quantitative performance assessment of surgical robot...

Quantitative Performance Assessment of Surgical Robot Systems:

TeleRobotic FLS

Mitchell J.H. Lum

A dissertation submitted in partial fulfillment ofthe requirements for the degree of

Doctor of Philosophy

University of Washington

2008

Program Authorized to Offer Degree:Electrical Engineering

University of WashingtonGraduate School

This is to certify that I have examined this copy of a doctoral dissertation by

Mitchell J.H. Lum

and have found that it is complete and satisfactory in all respects,and that any and all revisions required by the final

examining committee have been made.

Co-Chairs of the Supervisory Committee:

Blake Hannaford

Jacob Rosen

Reading Committee:

Blake Hannaford

Jacob Rosen

Thomas S. Lendvay

Andrew S. Wright

Date:

In presenting this dissertation in partial fulfillment of the requirements for the doctoraldegree at the University of Washington, I agree that the Library shall make its copiesfreely available for inspection. I further agree that extensive copying of this dissertation isallowable only for scholarly purposes, consistent with “fair use” as prescribed in the U.S.Copyright Law. Requests for copying or reproduction of this dissertation may be referredto Proquest Information and Learning, 300 North Zeeb Road, Ann Arbor, MI 48106-1346,1-800-521-0600, to whom the author has granted “the right to reproduce and sell (a) copiesof the manuscript in microform and/or (b) printed copies of the manuscript made frommicroform.”

Signature

Date

University of Washington

Abstract

Quantitative Performance Assessment of Surgical Robot Systems: TeleRobotic FLS

Mitchell J.H. Lum

Co-Chairs of the Supervisory Committee:Professor Blake Hannaford

Electrical Engineering

Professor Jacob RosenElectrical Engineering

Robot assisted surgery is revolutionizing the way in which many surgical interventions are

performed. Increased dexterity allows surgeons to perform procedures that were otherwise

impossible. TeleSurgery will allow expert care to be distributed anywhere in the world that

medical intervention is required. The development of the University of Washington, RAVEN

Surgical Robot, was a collaborative effort between surgeons and engineers from multiple

disciplines. Initial validation testing included, (1) operating through a digital-data-link on

an unmanned aerial vehicle, (2) long-distance teleoperation collaborations with researchers

in other countries, and (3) participation in a NASA training mission during which the

RAVEN was placed in an undersea habitat off the Florida Keys and teleoperated from

Seattle, WA. Based on the current standard in surgical skills evaluation, a method for

objective assessment of surgical robotic systems was developed. The effect of time delay

in telesurgery using the RAVEN and the effect of 3D vision using the ISI da Vinci R© robot

were evaluated. The results of these studies will guide further development of the RAVEN

and stimulate new ideas for telesurgery.

TABLE OF CONTENTS

Page

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Thesis Goal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.3 Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.4 Novel Contributions Of This Thesis . . . . . . . . . . . . . . . . . . . . . . . . 6

Chapter 2: RAVEN Surgical Robot - System Overview . . . . . . . . . . . . . . . 7

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 Clinical Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 Robot Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.4 Software and Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.5 System Validation Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Chapter 3: Society of American Gastointestinal Endoscopic Surgerons, Funda-mentals of Laparoscopic Surgery (SAGES FLS) . . . . . . . . . . . . . 30

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.2 Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.3 Impact/Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Chapter 4: TeleRobotic FLS: Methods . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.2 TeleRobotic FLS Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

i

4.3 Time Delay Experiments with the UW RAVEN Surgical Robot . . . . . . . . 37

Chapter 5: TeleRobotic FLS Pilot Study . . . . . . . . . . . . . . . . . . . . . . . 43

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Chapter 6: Effect of Time Delay on TeleRobotic FLS Block Transfer . . . . . . . . 48

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Chapter 7: Other TeleRobotic FLS Experiments . . . . . . . . . . . . . . . . . . . 63

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

7.2 Intracorpreal Knot Tying - RAVEN . . . . . . . . . . . . . . . . . . . . . . . 63

7.3 Comparison of End Effectors - RAVEN . . . . . . . . . . . . . . . . . . . . . 64

7.4 TeleRobotic FLS - Effect of Sterescopic View, ISI da Vinci R© . . . . . . . . . . 66

7.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

7.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

Appendix A: Plots from the Effect of Time Delay Study with RAVEN . . . . . . . . 77

A.1 Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

A.2 Time Delayed Block Transfer Plots . . . . . . . . . . . . . . . . . . . . . . . . 77

Appendix B: Subject Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

B.2 Feedback from Surgeons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

B.3 Feedback from non-Surgeons . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Appendix C: RAVEN Operation Tips . . . . . . . . . . . . . . . . . . . . . . . . . . 95

C.1 Procedural Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

ii

Appendix D: Suggested Improvements to the RAVEN . . . . . . . . . . . . . . . . . 99

D.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

D.2 Issues and Potential Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

iii

LIST OF FIGURES

Figure Number Page

2.1 The Blue Dragon system. (a) The system integrated into a minimally invasivesurgery operating room. (b) Graphical user interface. . . . . . . . . . . . . . . 8

2.2 Two parallel mechanism aluminum mock-ups. The parallel mechanism hasfour links and would have two actuated joints (the two base joints) if usedfor a surgical robot. It is clear from this picture that the parallel mechanismsuffers from collision problems. The dry-lab experiments underscored theneed for the most compact mechanism possible. . . . . . . . . . . . . . . . . . 9

2.3 (a) Close-up photo of two serial mechanisms in the animal lab set-up. (b)Surgeons manipulating conventional tools inserted through the last axis ofthe mock-ups using the serial configuration. . . . . . . . . . . . . . . . . . . . 10

2.4 The workspace is shown for the parallel mechanism with four equal linklengths of 60◦ as a function of three different base angle α12= 90◦,45◦,0◦.Black represents areas outside the reachable workspace or areas near kine-matic singularity. The circular area in the center of the workspace for the90◦ and 45◦ bases and the stripe for the 0◦ base represent an area of greatestisotropy. Notice that for the 90◦ and 45◦ bases, an area of singularity cutsthrough the reachable workspace; this is a property that is highly undesirable. 11

2.5 CAD rendering of surgical manipulator shown with plastic covers removed.Mass: 12.3kg; Folded Dimensions 61cm x 53cm x 38cm, Extended dimensions:120cm x 30cm x 38cm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.6 Line drawing of tool interface, exploded view. . . . . . . . . . . . . . . . . . . 15

2.7 a) The RAVEN Patient Site and b) the Surgeon Site. . . . . . . . . . . . . . . 15

2.8 Control System State Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.9 RAVEN functional block diagram. The communication layer can take a fewdifferent forms including wireless, wired, or a combination of both. All ofthese configurations have been tested in various experiments which are sum-marized in Table 2.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.10 Surgical robot system deployed in a remote field in Simi Valley, CA. . . . . . 21

2.11 (a) Experimental protocol was performed on a gloved box. (b) Successfulsuture tied on gloved box. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.12 Tool tip trajectory for Task 2 (touch each dot with left hand) while operatingthrough UAV. The x’s represent the location of each dot. . . . . . . . . . . . 23

iv

2.13 Tool tip trajectory for Task 4 (trace the circle with right hand) while oper-ating between Seattle, WA and London, England. . . . . . . . . . . . . . . . . 23

2.14 (a) The surgeon, Dr. Andrew Wright, controls (b) RAVEN and successfullyties a knot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.15 The SAGES FLS Block Transfer task board set up with the RAVEN movinga block from left to right. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.16 Histogram of packets with respect to delay between (a) UW and Aquariusand (b) UW and NURC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.17 Average block transfer completion times of surgeon E1 during local trainingon the RAVEN as well as during the NEEMO mission. Completion timesusing an ISI da Vinci are included for comparison. . . . . . . . . . . . . . . . 28

4.1 FLS Blocks being transferred using RAVEN . . . . . . . . . . . . . . . . . . . 34

4.2 TeleRobotic FLS peg numbering . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.3 Intracorporeal knot tying on a Penrose drain . . . . . . . . . . . . . . . . . . 37

4.4 FLS Pattern Cutting - concentric circles on the gauze pad . . . . . . . . . . . 38

4.5 Teleoperation communication flow . . . . . . . . . . . . . . . . . . . . . . . . 39

4.6 Training Task Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.7 Training Pegs Diagram, order of pegs . . . . . . . . . . . . . . . . . . . . . . . 42

5.1 The SAGES FLS Block Transfer task board set up with the RAVEN movinga block from left to right. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

6.1 Two types of errors are shown, recovered and unrecovered, for each of the 15subjects. The number of errors is the total over three trials of each condition. 52

6.2 Delay Effect versus block transfer time. 0ms average block transfer timewas 36.95sec; 250ms average block transfer time was 53.60sec; 500ms averageblock transfer time was 75.44sec. . . . . . . . . . . . . . . . . . . . . . . . . . 58

6.3 Delay Effect versus path length. 0ms average tool tip path length was 7.629m;250ms average tool tip path length was 9.781m; 500ms average tool tip pathlength was 11.724m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.4 Surgeon Effect versus block transfer time (data included from all three delayconditions). The non-surgeon group had an average block transfer time of55.03sec and the surgeon group had and average block trasnfer time of 55.79seconds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6.5 Surgeon Effect versus path length. . . . . . . . . . . . . . . . . . . . . . . . . 61

6.6 Surgeon Effect + Delay Effect versus block transfer time . . . . . . . . . . . . 61

6.7 Surgeon Effect + Delay Effect versus path length . . . . . . . . . . . . . . . . 62

A.1 Training Task 1A - 0ms delay . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

v

A.2 Training Task 1B - 0ms delay . . . . . . . . . . . . . . . . . . . . . . . . . . . 78





A.7 Training Task 1A - 250ms delay . . . . . . . . . . . . . . . . . . . . . . . . . . 79

A.8 Training Task 1B - 250ms delay . . . . . . . . . . . . . . . . . . . . . . . . . . 80





