shown in Figure 7. Nor-malization of data was done by dividing the respec-tive data by the average of the mean of the manualand RAMS data. The reason for normalization wasto determine a performance indicator-reflected im-provement or degradation in performance with re-spect to the capability of the subject. The data from

one subject from group 3 were discarded as anoutlier because they represented data points threestandard deviations from the mean. The data arepresented for the combined pool of subjects (G-All)and for the groups (G-1, G-2, and G-3) of subjects.

On the plots above, the manual mode refers tothe mode of performing the task with the hand-heldprobe. The RAMS-combined mode is the combineddata for both 5 and 30 Hz tremor filtering using theRAMS system. The RAMS-5 Hz mode representsdata from the use of the RAMS system to perform thetask with the tremor filter set to cutoff motion above5 Hz; and similarly, the RAMS-30 Hz mode is withthe tremor filter set to cutoff motion above 30 Hz.

The data for normalized task completiontimes clearly show that subjects performed thetasks significantly more quickly with the manualmode. The data for normalized penalties accumu-lated are more interesting, and a single-factorANOVA was performed on the penalties accumu-lated data to discern differences between the alter-native modes available. We use thep value torepresent the difference between any two sets ofdata. If the two populations really had the samemean, thep value was the probability that randomsampling would result in means as far apart (ormore so) as those observed in the experiment. Asmall p value indicated that the probability wassmall. For statistical significance, ap value of#.05was generally accepted as a significant difference,i.e., unlikely to be produced by random sampling of

Fig. 6. Normalized task completion times (means and standard deviation).


distributions with the same mean. The results arepresented in Table 1. Tests that had a significantresult are shown in bold.

At the end of each session, the subjects wereasked to fill out a questionnaire. Two easily quan-tifiable questions asked were:

● Did the subjects feel that the RAMS systemimproved or worsened their performancecompared to manual positioning of the probe?

● Did subjects feel that with more training theywould be able to improve their performanceusing the RAMS system compared to manualpositioning of the probe?

The subjects were also asked to rate on ascale of 1 to 5 (15 highly unconfident; 55 highlyconfident) their confidence using the manual,

RAMS–5 Hz, and RAMS–30 Hz modes for per-forming the task. The responses were:

● In the group of medical students (group 1),four felt that the RAMS system improvedtheir performance and three were either notsure or felt that they were the same.

● All subjects in group 1 felt that their perfor-mance with the RAMS system would be betterwith more training.

● In the group of surgeons (group 2), four feltthat the RAMS system improved their perfor-mance and three felt that it worsened theirperformance.

● Five of seven surgeons felt that their perfor-mance would improve with training, and twofelt that they were not sure, or did not know.

Fig. 7. Normalized penalties accumulated (means and standard deviation).

Table 1. ANOVA Test Results of Operator Performance Comparing Pairs of Alternative Modesof Performing the Task

ANOVA Test Hypothesisp

ValueFor all subjects: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the manual mode .021For group 1: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the manual mode .008For group 2: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the manual mode .543For group 3: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the manual mode .019

For all subjects: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the RAMS-5 Hz mode .003For group 1: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the RAMS-5 Hz mode .026For group 2: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the RAMS-5 Hz mode .458For group 3: RAMS-30 Hz mode resulted in fewer normalized penalties accumulated than the RAMS-5 Hz mode .036


● In group 3 (JPL engineers), six of nine sub-jects felt that their performance improved withthe RAMS system, two were not sure or didnot know, and one felt that his performanceworsened.

● Seven subjects in group 3 felt that their per-formance would be better with more training,and two were not sure or did not know.

The ratings results are plotted in Figure 8, and thecorresponding ANOVA results are shown in Table2. The following summarizes the results and isfollowed by a discussion of the details.

● Task completion times are significantlyshorter in the manual mode than when usingthe RAMS with either the 30- or 5-Hz tremorfilter.

● The performance of the group of ophthalmol-ogy surgeons (group 2) did not indicate a

conclusive advantage with any of the modesof performing the task.

● Groups 1 and 3 incurred significantly fewernormalized penalties with the RAMS systemset to the 30-Hz tremor filter than in the man-ual mode; i.e., the RAMS system was shownto increase positioning accuracy relative tothat achieved manually.

● The pooled groups of subjects and groups 1and 3 incurred fewer normalized penaltieswith the RAMS system set to the 30-Hztremor filter than with the RAMS system setto the 5-Hz tremor filter.

● The subjects felt more confident using theRAMS system (30 Hz) than the manual mode.

● The subjects felt that their performance withthe RAMS system would be even better ifthey were given more training on it.

Fig. 8. Subjective ratings of different modes (means and standard deviation).