A.13 Time Delayed Block Transfer Task, Subject 1, non-surgeon . . . . . . . . . . 82





A.18 Time Delayed Block Transfer Task, Subject 6, surgeon . . . . . . . . . . . . . 84



A.21 Time Delayed Block Transfer Task, Subject 9, surgeon . . . . . . . . . . . . . 86

A.22 Time Delayed Block Transfer Task, Subject 10, surgeon . . . . . . . . . . . . 86





vi

LIST OF TABLES

Table Number Page

2.1 Summary of Telesurgery Experiments. . . . . . . . . . . . . . . . . . . . . . . 18

2.2 Summary of experiments, mean network latency and significance of each. . . 26

4.1 Distances between pairs of pegs on the FLS Block Transfer Task Board . . . 35

5.1 Mean completion time for single block transfer . . . . . . . . . . . . . . . . . 45

5.2 Percentage of block transfers that resulted in error (dropped block) . . . . . . 46

5.3 p-values for paired t-tests. Significance level of p < 8.333e−3 required to showsignificant difference. *Note: Subject 1 comparisons substitute Treatment Dfor C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

6.1 Number of repetitions for each of the training tasks . . . . . . . . . . . . . . . 49

6.2 Three different conditions were presented to each subject . . . . . . . . . . . 50

6.3 Treatment order for each of the six permutations . . . . . . . . . . . . . . . . 51

6.4 Subject by Subject Mean Block Transfer Times reported in seconds for eachof the three conditions. The mean times for each subject were fitted to alinear regression. The Delay Sensitivity is the slope of the linear fit. The R2

value associated with the linear regression is also listed. . . . . . . . . . . . . 53

6.5 Errors for each of the three treatments over three trials for each and the totalerrors over all nine trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

6.6 Subject by Subject Mean Block Transfer Path Length reported in meters foreach of the three conditions. The mean path length for each subject werefitted to a linear regression. The Delay Sensitivity is the slope of the linearfit in millimeters of increased path length per millisecond of increased delay.The R2 value associated with the linear regression is also listed. . . . . . . . . 56

6.7 ANOVA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

7.1 Intracorporeal Knot Tying Task, RAVEN Surgical Robot . . . . . . . . . . . 65

7.2 Block Transfer Task, ISI da Vinci R© . . . . . . . . . . . . . . . . . . . . . . . . 69

7.3 Intracorporeal Knot Tying Task, ISI da Vinci R© . . . . . . . . . . . . . . . . . 70

7.4 Pattern Cutting, ISI da Vinci R© . . . . . . . . . . . . . . . . . . . . . . . . . . 71

7.5 MISTELS task cutoff times . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

A.1 Order of the nine trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

vii

GLOSSARY

AVULSED: The forcible tearing away of a body part by trauma or surgery.

FLS: Fundamentals of Laparoscopic Surgery. The SAGES FLS Program has a mission to

provide surgical residents, fellows and practicing surgeons an opportunity to learn the

fundamentals of laparoscopic surgery in a consistent, scientifically accepted format;

and to test cognitive, surgical decision-making, and technical skills, all with the goal

of improving the quality of patient care.

PENROSE DRAIN: A Penrose drain is a hollow rubber tube that removes fluid from a

wound area.

PSEUDORANDOM: A pseudorandom process is a process that appears random but is

not. Pseudorandom sequences typically exhibit statistical randomness while being

generated by an entirely deterministic causal process. Such a process is easier to

produce than a genuine random one, and has the benefit that it can be used again

and again to produce exactly the same numbers, useful for testing and fixing software.

SAGES: The Society of American Gastrointestinal and Endoscopic Surgeons

SCARA MANIPULATOR: The SCARA acronym stands for Selective Compliant Assembly

Robot Arm or Selective Compliant Articulated Robot Arm. It has three parallel

revolute joints, allowing it to move and orient in a plane, with a fourth prismatic joint

for moving the end-effector normal to the plane.

TROCAR: A trocar is a hollow cylinder with a sharply pointed end, often three-sided,

that is used to introduce cannulas and other similar implements into blood vessels or

viii

body cavities. Trocars are also used as ports in laparoscopic surgery.

ix

ACKNOWLEDGMENTS

The author wishes to express sincere appreciation to those who have contributed to

my ability to get here. First and foremost, my friends and family for all of the support.

My committee members, Blake Hannaford, Jacob Rosen, Mika Sinanan, Thomas

Lendvay, Andrew Wright and Mehran Mesbahi. Karen Fisher of the BioRobotics

Lab. Helene Obradovich, Stephen Graham, Amy Feldman-Bawarshi, Frankye Jones,

Angel Bailey, Johnny Young, Laura Haas, Tom Jones, Helena Dworakowski and

Davida Clyde from the EE office. Joshua R. Smith, Anthony LaMarca, Jean Moran

and Ryan Wistort from Intel Research Seattle. Jed Kaufman from Swedish Hospital,

Lily Chang, John Corman and Marco Salizar from Virginia Mason Medical Center.

Lisa Jukelevics the SAGES FLS Project Manager. Prof. Paul Sampson from the

Department of Statistics.

x

1

Chapter 1

INTRODUCTION

In the late 1980’s and early 1990’s, minimally invasive techniques revolutionized the

many surgical procedures were performed. Minimally invasive surgery allows the surgeon to

make a few small incisions in the patient, rather than making one large incision for access.

This technique allows for significantly faster recovery times, less trauma, and decreased pain

medication requirements for the patient.

With the advent of surgeon controlled robotic systems for minimally invasive surgery

(MIS) since the mid-1990’s, the hope for telesurgery has come to fruition [20]. Through a

large group effort within the University of Washington BioRobotics Lab and their collabo-

rators, a new surgical robot system, the RAVEN, has been developed.

The BioRobotics Lab has been a leader in quantitative assessment of surgical skill.

These efforts served as a foundation for developing the RAVEN Surgical Robot. The goal

of telesurgical intervention is to bring expert care anywhere in the world and potentially to

provide the best surgical experience for patients who do not live where this expertise exists.

This thesis will focus on quantitative assessment of surgical robotic systems’ performance

in concert with surgeon skills. Time delay is a factor in all global communication systems.

A key question to be studied in these experiments is the effect of delay on telesurgical dry

lab tasks.

1.1 Literature review

1.1.1 Surgical Robotics

There are a few excellent papers that survey the field. Cleary and Nguyen [6] as well as

Davis [7] are particularly relevant. Satava provides an excellent first-hand narrative of his

2

involvement in the development of this field [27].

One of the earliest uses of robotics in surgery was in 1985 when Kwoh used a modified

Puma 560 as a positioning device to orient a needle for biopsy of the brain on a 52 year-old

male [14]. In parallel to Kwoh, Taylor was working at IBM on a bone cutting robot for the

pocket hip-replacement [29]. This device would become ROBODOC.

The late 1980’s also brought on a revolution in surgical intervention. Jacques Perrisat,

MD, from Bordeaux, France presented a video clip at SAGES (Society of American Gas-

trointestinal Endoscopic Surgeons) of the first laparoscopic cholecystectomy (gall bladder

removal). Minimally invasive surgery techniques greatly influenced the approaches that

roboticists have taken toward robot assisted interventions.

At MIT, Madhani, Niemeyer and Salisbury developed an 8-degrees of freedom (dof)

micro-macro based surgical robot system called Black Falcon in 1998. The first 4-dof were

used for gross positioning and the last 4-dof were used for fine dexterous manipulation. The

system used a modified Sensable Phantom as the master device [19]. Around the same time,

at the University of California, Berkeley, Cavusoglu working with Tendick’s group, built “A

Laparoscopic Telesurgery Workstation” [5].

Intuitive Surgical Inc. and Computer Motion Inc. both produced commercially available

FDA-approved surgical robot systems for MIS. Computer Motion’s Zeus surgical robot held

a surgical tool on a SCARA-like manipulator. The Intuitive Surgical Inc. da Vinci R©

uses a parallel 5-bar linkage. In 2003, after years of litigation and counter-litigation over

intellectual property rights, the two companies merged under the name Intuitive Surgical

Inc. (ISI). Currently there are over 400 da Vinci R© systems in use throughout the world.

Their system puts all the power amplifiers for the slave into the master console. In a typical

operating room where the surgeon is sitting a few feet from the patient, this does not pose

any problems. However, when the master and slave are electrically tethered to each other,

telesurgery becomes more difficult.

Telesurgery on a human patient was accomplished on September 9, 2001 by Marescaux

and Gagner. In collaboration with Computer Motion, they used a modified Zeus system

to teleoperate between New York City and Strasbourg, France under a 155ms time delay

using a dedicated Asynchronous Transfer Mode (ATM) communication link [20,21].

3

In Asia, a group from the University of Tokyo has recently been working on a new

telesurgery system [22] [28], and has completed laparoscopic cholcystectomy on a porcine

model between sites in Japan, and moret recently between Japan and Thailand [1] [2].

In [1], they state the experimental result as “a laparoscopic cholecystectomy on a pig was

successfully carried out. The completion time of the surgery was about 90 min, which is

roughly equal to a conventional laparoscopic cholecystectomy.”

Morel’s group from University of Paris, Laboratoire de Robotique de Paris (LRP) uses

a spherical mechanism similar to the RAVEN [30]. Their device is relatively simple, but

what is novel is that it moves the trocar in addition to the tool. This has allowed them to

embed force sensors in the device that give a direct reading of the forces at the tool tip,

instead of the combined interaction forces of the tool/tissue and trocar/abdomen.

Berkelman, at the University of Hawaii, Manoa, has further developed the Light En-

doscopic Robot (LER), on which he began work on while with Taylor’s group at Johns

Hopkins University. This device was designed to guide an endoscopic camera, but is now

capable of holding disposable endoscopic graspers [4] [3]. A tool with wrist articulation is

currently in development.

1.1.2 Quantitative Assessment of Surgical Skill

The old paradigm in surgical education was based around the concept of “See one, do one,

teach one”. This “mentor and student” approach, whereby expert surgeons mentor and train

novice residents, allows those novices to progress to the point where they become the next

generation of experts. The surgical expert would mentor trainees, observing their progress

and determining when they were “good enough”. Certainly an experienced surgeon would

be well-qualified to evaluate whether or not a resident is ready to operate on a patient, and

certainly the experienced surgeon would never allow patient safety to come into jeopardy.

However, this kind of qualitative assessment is easily subject to large variability from surgeon

to surgeon. This means that residency programs across the country will be graduating

surgeons with a wide range of skill that cannot be quantifiably recognized. The old means

of qualitative skill assessment is no longer sufficient. The community medical has demanded

4

a shift toward quantitative measures.

Blue Dragon; BioRobotics Lab/CVES

A quantitative assessment of skill was the goal of Rosen, et. al. [24]. In this research, Jeffrey

D. Brown developed the Blue Dragon, a device used to measure surgical forces, torques and

displacements during in vivo procedures. In conjunction with an Instrumented Endoscopic

Grasper (IEG) [13], the complete system can be likened to a flight data recorder on an

airplane, giving the ability to analyze and recreate all the kinematics and dynamics of the

surgical tools during a minimally invasive procedure. The experimental protocol took data

on 30 surgeons of different skill levels ranging from first year residents (novices) to expert

surgeons, performing a number of tasks during in-vivo procedures on a porcine model. Using

a discrete Markov Model, the surgeons’ skill level can quantitatively be determined with

respect to the model of an expert.

Fundamentals of Laparoscopic Surgery; SAGES

The Society of American Gastrointestinal Endoscopic Surgeons (SAGES) formed a com-

mittee for the Fundamentals of Laparoscopic Surgery (FLS) in the late 1990’s. Peters [23]

discusses the background behind the development of the FLS program. The program fea-

tures both cognitive surgical understanding as well as hands-on technical skills. The tech-

nical skills set is derived from the McGill University MISTELs program [10, 11]. The FLS

program has been implemented around the world with thousands of surgeons from first

year residents to veteran surgeons tested. It provides a well structured and well defined

quantitative means by which a surgeon’s skill can be evaluated.

Red Dragon, BioRobotics Lab/ISIS

Based on the preliminary design of the spherical manipulator in the RAVEN, a new pas-

sive measurement device was created and named the “Red Dragon,” a more compact and

lightweight next generation device than the Blue Dragon [12]. The Red Dragon is currently

being used in much the same way as the Blue Dragon, except that the experimental proto-

5

col uses a subset of the FLS skills tasks performed in a dry-lab environment instead of the

in-vivo porcine experiments.

A major part of the validation of the RAVEN system will include performing FLS skills

tasks. Both the subjects FLS score and kinematic data using the RAVEN in different

operating conditions can be compared with their Red Dragon data.

1.2 Thesis Goal

The goal of my dissertation is to define a methodology by which the performance of a surgi-

cal robot can be quantitatively assessed. The ultimate goal of surgical innovation is better

patient outcomes. In order to innovate, both the clinical and technological capabilities and

limitations must be known. As an engineering student working closely with both talented

engineering faculty and surgical faculty, I gained a strong understanding of minimally in-

vasive surgery and the requirements for a potential new surgical robot manipulator. My

work with the BioRobotics Lab over the last eight years has laid the foundation for my PhD

dissertation.

1.3 Organization of the Thesis

This document begins with an Introduction that includes a literature review and background

on the work that led up to the development of the RAVEN Surgical Robot. In Chapter 2,

the RAVEN Surgical Robot - System Overview provides information and background on the

development of the RAVEN Surgical Robot and some of the preliminary experiments that

led up to the main body of work. Chapter 3, the Society of American Gastrointestinal En-

doscopic Surgeons, Fundamentals of Laparoscopic Surgery (SAGES FLS) describes the FLS

curriculum and tasks and their relevancy in surgical training, education, and certification.

This leads into the main work of Chapter 4, the TeleRobotic FLS: Methods, which describes

in detail the methodology for applying the SAGES FLS skills tasks to Surgical TeleRobotics.

The goal is to create a standardized means by which engineers and researcher in surgical

robotics can test their systems or different aspects of surgical teleoperation. Chapter 5

describes the TeleRobotic FLS Pilot Study that was performed to debug the experimental

methods and collect preliminary data. Chapter 6 then gets to the primary experimental

6

component of the thesis, which is the Effect of Time Delay on TeleRobotic FLS Block Trans-

fer. This study was conducted using a total of 15 subjects, 6 surgeons and 9 non-surgeons,

to investigate the effects of time delay on the one of the FLS tasks. Chapter 7, Other

TeleRobotic FLS Experiments, describes some additional experiments that were run using

the TeleRobotic FLS method, including suturing with the RAVEN, as well as all three tasks

performed with the ISI da Vinci R© to investigate the effects of stereoscopic vision. The the-

sis contains an appendix with Suggested Improvements to the RAVEN that are based on

observations during the experiments and feedback from the subjects who participated in

the TeleRobotic FLS studies.

1.4 Novel Contributions Of This Thesis

This thesis presents a method for integrating three of the SAGES FLS skills tasks for

telerobotic applications. This method has been applied to investigate the effect of delay on

surgeon and non-surgeon performance of the Block Transfer task using the RAVEN. It has

also been used to compare the effect of different end-effectors on non-surgeon performance

of the Block Transfer task in the RAVEN. The method has been used for investigating the

effect of 3D on surgeon performance of all three tasks using the ISI da Vinci R©.

TeleRobotic FLS can be applied to track improvements in the RAVEN or test new

hypotheses in surgical telerobotics. The hope is that other groups working in this field will

also adopt the TeleRobotic FLS as a standard method of testing that will give engineers

and surgeons a set of results that are meaningful to both groups.

7

Chapter 2

RAVEN SURGICAL ROBOT - SYSTEM OVERVIEW

2.1 Introduction

This chapter will discuss the design, development, and accomplishments of the RAVEN

Surgical Robot.

2.2 Clinical Requirements

For over a decade, a strong collaboration between engineers in the BioRobotics Lab and

surgeons in the Center for Video Endoscopic Surgery has focused on answering clinically

relevant problems. Surgical training follows the mentor/student model in which the expert

surgeon shows a novice how to perform a task and the novice then mimics the expert. The

evaluation of surgical skill has historically been a subjective process.

In order to move toward more objective measures, extensive work has been performed

in the area of surgical measurement and skill assessment [25]. The Blue Dragon, a passive

device instrumented with sensors, was developed for measuring surgical tool displacements,

forces, and torques during in-vivo animal surgeries (Figure 2.1). Using the Blue Dragon,

an extensive database was created of in-vivo tissue handling/examination, dissection, and

suturing tasks performed by 30 surgeons. Analysis of this data indicated that, 95% of

the time, the surgical tools were located within a conical range of motion with a vertex

angle 60◦ (termed the dexterous workspace, DWS). A measurement taken on a human

patient showed that, in order to reach the full extent of the abdomen, the tool needed to

move 90◦ in the mediolateral (left to right) and 60◦ in the superior/inferior direction (head

to foot). The extended dexterous workspace (EDWS) was defined as a conical range of

motion with a vertex angle of 90◦ and is the workspace required to reach the full extent

of the human abdomen without reorientation of the base of the robot. These parameters,

obtained through surgical measurement, served as a basis for the kinematic optimization of

8

(a) (b)

Figure 2.1: The Blue Dragon system. (a) The system integrated into a minimally invasivesurgery operating room. (b) Graphical user interface.

the RAVEN spherical mechanism [15] [18].

2.3 Robot Design

The RAVEN Surgical Robot consists of three main pieces: the patient site, the surgeon site,

and a network connecting the two. Using the typical teleoperator system nomenclature, the

surgeon site is the ‘master’ and the patient site is the ‘slave’. The patient site consists of

two surgical manipulators that are positioned over the patient. The surgeon site consists of

two control devices and a video feed from the operative site. The communication layer can

be any TCP/IP network, including a local private network, the Internet, or even a wireless

network.

2.3.1 The Patient Site

Much of the engineering effort was focused on developing the patient site. Starting with the

range of motion required for surgery, the spherical mechanism was analyzed and optimized

for our application [18]. Once the optimal geometry of the mechanism was determined, a

detailed design of the arms and tool interface was performed.

9

Design Approach

The pivot point constraint in MIS makes the spherical manipulator a strong candidate for

a surgical robot. An adjustable passive aluminum mock-up was fabricated to model the

kinematics of the spherical manipulator, in parallel and serial configurations. The mock-

up was designed such that a standard MIS tool with 5mm shaft could pass through the

distal joint. In a dry-lab set-up, a number of kinematic configurations were compared on

a training torso (Simulab, Seattle, WA) to assess range of motion and collision problems.

These dry-lab experiments showed that a parallel configuration had a limited workspace with

kinematic singularities contained in the workspace, self-collision problems (where an arm

collided with itself), robot-robot collisions (between two robots within the surgical scene)

and robot-patient collisions (Figure 2.2). Based on some of these practical constraints, it

was determined that the best configuration was two serial manipulators.

Figure 2.2: Two parallel mechanism aluminum mock-ups. The parallel mechanism has fourlinks and would have two actuated joints (the two base joints) if used for a surgical robot.It is clear from this picture that the parallel mechanism suffers from collision problems. Thedry-lab experiments underscored the need for the most compact mechanism possible.

The animal lab experiment applied results from the dry-lab experiment; two serial ma-

nipulators were evaluated with surgeons performing suturing and tissue handling tasks in-

vivo on a porcine model, as shown in Figure 2.3. For this evaluation, the link angles were

10

(a) (b)

Figure 2.3: (a) Close-up photo of two serial mechanisms in the animal lab set-up. (b)Surgeons manipulating conventional tools inserted through the last axis of the mock-upsusing the serial configuration.

set to 75◦ and the surgeons were able to perform all the required tasks, without robot-robot

or robot-patient collisions. The animal lab experiment validated that two serial spherical

manipulators in the surgical scene would be feasible for a surgical robotic system.