Table 2. ANOVA Test Results of Comparison of Subject Ratings between Pairs of Modes

ANOVA Test Hypothesisp

ValueFor all subjects: Subjects felt more confident using the RAMS-30 Hz mode versus the manual mode. .017For group 1: Subjects felt more confident using the RAMS-30 Hz mode versus the manual mode. .033For group 2: Subjects felt more confident using the RAMS-30 Hz mode versus the manual mode. .534For group 3: Subjects felt more confident using the RAMS-30 Hz mode versus the manual mode. .083

For all subjects: Subjects felt more confident using the RAMS-30 Hz mode versus the RAMS-5 Hz mode. .000For group 1: Subjects felt more confident using the RAMS-30 Hz mode versus the RAMS-5 Hz mode. .006For group 2: Subjects felt more confident using the RAMS-30 Hz mode versus the RAMS-5 Hz mode. .073For group 3: Subjects felt more confident using the RAMS-30 Hz mode versus the RAMS-5 Hz mode. .002


DISCUSSIONThe experiment showed that the time taken to per-form the precision positioning task clearly wasclearly longer when using the RAMS system. Theresults for errors occurring (normalized penaltiesaccumulated) were mixed. The tremor filter set tolimit hand motion to below 5 Hz appeared to de-grade performance of most subjects despite beingdesigned to eliminate tremor in hand motion. Eventhough the groups of medical students (group 1)and JPL robotics engineers (group 3) had a signif-icant advantage with the RAMS system set to the30-Hz filter, the group composed of ophthalmologysurgeons did not appear to have a significant ad-vantage when using the 30-Hz tremor filter in theRAMS system. Groups 1 and 3 performed signifi-cantly better with the RAMS system when itstremor filter was set to allow hand motion below 30Hz than with the hand-held probe. The ophthalmol-ogy surgeons did not have a significant perfor-mance improvement when comparing the RAMSsystem to the manual mode.

The results from the questionnaire confirmthe earlier conclusions. Generally, subjects signif-icantly preferred the RAMS system with the 30-Hztremor filter setting over the RAMS system withthe 5-Hz tremor filter. For the pooled groups ofsubjects, there is also a significant preference forthe RAMS system with the 30-Hz tremor filter overthe manual mode. There was no indication of pref-erence for the group of surgeons, however.

The surgeons who participated in the experi-ment were available only as their schedules permit-ted. In some cases, the experimental session wasinterrupted by the urgent needs of the subject’spatients and the experimental session was resumedwhen the subject became available again. In mostcases, the subjects were in a hurry to complete thetest. Some also appeared to be nervous about mea-surements made on their manual dexterity. Theseeffects could have had a significant impact on theirperformance.

The experiment that we conducted has pro-vided much practical insight into issues in conduct-ing such tests.

● Subjects had varied preferences in the config-uration of the experimental equipment (e.g.,position of the input device, position of themicroscope) and it would have been advisableto have had the configuration be easily adjust-able to adapt to the subject’s preferences.

● The conditions under which subjects partici-pated greatly affected their performance. The

surgeons in our study participated under rela-tively stressful conditions. Performance wasalso dependent on the motivation of subjectsto perform well. A third factor that affectedperformance was the amount of training sub-jects received. These factors should be con-trolled to minimize their effect on the exper-imental result.

● Alternative measures provide different indicesof performance and the metric used shouldaccurately reflect the performance index ofinterest. The measure of penalties in this ex-periment accumulated errors at a fixed rateover time so that a slower experimental run ofthe exact same motion would result in agreater number of penalties.

● The experimental equipment should have re-placement parts for elements that may wearout or break down during the experiment.

● Thorough testing of the equipment in experi-ment-like conditions can provide experiencethat is helpful when running the actual exper-iment. Using a research prototype in thesetests raised issues of reliability and robustnessof the prototype and experimental fixtures, inaddition to the usual problems of dealing withthe variability and performance of human sub-jects.

The research effort to develop the RAMSsystem was recently completed at JPL, so confir-mation of these results with further experiments onthe same hardware is no longer possible. However,the results learned from this development and theexperiment conducted indicate that there is poten-tially a significant benefit to be gained from atelerobotic tool designed to assist surgeons in per-forming microsurgery.

ACKNOWLEDGMENTS

This work was done at the Jet Propulsion Labora-tory, California Institute of Technology, under acontract with the National Aeronautics and SpaceAdministration (NASA). The RAMS system devel-opment is the result of a collaboration betweenNASA/Jet Propulsion Laboratory and Dr. SteveCharles, and MicroDexterity Systems, Inc. The ex-periment described in this article was conducted atthe Doheny Eye Institute, University of SouthernCalifornia, and at the Jet Propulsion Laboratory.The authors thank their subjects for kindly partic-ipating in the experiment.