A detailed numerical analysis in [15] analyzed both the parallel and serial mechanism

and confirmed the results of the experimental evaluation. A kinematic optimization was

performed to determine the optimal link angles based on the workspace required for surgery.

One striking result is that, for base angles greater than zero (both joint axes collinear), the

parallel mechanism is plagued by an area of kinematic singularity within the center of its

workspace (Figure 2.4).

It was shown both experimentally and analytically that the serial mechanism is better

suited for a surgical manipulator. In this study, optimization criteria consisted of kine-

matic isotropy (the ratio of singular values of the Jacobian matrix) in the numerator and

a link-length penalty in the denominator. The combined criterion rewards good kinematic

performance and penalizes size. With this criterion at its core, the optimization was per-

formed comprehensively over the design space with all combinations of each link ranging

from 30◦-90◦. Within each design candidate, the target workspace was the DWS, the 60◦

11

Figure 2.4: The workspace is shown for the parallel mechanism with four equal link lengthsof 60◦ as a function of three different base angle α12= 90◦,45◦,0◦. Black represents areasoutside the reachable workspace or areas near kinematic singularity. The circular area inthe center of the workspace for the 90◦ and 45◦ bases and the stripe for the 0◦ base representan area of greatest isotropy. Notice that for the 90◦ and 45◦ bases, an area of singularitycuts through the reachable workspace; this is a property that is highly undesirable.

cones. Only the designs that could also reach the EDWS were considered. The optimization

resulted in a design of 75◦ for the first link angle and 60◦ for the second link angle. The

optimized link angles served as the foundation for extensive mechanical design.

Surgical Manipulators

The 7-DOF cable-actuated surgical manipulator, shown in Figure 2.5, is broken into three

main pieces; the static base that holds all the motors, the spherical mechanism that positions

the tool, and the tool interface. The motion axes of the surgical robot are:

1. Shoulder Joint (rotational)

2. Elbow Joint (rotational)

3. Tool Insertion / Retraction (translational)

4. Tool Rotation (rotational)

12

5. Tool Grasping (rotational)

6. Tool Wrist-1 Actuation (rotational)

7. Tool Wrist-2 Actuation (rotational)

The first four joint axes intersect at the surgical port location, creating a spherical

mechanism that allows for tool manipulation similar to manual laparoscopy. The mechanism

links are machined from aluminum and are generally I-section shapes with structural covers.

These removable covers allow access to the cable system while improving the torsional

stiffness of the links. The links are also offset from the joint axis planes, allowing for a

tighter minimum closing angle of the elbow joint.

The RAVEN utilizes DC brushless motors located on the stationary base; these motors

actuate all motion axes. Maxon EC-40 motors with 12:1 planetary gearboxes are used for

the first three axes, which see the highest forces. The first two axes, those under the greatest

gravity load, have power-off brakes to prevent tool motion in the event of a power failure.

The fourth axis uses an EC-40 without a gearbox, and Maxon EC-32 motors are used for

the remaining axes. Maxon DES70/10 series amplifiers drive these brushless motors. The

motors are mounted onto the base via quick-change plates that allow motors to be replaced

without the need to disassemble the cable system.

The cable transmission system for each motion axis is comprised of a capstan on each

motor, a pretension adjustment pulley, various pulleys to redirect the cables through the

links, and a termination point. The shoulder axis is terminated on a single partial pulley.

The elbow axis has a dual-capstan reduction stage terminating on a partial pulley. The tool

insertion / retraction axis has direct terminations of the cables on the tool holder. The tool

rotation, grasping, and wrist cables are terminated on capstans on the tool interface.

The cable system transmission ratios for positioning the tool tip are:

1. Shoulder: 7.7:1 (motor rotations: joint rotations)

2. Elbow: 7.3:1 (motor rotations: joint rotations)

13

3. Insertion: 133:1 (radians: meters)

Each axis is controlled by two cables, one for motion in each direction, and these two

cables are pretensioned against each other. The cables are terminated at each end to

prevent any possibility of slipping. The cable system maintains constant pretension on the

cables through the entire range of motion. Force and motion coupling between the axes is

accommodated in the control system.

Laser pointers attached to the shoulder and elbow joints allow for visual alignment of the

manipulator relative to the surgical port. When the two dots converge at the port location,

the manipulator is positioned such that its center of rotation is aligned with the pivot point

on the abdominal wall.

The power-off brakes can be released by flipping a switch located on the base. The brakes

are normally powered by the control electronics, but also have a battery plug-in for easy

set-up and break-down when the system is not powered. ABS plastic covers were created

on a 3D printer to encapsulate the motor pack, thereby protecting actuators, encoders and

electrical wiring. Figure 2.7(a) shows the patient site without electrical cables.

The tool interface, shown in Figure 2.6, controls the tool rotation, grasp, and wrist axes,

and allows for quick changing of tools. The coupler is designed for one-handed engage-

ment/disengagement of the surgical tool to the manipulator. The tools used are Micro-Joint

tools from the Zeus surgical robot that have been adapted for use on the RAVEN. The tools’

grasp and wrist axes are actuated by pushrods in the tool shaft. High-pitch acme threads

in the tool interface convert the rotational motion of the cable system capstans into linear

motion of the tool pushrods. Because the modified Zeus tools only feature one wrist axis,

the surgical robot currently utilizes only one of its two wrist axes.

2.3.2 The Surgeon Site

The surgeon site was developed to be low cost and portable, a choice that allows for easier

telesurgical collaboration. It consists of two PHANToM Omni devices (SensAble Tech-

nologies, Woburn, MA), a USB foot-pedal, a laptop running the surgeon’s graphical user

interface software, and a video feed of the operative site as shown in Figure 2.7(b). Sens-

14

Figure 2.5: CAD rendering of surgical manipulator shown with plastic covers removed.Mass: 12.3kg; Folded Dimensions 61cm x 53cm x 38cm, Extended dimensions: 120cm x30cm x 38cm.

Able’s PHANToM haptic devices are well established among haptics researchers and have

a development environment that is straightforward to use. The Omni is a cost effective so-

lution that allowed us to quickly implement a surgeon interface device for our master/slave

system. It features 3-DOF force-feedback, 6-DOF sensing, and two momentary switches

on the stylus. We are not currently utilizing the force-feedback capability of the Omni.

The foot-pedal enables and disables the coupling between the patient site and surgeon site,

allowing for position indexing.

15

Figure 2.6: Line drawing of tool interface, exploded view.

(a) (b)

Figure 2.7: a) The RAVEN Patient Site and b) the Surgeon Site.

2.4 Software and Control

2.4.1 Patient Site

Control software is running in the kernel space of an RTAI Linux computer at a rate of

1kHz. The link between the control software and the motor controllers is a USB 2.0 interface

board. The USB board features eight channels of high-resolution 16-bit D/A for control

signal output to each controller and eight 24-bit quadrature encoder readers.

16

Software and Safety Architecture

The control system and surrounding electronic hardware were designed to incorporate safety,

modular design, and flexibility. Because this is a medical device, the most critical of these

aspects is safety. Inherent to a safe system is robustness, reliability, and some level of

automatic override. To achieve reliability we defined four software states in which our

system can operate: Initialization, Pedal Up, Pedal Down, and Emergency Stop (Figure

2.8). At power-up, the manipulators should be resting against hard stops. The initialization

state takes each manipulator from its resting position and moves it into the surgical field.

Once the initialization is complete, the system automatically transitions into the Pedal Up

state. In the Pedal Up state, the robot is not moving and brakes are engaged. The system

enters Pedal Up when the surgeon lifts his/her foot from the foot-pedal, decoupling the

master from the surgical manipulator. This is done to perform tool indexing or to free the

surgeon’s hands for peripheral tasks. The Pedal Down state is initiated when the surgeon

pushes the foot-pedal down, releasing the brakes and allowing the master device to directly

control the surgical manipulator.

E-STOP Init Start (automatic)

Pedal Pressed

Pedal Released

Pedal Down

Pedal Up

Fault Condition

Figure 2.8: Control System State Diagram

A Direct Logic 05 programmable logic controller (PLC) controls motor-enable, brakes,

and the system states based on inputs received from the system. PLCs are a robust technol-

ogy used extensively in automation applications. PLC technology is reliable and provides

built-in, easy-to-use safety circuitry. In addition to monitoring the system hardware, the

PLC monitors the state of the control software through the use of a watch-dog timer. The

watch-dog timer monitors a square-wave signal generated by the control software, output

from the parallel port of the Linux PC. In the event of a software or computer hardware

17

failure, the PLC will detect the loss of the square-wave and immediately put the system

into the Emergency Stop state, enabling the brakes and disabling the motors. An array

of status LEDs displays the current state of the system. The RTAI Linux control software

detects state transitions of the PLC and follows them within 1ms.

Engineers’ Interface

The Engineers’ Interface (EI), a low-level interface to the states and mechanisms of the

control software, assists robot development. Developers are presented with an intuitive

graphical user interface (GUI) with easy access to robot features. In development stages,

the system run level (stop, init, run, e-stop) can be set manually with the click of a button.

Control commands can be sent to any degree of freedom or the entire robot. For example,

a 40◦ sine wave can be output on the shoulder joint, or motor controller number two can

output 30% maximum current, or the end-effector position can be instructed to move 3cm

left. Robot information (such as motor output, joint position, and end-effector position) is

displayed on-screen in real-time and also logged for later evaluation.

The EI can connect to the RTAI Linux control system using either FIFO device nodes

or a single, bi-directional (TCP/IP) network socket. Two types of data are exchanged: a

packet containing all robot-state information is received by the EI, and a command packet

with all instruction parameters is sent from the EI to the control software. This link is

independent of the master-slave link.

2.4.2 Surgeon Site

The Surgeon Site software provides the surgeon with a GUI to log into and connect to

the Patient Site. It allows for unique identification of each user, keeping a detailed log of

when each user logs into the system, connects to the Patient Site, and transitions between

pedal-up and pedal-down states. It provides an automatic means by which each user’s time

on the system can be tracked.

18

2.5 System Validation Experiments

The first teleoperation of the RAVEN took place on October 15, 2005 in a cross-campus

demonstration at the University of Washington (UW) with the surgeon site in a lecture

hall and the patient site in the BioRobotics Lab (BRL). The surgical manipulator’s first

three degrees of freedom - the shoulder, elbow and tool insertion joints - were actuated. A

PHANToM Omni was used to control the endpoint of the surgical tool through the UW’s

campus network with no noticeable delay.

The implementation of a low-cost and portable surgeon site has provided the opportu-

nity for telesurgical collaboration. The telesurgery experiments summarized in Table 2.5

have included many topologies, including within one lab, between labs, and mobile robotic

telesurgery experiences. Figure 2.9 is a functional block diagram of the system, illustrat-

ing the key components of the patient site, the surgeon site, and the communication layer

between them. The RAVEN has been tested in a variety of environments, using multiple

communication layer topologies, and has demonstrated its portability and robustness.

Table 2.1: Summary of Telesurgery Experiments.

Experiment Date(s) Patient Site Surgeon SiteCommunication Layer

VideoNetwork

Architecture

HAPs/MRTJune 5-9, Field, Field, HaiVision

Wireless via UAV2006 Simi Valley, CA Simi Valley, CA Hai560

ICLJuly 20, BioRobotics Lab, Imperial College, iChat or

Commercial Internet2006 Seattle, WA London, England Skype

Animal LabMarch 8, CVES, CVES, Direct

LAN2007 Seattle, WA Seattle, WA S-video

NEEMO

Aquarius

May 8-9,

2007

Aquarius Undersea

Habitat, 3.5 miles off

Florida Keys, 60 ft

depth

University of

Washington, Seattle

HaiVision

Hai1000

Commercial Internet

between Seattle, WA and

Key Largo, FL; microwave

communication link across

10 miles, Key Largo to

Aquarius

NEEMO NURCMay 12-13,

2007

National Undersea

Research Center,

Key Largo, FL

University of

Washington, Seattle

HaiVision

Hai200Commercial Internet

19

Portable Surgical Console (Master)

Surgeon

Paddle Video

Monitor Haptic Master

Input

Position Sensor

USB A/D Converter

Master PC

Software

UDP

Video Decompression

Communication Layer

Option 1 - Wireless (UAV)

Portable Surgical Robot (Slave)

Patient

Video Camera

Surgical Robot

Position Sensor

Servo Control

Slave PC

Software

UDP

Video Compression Motors & Brakes

Safety System

Links & Tools

USB A/D Converter

RF Antenna RF Antenna

Unmanned Aerial Vehicle (UAV)

Pilot

Network Hub

Network Hub Option 3 - Hybrid

Option 2 - Wired

Internet / Intranet

Internet / Intranet RF

Antenna RF

Antenna

Figure 2.9: RAVEN functional block diagram. The communication layer can take a few dif-ferent forms including wireless, wired, or a combination of both. All of these configurationshave been tested in various experiments which are summarized in Table 2.5.

2.5.1 High Altitude Platforms/Mobile Robotic Telesurgery (HAPs/MRT)

Many research systems live out their entire life cycle, from conceptions to decommission, in

a laboratory environment, and are never challenged to move outside of that environment.

An evaluation of the RAVEN’s robustness was its first field deployment on June 5-9 2006.

Dr. Timothy Broderick, MD and Brett Harnett of the University of Cincinnati led the

HAPs/MRT project to evaluate surgical robotics in field conditions. The RAVEN was

taken from the BRL in Seattle, WA and deployed in the desert of Simi Valley, CA for

telesurgery experiments on an inanimate model (see Figure 2.10). The system was powered

by gas generators and was set up under portable tents in an isolated field. Separated by a

distance of 100 meters, the surgeon and patient sites were connected via an aerial digital

datalink onboard AeroVironment’s PUMA unmanned aircraft. The datalink provided by

AeroVironment utilized Internet-style communication at a rate of 1MB per second between

the two sites, allowing the network architecture to remain unmodified. HaiVision Inc.

(Montreal, Canada) provided a hardware codec that used MPEG-2 and transmitted the

video signal at 800kbps.

20

Two surgeons, Dr. Broderick and Dr. Lynn Huffman, from the University of Cincinnati,

performed a set of tasks including touching a series of landmarks and suturing on a latex

glove stretched over an aluminum box (“gloved box”). The gloved box was marked with a

circle and a grid of landmarks spaced 1cm apart left to right and 0.5cm apart toward and

away as shown in Figure 2.11(a). The landmarks were put in a numeric sequence starting

with 1 at the upper left, 2 upper right, moving down through the rows, and finishing at

the lower right. The following five tasks were part of the experimental protocol we entitled

Touch the Dots:

1. Right tool touches each landmark in numeric order.

2. Left tool touches each landmark in numeric order.

3. Touch each landmark in numeric order using alternating left and right tool. Right

tool touches the odd numbered landmarks (left column), left tool touches the even

numbered ones (right column).

4. Right tool traces inner edge of circle in a clockwise direction.

5. Left tool traces inner edge of circle in a clockwise direction.

During three days of field deployment, kinematic data of the surgeons’ commands and

the surgical manipulators’ motions were collected along with network characterization data.

Figure 2.12 shows the tool tip path of Dr. Broderick touching each of the dots with his

left hand. Deploying the system into a field environment and successfully executing the ex-

perimental protocol demonstrated the feasibility of performing Mobile Robotic Telelsurgery

(MRT) through a wireless communication link with limited bandwidth and variable time

delays in an extreme or remote environment.

21

Figure 2.10: Surgical robot system deployed in a remote field in Simi Valley, CA.

2.5.2 Imperial College, London, England (ICL) to University of Washington, Seattle, WA,

USA.

A collaboration with Julian Leung, George Mylonas, Sir Ara Darzi and Ghuang Zhong

Yang from Imperial College (London, England) demonstrated the ability to operate across

a long distance. On July 20, 2006, in the lab in London, the surgeon site was set up with two

PHANToM 6-DOF Premium haptic devices and our surgeon console software. iChat (Apple

Computer Inc) was used for video feedback. The patient site was run from the BioRobotics

Lab in Seattle, WA. Time delay between the patient and surgeon sites was about 140ms for

Internet latency (measured by ping) and about 1 second for video encoding/decoding. This

experiment showed that the master console software was general enough to adapt to other

PHANToM devices, and also demonstrated the system’s ability to teleoperate across long

distances. During this experiment, the remote surgeons performed the same set of tasks on

the gloved box as were performed during HAPs/MRT. Figure 2.13 shows the tool path of

Dr. Leung tracing out the circle.

22

(a) (b)

Figure 2.11: (a) Experimental protocol was performed on a gloved box. (b) Successfulsuture tied on gloved box.

2.5.3 Animal Lab

On March 8, 2007, in collaboration with the University of Washington Center for Video

Endoscopic Surgery (UW CVES), surgeons Mika Sinanan, Andrew Wright, and Thomas

Lendvay performed surgical tasks on a live porcine model (UW-IACUC approval #2469-04,

“Robotic Surgery”). The tasks involved measuring out a specified length of bowel as well

as tying a suture. The patient site was set up in the animal lab, with the surgeon site in an

adjacent office. Video feedback was sent directly through an S-video cable that ran between

the two rooms. Figure 2.14 shows one of the surgeons in one room tying, successfully tying

a suture on a piece of bowel, with the patient in the next room. This experiment was a

step toward proving that the RAVEN could operate on an animal, not just on dry lab task

boards.

2.5.4 NASA Extreme Environment Mission Operations (NEEMO) XII

In the area of surgical robotics, there is no clinically relevant testing standard. As in

HAPs/MRT and the Imperial College collaborations, each set of researchers devises their

own experimental protocol to test their system. The same was true in conventional surgery

23

Figure 2.12: Tool tip trajectory for Task 2 (touch each dot with left hand) while operatingthrough UAV. The x’s represent the location of each dot.

Figure 2.13: Tool tip trajectory for Task 4 (trace the circle with right hand) while operatingbetween Seattle, WA and London, England.

until the late 1990’s, when the Society of American Gastrointestinal and Endoscopic Sur-

geons (SAGES) created a committee to develop curriculum for teaching the Fundamentals

of Laparoscopic Surgery (FLS). The outcome is a curriculum that includes both cognitive

and psychomotor skills. The FLS skills tasks (described in Chapter 3) have been validated

to show significant correlation between score and postgraduate year [8]. These tasks have

been used to quantitatively assess the skill of thousands of surgeons, ranging from novice

to expert, and are considered the “gold standard” in surgical skill assessment.

To move toward a standard for surgical robot evaluation and testing, we have adopted

24

(a) (b)

Figure 2.14: (a) The surgeon, Dr. Andrew Wright, controls (b) RAVEN and successfullyties a knot.

the FLS skills tasks to use in our experiments. The NASA NEEMO XII mission was our

first use of the new task set, with the FLS Block Transfer task, shown in Figure 2.15, chosen

as the primary skills task. The NEEMO missions are training analogs to space flight that

train astronauts and support personnel on how to run missions. These missions take place

in the Florida Keys at the National Undersea Research Center (NURC) in Key Largo, FL

and at the Aquarius Undersea Habitat, 3.5 miles offshore at a depth of 60 feet.

Prior to the mission, two of the NEEMO XII “Aquanauts” and an astrogeologist came

to the lab in Seattle for two days of training on set-up, running, breakdown, storage and

troubleshooting the RAVEN. This training was essential for mission success because, during

the mission, support personnel from the BioRobotics Lab would not be able intervene within

Aquarius, just as in an actual space mission. Once on site in Key Largo, FL, the Aquanauts

as well as habitat technicians received further training before the system was deployed into

the habitat. In order to safely transport the RAVEN through 60 feet of Atlantic Ocean

into the habitat, support divers packed all the components into sealed steel “pots” and

neoprene dry-bags. On May 7, 2007, the RAVEN began its 3-day deployment as part of the

large-scale 12-day training exercise.