REFERENCES1. Adelstein BD, Ellis SR. Effect of head-slaved visual

image roll on spatial situation awareness. In: Proceed-ings of the 37th annual meeting of the Human Factorsand Ergonomics Society, Seattle, Washington; 1993.

2. Bejczy AK. Robots as man-extension in space. In:Proceedings of the Ninth World Congress of the In-ternational Federation of Automatic Control, Budap-est, Hungary; 1984.

3. Bose B, Kalra AK, Thukral S, Sood A. Tremor com-pensation for robotics assisted microsurgery. In: Pro-ceedings of the annual international conference of theIEEE Engineering in Medicine and Biology Society;1992. p 1067–1068.

4. Charles S, Das H, Ohm T, Boswell C, Rodriguez G,Steele R, Istrate D. Dexterity-enhanced teleroboticmicrosurgery. In: Proceedings of the Eighth Interna-tional Conference on Advanced Robotics, Monterey,California; 1997. p 5–10.

5. Charles S, Williams RE, Hamel W. Design of a sur-geon-machine interface for teleoperated micro-sur-gery. In: Proceedings of the 1989 IEEE Engineering inMedicine and Biology Society annual conference, Se-attle, WA; 1989. p 883–884.

6. Dario P, Carroza MC, Leniconi L, Magnani B,D’Attanasio S. A micro robotic system for colonos-copy. In: Proceedings of the 1997 International Con-ference on Robotics and Automation, Albuquerque,New Mexico; 1997.

7. Das H, Zak H, Kim WS, Bejczy AK, Schenker PS.Performance experiments with alternative manualcontrol modes in teleoperation. PRESENCE: Teleop-erators Virtual Environments 1992;1:201–217.

8. Hannaford B, Marbot PH, Moreyra M, Venema S. A5-axis mini direct drive robot for time delayed teleop-eration. In: Proceedings of the Intelligent Robots andSystems Conference (IROS ’94), Vol. 1; 1994. p 555–562.

9. Hunter IW, Doukoglou TD, Lafontaine SR, CharettePG, Jones LA, Sagar MA, Mallinson GD, Hunter PJ.A teleoperated microsurgical robot and associated vir-tual environment for eye surgery. PRESENCE: Tele-operators Virtual Environments 1993;2:265–280.

10. Kavoussi LR, Moore RG, Adams JB, Partin AW.Urologists at work: Comparison of robotic versus hu-man laparoscopic camera control. J Urol 1995;154:2134–2136.

11. Matthews PCB, Muir RB. A comparison of electro-myogram spectra with force spectra during humanelbow tremor. J Physiol 1980;302:427–441.

12. Salcudean SE, Yan J. Towards a force-reflecting mo-tion-scaling system for microsurgery. In: Proceedingsof the IEEE International Conference on Robotics andAutomation, Vol. 3; 1994. p 2296–2301.

13. Schenker PS, Charles S, Das H, Ohm T. Developmentof a telemanipular for dexterity enhanced microsur-gery. In: Proceedings of the Second Annual Interna-tional Symposium on Medical Robotics and ComputerAssisted Surgery, Baltimore, Maryland; 1995. p 81–88.

14. Schenker PS, Das H, Ohm T. A new robot for highdexterity microsurgery. In: Ayache N (ed): ComputerVision, Virtual Reality and Robotics in Medicine:Lecture Notes in Computer Science. Berlin: Springer-Verlag; 1995. p 115–122.

15. Siemionow M. Robot assisted microsurgery: Experi-ence with the RAMS system in plastic surgery re-search. In: Proceedings of the Minimally InvasiveCoronary Artery Bypass Grafting Workshop, Utrecht,The Netherlands; 1997.

16. Tanikawa T, Arai T, Masuda T. Development of amicro manipulation system with two-finger microhand. In: Proceedings of the Intelligent Robots andSystems Conference (IROS ’96), Vol. 2; 1996. p 981–987.

17. Tendick F, Bhoyrul S, Way LW. Comparison of lapa-roscopic imaging systems and conditions using a knottying task. In: Proceedings of the Second AnnualInternational Symposium on Medical Robotics andComputer Assisted Surgery, Baltimore, Maryland;1995. p 238–245.

18. Tendick F, Stark L. Analysis of the surgeon’s graspfor telerobotic surgical manipulation. In: Proceedingsof the 1989 IEEE Engineering in Medicine and Biol-ogy Society annual conference, Vol. 3; 1989. p 914–915.


evaluation of a telerobotic system to assist surgeons in microsurgery

Documents