25

Figure 2.15: The SAGES FLS Block Transfer task board set up with the RAVEN movinga block from left to right.

During the experiment, the surgeon site was set up in a conference room in Seattle, WA.

The patient site was set up and supported by two surgeons inside Aquarius. Communication

between the patient and surgeon sites traveled between UW and NURC via commercial

Internet, then from NURC across a wireless microwave communication link to the Life

Support Buoy, and down a hardwired umbilical cable into Aquarius.

In order to gather network performance characteristics, a UDP packet reflector program

was placed at the servers at NURC and Aquarius in Florida. The UDP packet reflector

program receives the UDP data packets and routes them to back to the sender, in this case,

back to the workstation at the UW. A data structure similar to the UDP packets used in

the telesurgery experiments was used for the performance measurements. Each UDP data

packet was time-stamped at the workstation in UW and sent to the servers at NURC and

Aquarius; the reflected packets were used to measure the elapsed round-trip time between

the two locations. The UDP packet sequence number was also used to measure the number

of lost and out-of-sequence packets during the tests.

26

2.6 Results

The RAVEN was conceived from a close collaboration between engineers and surgeons. The

system is a new platform for telesurgery experiments. Table 2.2 summarizes the mean net-

work latency during five different experiments. The total delay experienced by the surgeon

during teleoperation is a function of both network latency as well as video compression and

decompression times. Depending on the video codec used, video latency can vary dramat-

ically and is difficult to measure accurately. During these telesurgery experiments, data

were collected to characterize the network conditions. Figure 2.16 shows a histogram of the

network conditions during NEEMO.

Table 2.2: Summary of experiments, mean network latency and significance of each.

Experiment Mean Network

Latency (ms)

Significance

HAPs/MRT 16 Operated in a field environment to test ruggedness

and portability. Communicated via wireless through

a UAV.

ICL 172 Adaptability of surgeon site to other Sensable devices.

Teleoperation over long distance.

Animal Lab 1 Demonstrated ability to operate on an animal

through MIS ports.

NEEMO

Aquarius

76 Telerobotic FLS for performance measurement. Op-

erating in a unique environment. Communicating

across both commercial Internet and long distance

wireless.

NEEMO

NURC

75 Additional opportunity to collect Telerobotic FLS

data over long communication network.

The SAGES FLS skills tasks are well defined and the kit readily available for purchase.

Developing a “Telerobotic FLS” protocol will give consistency to telesurgical experiments.

27

(a) (b)

Figure 2.16: Histogram of packets with respect to delay between (a) UW and Aquarius and(b) UW and NURC.

Figure 2.17 summarizes the mean completion time for expert surgeon E1 performing the

Block Transfer task. In each of the first three weeks of training, E1 performed three rep-

etitions of the Block Transfer in the lab environment with effectively no delay. There is

a learning effect, as E1’s mean time improved from week to week. During the NEEMO

mission, there was limited time, so E1 was only able to complete a single repetition with

the RAVEN in Aquarius and another single repetition with it on-shore in Key Largo. While

these results do not show statistical significance, one can observe a learning effect which is

probably due to accommodating for telesurgery latency. For comparison, the same surgeon,

who uses a da Vinci, clinically was able to complete the block transfer task in about one

minute using the da Vinci, taking only slightly longer with the stereo capability disabled.

The da Vinci results are also included in Figure 2.17.

2.7 Conclusions

Starting with an extensive database of in-vivo minimally invasive surgical measurements,

the requirements for tissue manipulation and tool handling were defined. Using a clinically

relevant design specification, a kinematic optimization was performed on a spherical mecha-

28

Figure 2.17: Average block transfer completion times of surgeon E1 during local trainingon the RAVEN as well as during the NEEMO mission. Completion times using an ISI daVinci are included for comparison.

nism to obtain the ideal link lengths for the surgical manipulator. The mechanical design of

the manipulators minimizes inertia through careful design of the link structure and place-

ment of all the actuators on a stationary base. RTAI-based control software was developed

in conjunction with a USB-interface board allowing for high performance real-time control

of the system. Integrating commercially available haptic devices into the surgeon console

provided an inexpensive solution to surgeon site control of the surgical manipulators and

enabled many collaboration opportunities (summarized in Table 2.2)

Acknowledgments

Development of the RAVEN was supported by the US Army, Medical Research and Ma-

teriel Command, grant number DAMD17-1-0202. The HAPs/MRT project was supported

29

by the US Army, Medical Research and Materiel Command grant number W81XWH-05-

2-0080. The authors would like to thank our HAPs/MRT collaborators at the University

of Cincinnati, AeroVironment, and HaiVision, as well as our collaborators in London at

Imperial College. The NEEMO XII participation has been supported by the US Army

TATRC grant number W81XWH-07-2-0039. The authors would like to thank our NEEMO

collaborators from University of North Carolina at Wilmington, US Navy, National Un-

dersea Research Center, National Oceanographic and Atmospheric Administration, NASA,

and the University of Cincinnati.

30

Chapter 3

SOCIETY OF AMERICAN GASTOINTESTINAL ENDOSCOPICSURGERONS, FUNDAMENTALS OF LAPAROSCOPIC SURGERY

(SAGES FLS)

3.1 Introduction

SAGES created a committee in the late 1990’s to develop curriculum for teaching the Fun-

damentals of Laparoscopic Surgery (FLS) [23]. The outcome has been a curriculum that

includes both cognitive as well as psychomotor skills. The FLS tasks have been used to

quantitatively assess the skill of thousands of surgeons, ranging from novice to expert, and

are considered the “gold standard” in surgical skill assessment.

3.2 Tasks

The FLS skills assessment consists of five tasks. Two of these tasks, placement of ligating

loop and extracorporeal knot tying, require specialized tools. Placement of a ligating loop

requires a ligating loop and extracorporeal knot tying requires a knot pusher. In order to

make for a more generic set of tasks that can be performed by a wide variety of surgical

robots, these two tasks were eliminated from the TeleRobotic FLS list of tasks. The three

SAGES FLS tasks were chosen for TeleRobotic FLS are:

• Pegboard Transfer (Block Transfer) 1

• Pattern Cutting

• Intracorporeal knot tying

1The Pegboard Transfer is often referred to in short as “Peg Transfer,” which has lead to come confusion,since blocks are being transferred not pegs. Some surgeons refer to the task as “Block Transfer” tomake the name more descriptive of the task performed. For TeleRobotic FLS we refer Block Transfer

synonymously with the SAGES Pegboard Transfer.

31

These three tasks represent all the necessary movements, skills, and cognitive tasks

found in a typical laparoscopic procedure. Further, Derossis [9] reported that these three

tasks shown a significant correlation (P<0.05) between PGY (postgraduate year) and score.

The tools required are graspers for the Block Transfer and Intracorporeal Knot Tying and a

grasper and scissors for the textitPattern Cutting The following descriptions of the tasks are

from [11]. In order to avoid skewing the description of the tasks they are quoted verbatim.

For the knot tying task, only the intracorproal knot will be performed.

Block Transfer / Pegboard Transfer

A series of 6 plastic rings are picked up in turn by a grasping forceps from a

pegboard on the surgeon’s left, transferred in space to a grasper in the right

hand, then placed around a post on the corresponding right-sided pegboard.

After all rings are transferred from the left to right, the process is reversed,

requiring transfer from the right to left hand. This task was designed to develop

depth perception and visual-spatial perception in a monocular viewing system

and the coordinated use of both the dominant and nondominant hands. It also

replicates the important action of transferring and positioning a needle between

needle holders when suturing. This exercise is timed, and a penalty score is

assessed whenever a ring is dropped outside the surgeon’s view.

Pattern Cutting

In this exercise, a 4-x-4-inch gauze is suspended by alligator clips. The surgeon

is required to cut a precise circular pattern from the gauze along a premarked

template. It requires precision and the use of the nondominant hand to pro-

vide appropriate traction to the material and to position the gauze so that the

dominant hand holding endoscopic scissors may cut it accurately. Any devia-

tion of the cut from the premarked template results in a penalty. This exercise

teaches the concept of traction and the need to use the nondominant hand to

provide a convenient working angle for the dominant hand, working through the

32

constraints of fixed trocar positions.

Intracorporeal Knot Tying

It is essential that a surgeon doing laparoscopic surgery be able to place a stitch

and tie a knot. In this exercise, a 0-silk suture with a curved needle is introduced

through the trocar and positioned properly using the needle holders. A stitch

is then placed through target points on either side of a slit Penrose drain, and

the suture is tied using either an intracorporeal (IC, instrument) tie (Task 4),

or an extracorporeal (EC) tying technique with the aid of a knot pusher (Task

5). In these exercises, penalty scores are applied if the needle is not passed

precisely through the target dots and if the stitch is not tied sufficiently tightly

to approximate the edges of the slit and close the defect. Additional penalty is

applied if the knot slips when tested and if the Penrose drain is avulsed from the

underlying block to which it is attached by velcro, indicating that uncontrolled

upward force was applied.

3.3 Impact/Relevance

The FLS Program has become pervasive in surgical skills training and assessment. It has

already been adopted by many surgical residency programs worldwide, and with a $1.8

million grant from Covidien, the FLS Program will be provided to over 250 surgical residency

programs in the United States and Canada, reaching thousands of surgeons-in-training.

[http://www.flsprogram.org/]

The FLS tasks were chosen to demonstrate robotic performance because (1) robotic

surgery employs the same visuo-spatial relationship processing that laparoscopy employs

and (2) FLS is the most reproducible and validated method for demonstrating laparoscopic

proficiency. We chose to select tasks that would be able to be analyzed with a common

language for surgeons and trainees, as wells as researchers and engineers developing solutions

for the surgical community.

33

Chapter 4

TELEROBOTIC FLS: METHODS

4.1 Motivation

The RAVEN Surgical Robot is a system designed for telesurgical applications (Chapter 2).

During its initial validation and field experiments, the RAVEN was tested in a number of

teleoperation modes including operating through a digital data link on board an unmanned

aerial vehicle [17] and in the Aquarius Undersea Habitat [16]. The system has also teleop-

erated with the Patient Site in Seattle, WA and the surgeon site located in Cincinnati, OH;

Tokyo, Japan; Montpellier, France; and London, England connecting through commercial

Internet. Network statistics were collected on packet loss, delay and correlation.

A review of the active groups in surgical robotics research (see Chapter 1) shows that

each group has created their own test of tasks and metrics. Following the HAPs/MRT

(Section 2.5.1) and Imperial College (Section 2.5.2) experiments the shortcomings of the

Touch the Dots tasks were realized. This set of tasks is not standard to any other set

of engineering researchers nor does it provide any surgically relevant results. In order to

converge toward a standard set of tasks that have been shown to be surgically relevant, the

SAGES FLS skills tasks have been adopted. In this chapter, we present a methodology for

incorporating these tasks for telesurgical performance evaluation.

4.2 TeleRobotic FLS Tasks

Of the five FLS skills tasks, three are directly applicable to use with surgical robotic systems.

This section describes each of three tasks, how they are performed, and the way that the

results are reported.

34

4.2.1 TeleRobotic Block Transfer

In the SAGES FLS Block Transfer, the subject is allowed to start with all blocks on either

side, then transfer all blocks from one side to the other and then back again. It is only

considered an error if a block is dropped outside the viewable area, bounded by a black

rectangle on the task board (see Figure 4.1. The score is a proprietary formula based on an

aggregate of completion time with a penalty for errors.

Figure 4.1: FLS Blocks being transferred using RAVEN

The TeleRobotic Block Transfer, by contrast, is more structured. A single trial consists

of moving all six blocks from the left side to the right side and then back again. There is

a pause between left-to-right and right-to-left. Each peg is numbered as shown in Figure

4.2. The trial begins with the tools touching or near the first block. The blocks must be

moved in numeric order from the peg on one side to the corresponding peg on the other

side. Should the subject attempt to move blocks out of order or to the wrong peg, the

experimenter will remind the subject that order matters.

Errors are defined as any time a block is not set down on its corresponding numbered

peg. This often occurs in the form of a dropped block. If a block is dropped and recovered,

that is marked as one type of error. If a block is dropped and not recovered that is noted as

35

Peg Number Distance (cm)

1 3.8

2 5.7

3 5.7

4 5.2

5 5.2

6 6.9

Table 4.1: Distances between pairs of pegs on the FLS Block Transfer Task Board

another type of error. For system development purposes the experimenter should not the

cause of the error if that can be immediately determined.

When the experimenter says “Go!,” the time starts. The individual “lap time” for each

block is recorded and measured at the point where the block is fully in contact with the

surface of the peg board (i.e. not partially dangling on the peg or partially sitting on an

adjacent block). The time ends when the sixth block is placed. From the “lap times,” the

“split times,” or individual block transfer times, may be determined.

The time-based results can either be reported as the average individual block transfer

time or the average time for all six blocks, as well as the number of blocks dropped and

recovered and the number of blocks dropped and not recovered. Optimally the surgeon

performs the task carefully enough that no errors occur.

4.2.2 TeleRobotic Intracorporeal Knot Tying

The intracorporeal knot tying task board is a foam-covered board to which a Penrose drain

is velcroed. The Penrose drain has two dots on it; these are targets for the needle to enter

and exit, as shown in Figure 4.3. The task requires the surgeon to suture the drain and

tie three knots. The data reported are (1) completion time, (2) distance of the suture from

the entrance target, (3) distance of the suture from the exit target, (4) presence of a gap in

the drain and width of the gap, and (5) if the knot is secure. Time is measured in seconds,

36

4

6

1

2

3

5

6

1

4

5

2

3

Left Right

UW BioRobotics Lab TeleRobotic FLS

Figure 4.2: TeleRobotic FLS peg numbering

distances in millimeters and knot security as “yes” or “no”.

4.2.3 TeleRobotic Pattern Cutting

In the pattern cutting task, a gauze pad with two concentric circles is clipped and secured

to the FLS task board, as shown in Figure 4.4. The set-up requires a scissors end-effector

for the dominant hand and a grasper for the non-dominant hand. The subject is required to

grasp the gauze with one end-effector to stabilize it while cutting the circle shape with the

other end effector. The task measures the subjects ability to orient the material grasped

optimally for cutting (dissecting), a task frequently used in surgery. An error is defined

as cutting outside the outer black ring or inside the inner black ring. Once the subject

37

Figure 4.3: Intracorporeal knot tying on a Penrose drain

has completed the task, the circular piece of gauze can be inspected for errors. Any white

outside the outer black ring or white inside the inner black ring signifies an error. The

number of errors and the completion time in seconds are reported.

4.3 Time Delay Experiments with the UW RAVEN Surgical Robot

This section describes a specific implementation of the TeleRobotic FLS methodology to

study the effect of time delay on performance using the UW RAVEN Surgical Robot. This

experimental set-up was used for both the pilot study described in Chapter 5 and the main

study described in Chapter 6.

4.3.1 Teleoperation

Real Teleoperation: In an actual teleoperation, physical distance and a real network separate

the patient and surgeon sites with time-varying delays. When a surgeon makes a gesture

using the master device, motion information is sent through the network to the Patient

38

Figure 4.4: FLS Pattern Cutting - concentric circles on the gauze pad

Site with a network time delay (Tn). The manipulator moves and the audio/video (a/v)

device observes the motion. Digital a/v is compressed (Tc), sent from the Patient Site to

the Surgeon Site through the network (Tn), then decompressed (Td) and observed by the

surgeon. The surgeon has experienced a total delay T = 2Tn + Tc + Td, from the time (s)he

made the gesture to the time that action was observed.

Emulated Teleoperation: In the emulated teleoperation, the Surgeon and Patient sites are

not separated by physical distance. Instead they are connected through a Linux PC with two

network cards running NISTNET that emulates a real network. This emulator allows the

experimenter to adjust the average packet delay between the Surgeon and Patient sites [26].

The a/v feed is connected directly from the camera at the Patient Site to the monitor at

the Surgeon Site through S-video eliminating any delay due to compression/decomression.

The surgeon experiences a total delay, Te due to the emulator, from the time (s)he made

the gesture to the time that action was observed.

The flow of information is illustrated in Figure 4.5. By setting Te = 2Tn + Tc + Td,

39

Patient Site Surgeon Site

Network (T n )

A/V Compression

(T c )

Network (Tn)

A/V Decomression

(T d )

Patient Site Surgeon Site

Emulator (T e )

Direct A/V (T=0)

T e = 2*T n + T c + T d

Figure 4.5: Teleoperation communication flow

one can emulate any real teleoperation condition. In this study, because the camera is

connected directly to the monitor, there is no degradation of the video or audio signals due

to compression techniques. Performance in telesurgery as a function of video degradation

could be the subject of a future study, but is not a factor in this case.

4.3.2 Experimental Methods

Experimental Set-up

The RAVEN Patient and Surgeon sites are located in the same room and are connected

through the network emulator. The video feed comes directly from a Sony DCR-VX1000

40

3-chip digital video camera to a Sony Trinitron PVM-14M2MPU color monitor through an

S-video cable.

Training

Each subject received specific training on the system prior to the main study. Each subject

watched an orientation video about the RAVEN surgical robot and how to perform tele-

manipulation tasks. The video broke down manipulation into three parts: (1) positioning,

(2) orienting, and (3) grasping, first with dominant hand, then with non-dominant hand.

The subjects were instructed on three tasks (described below) that would enable them to

successfully teleoperate using the RAVEN. During the pilot study (Chapter 5), each task

was performed until the subject’s completion time for that task did not improve over three

trials. Based on subject feedback as well as analysis of the training data, a fixed number of

repetitions for each of the tasks were performed for the main study (Chapter 6). Once the

subject had completed a tas,k (s)he was allowed to move on to the next task. The subjects

trained until they had completed all three tasks. The subjects then repeated the same train-

ing tasks under a delay condition of 250ms. By first training the subjects with no-delay,

they were able to learn the psychomotor skills necessary to telemanipulate objects under the

most ideal conditions. Then, by repeating the training tasks under a delay condition, they

learned to accommodate for delay. Ideally, in order to reduce subject fatigue, the non-delay

and delay training were completed on separate days. In practice, due to scheduling this was

not always possible.

The training task board was built on a 4” x 2.5” piece of plexiglass. Six 1” x 1/4”-20

countersink screws were arranged in a grid of two rows of three. The screws were capped

by 1” pieces of 1/4” inner-diameter rubber tubing and arranged with 1” spacing between

each of the three columns and 7/8” spacing down between each of the two rows. Each of

the six pegs were numbered 1-6 (see Figure 4.6). The following list describes the training

tasks for the dominant hand. Tasks 1B, 2B and 3B, the tasks for the non-dominant hand,

are similar.

• Task 1A Dominant Hand Positioning Using the dominant hand’s tool, touch each peg

41

in sequence 1 through 6, while keeping the non-dominant hand’s tool in the field of

view. You will know you’ve touched the peg when you see it deflect.

• Task 2A Dominant Hand Orientation Using the dominant hand’s tool, orient the

grasper tips and place the tips into the center of each peg in sequence 1 through 6

while keeping the non-dominant hand’s tool in the field of view.

• Task 3A Dominant Hand Grasping Using the dominant hand’s tool, open the grasper

tips, place the tips with one jaw in the center of each peg and one jaw on the outside

of the peg, then grasp the peg wall. Grasp each peg in sequence 1 through 6 while

keeping the non-dominant hand’s tool in the field of view. When grasping with the

right tool, grasp the right side of the peg. When grasping with the left tool, grasp the

left side of the peg.

Figure 4.6: Training Task Board

Experimental Tasks

Once the subject had completed training the progress to the main experimental tasks. The

tasks and experimental design are described in detail in their respective chapters.

42

4

1 2 3

5 6

Training Pegs

UW BioRobotics Lab TeleRobotic FLS

Figure 4.7: Training Pegs Diagram, order of pegs

Warm-up

If the subject had been away from the system for more than an hour, they were presented

with the training task board and were required to perform a 5-minute warm-up during

which they performed the training tasks. The warm-up was performed with no time delay

and subjects were allowed to move at their own pace.

43

Chapter 5

TELEROBOTIC FLS PILOT STUDY

5.1 Introduction

The goal of the pilot study was to use a limited number of non-surgeon subjects to debug the

TeleRobotic FLS experimental methods and to collect preliminary data for the time-delayed

Block Transfer task.

5.2 Methods

The experimental set-up, methods, and training were run as described in Chapter 4, Section

4.3. During this study, the RAVEN used modified Computer Motion Zeus Micro-wrist needle

drivers, as shown in Figure 4.1.

5.2.1 Subject Population

Three non-surgeon subjects, all right-handed, two male and one female, with ages ranging

from 28 to 39, completed this study under University of Washington Human Subjects Ap-

proval Number 01-825-E/B07. The subjects first performed the training tasks in order to

learn how to telemanipulate using the RAVEN. Within one week from the start of their

training, they returned to perform the pilot study.

5.2.2 Pilot Study

The Block Transfer task consists of moving six blocks, one at a time, first from the left

side of the FLS peg board (Figure 5.1) to the right side, by grasping with the left tool and

transferring to the right tool, then setting the block down on a peg on the right side. The

subject repeats this by moving all six blocks from right to left, grasping with the right tool

then transferring to the left tool. One trial consists of moving the blocks from the left to

the right, then back from the right to the left, for a total of twelve transfers.

44

Figure 5.1: The SAGES FLS Block Transfer task board set up with the RAVEN moving ablock from left to right.

Subjects performed three repetitions of each of three treatments for a total of nine trials.

The three treatments included a delay of 0ms as well as two non-zero delays. The nine trials

were arranged in a pseudo-random fashion. The treatments were grouped into three bins so

that each bin contained one of each of the three treatments. Designing the order so that the

subject received one of each of the treatments before receiving a second (and so on) meant

that, if a learning effect was present during the experiment, the improvement in performance

would be distributed more evenly across the different treatments. The subjects were allowed

a short break between each trial while the experimenter prepared the next treatment. After

both the third and sixth trials, subjects were required to take a 5-minute break. Though

each treatment was presented in random order, subjects were told which treatment they

would be receiving before each trial. By informing the subjects of which delay condition

they would experience, they could mentally accommodate for the delay condition, just as

they would if they were performing telesurgery on a patient. The experimenter was allowed

to answer clarifying questions about the task, but was not allowed to coach subjects on

strategies to approach the task.

The times were recorded for moving from left to right and right to left, as well as the

individual split times for each transfer. For this experiment only one type of error was

reported; any dropped block, whether recovered or not. The number of errors for each trial

were recorded.

45

Table 5.1: Mean completion time for single block transfer

Treatment - Delay (ms)Mean Completion Time (sec)

Sub1 Sub2 Sub3

A - 0 41.68 49.85 120.38

B - 250 68.85 77.15 126.38

C - 500 n/a 121.77 170.40

D - 1000 140.96 n/a n/a

5.3 Results

The first subject performed the study using delays of 0ms, 250ms and 1000ms. It was deter-

mined after the first subject that 1000ms delay made the overall experiment prohibitively

long and the subject commented about fatigue. The second and third subjects performed

the study using delays of 0ms, 250ms, and 500ms. The first two subjects completed all

nine trials for a total of 108 individual transfers. The third subject only completed five of

the nine trials before mechanical failures of the graspers prevented further use of the tools.

The results listed are the mean completion time for a single block transfer, based on the

aggregate of the six blocks transferred from left to right and six from right to left per trial.

Table 5.1 lists the mean completion time of each subject for each delay. Table 6.5 lists the

error rate as a percentage of errors with respect to the number of transfers attempted.

5.3.1 Statistical Analysis

Multiple paired t-tests were conducted to see if there was a statistically significant difference

in the mean block transfer time with different time delays for each subject. For a single

paired t-test, a significance level of 0.05 would be used to test the null hypothesis; however

as multiple hypotheses were being tested, a Bonferroni correction was used. The Bonferroni

correction states that, if an experimenter is testing n independent hypotheses on a set of

data, then the statistical significance level that should be used for each hypothesis separately

46

Table 5.2: Percentage of block transfers that resulted in error (dropped block)

Treatment - Delay (ms)Error rate

Sub1 Sub2 Sub3

A - 0 11% 11% 0%

B - 250 8% 19% 8%

C - 500 n/a 36% 8%

D - 1000 3% n/a n/a

Table 5.3: p-values for paired t-tests. Significance level of p < 8.333e − 3 required to showsignificant difference. *Note: Subject 1 comparisons substitute Treatment D for C

t-testp-value

Sub1* Sub2 Sub3

(A,B) 7.426e-8 9.420e-7 0.5322

(A,C) <2.2e-16 2.583e-11 2.510e-3

(B,C) 1.032e-12 3.842e-6 4.529e-3

(Alr,Arl) 0.1913 0.0480 0.5219

(Blr,Brl) 0.4132 0.4091 0.3288

(Clr,Crl) 0.7681 0.5247 0.1844

is 1/n times what it would be if only one hypothesis were tested. In this case, since n = 6, a

significance level of p < 0.05(1/6) = 0.00833 is required. Table 5.3 summarizes the results of

the paired t-tests. The tests (A,B), (A,C), (B,C) compare the block transfer time between

the three treatments (e.g. (A,B) compares the block transfer times for the 0ms and 250ms

delay conditions). The tests (Alr,Arl), (Blr,Brl) and (Clr,Crl) compare the block transfer

times moving left-to-right (lr) and right-to-left (rl) for each treatment.

The statistical analysis shows that, for all three subjects, there was no significant differ-

ence in the mean block transfer time, whether moving the blocks from the left to the right

47

or from the right to the left. This result indicates that this bi-manual task is symmetric

with respect to moving from left to right or right to left. There was a significant difference

in mean block transfer time between the different treatments for each of the three subjects,

except when comparing Treatment A and Treatment B for Subject 3.

5.4 Discussion

Intuitively, one might expect that as delay increases, so would the mean block transfer

time. Subject 1 had the lowest mean block transfer times and also showed fewer errors as

delay increased. Subject 1 commented that he was attempting to be more careful under

longer delay conditions. Subject 2 and Subject 3 both had more errors with greater delay.

Subject 3 did not show a significant difference in mean block transfer time between the

no-delay and 250ms delay conditions. Subject 3 had the highest mean completion time of

the three subjects. This suggests that a subject who moves more slowly (and potentially

more carefully) in the no-delay condition will suffer lower performance decreases as delay

increases when compared with subjects who generally move faster. This hypothesis may be

tested in future work.

The pilot study sucessfully served its purposes of collecting initial data and “debugging”

the experimental method. Appropriate changes were made for the main study (Chapter 6)

based on the experience gained in this smaller study.

48

Chapter 6

EFFECT OF TIME DELAY ON TELEROBOTIC FLS BLOCKTRANSFER

6.1 Introduction

This chapter describes a study that investigated the effect of time delay on surgeons and

non-surgeons performing the Block Transfer task using the RAVEN. Two primary research

questions were answered by this study. First, does delay significantly effect performance?;

second, is there a significant difference between surgeons’ and non-surgeons’ performance?

6.2 Methods

The experimental set-up, methods, and training were run as described in Chapter 4, Sec-

tion 4.3. During the pilot study (Chapter 5), the original mirco-wrist tools suffered from

fatigue due to over use and became non-functional. For this study, non-wristed, disposable

Maryland graspers from Covidien (Mansfield, MA) were adapted for use on the RAVEN.

6.2.1 Subject Population

Fifteen subjects, six surgeons and nine non-surgeons, nine male and six female, ages ranging

from 18 to 43, participated in this study under University of Washington Human Subjects

Approval Number 01-825-E/B07. The subjects performed the training tasks first with no

delay, and then with 250ms delay, in order to learn how to telemanipulate using the RAVEN.

Within one week from the start of their training, they returned to perform the Time Delayed

Block Transfer experiment. Due to scheduling constraints, many of the subjects completed

the training and the experiment on the same day.

49

Table 6.1: Number of repetitions for each of the training tasks

Task No Delay Delay (250ms)

1A, 1B 25 10

2A, 2B 15 10

3A, 3B 12 10

6.2.2 Experiment Overview

The subjects first watched three videos. The first was an orientation to the RAVEN that

provided them with an overview of the system, as well as instructions on how to telemanip-

ulate with it. The second video explained and demonstrated the training tasks (subsection

6.2.3). The third video explained and demonstrated the Block Transfer task (subsection

6.2.4). The subjects then read the tips sheet included in Appendix C.

The subjects performed the Block Transfer manually using Covidien Maryland 5mm

graspers on a trainer box with a clear plastic lid. This set-up allowed the subjects to

manipulate the blocks under direct vision of the operative site. The subjects performed

three trials of the Block Transfer task so that they would become familiar with the task.

Understanding the task would also help see the value in each of the three training tasks.

The subjects then performed the training, followed by the main portion of this study.

6.2.3 Training

During the pilot study (Chapter 5), the subjects performed the training tasks until their

competition time did not improve over three consecutive trials. It was noted that subjects

tended to plateau after approximately the same number of trials. In this study, for consis-

tency, each subject was required to perform the same number of repetitions for each of the

different training tasks (summarized in Table 6.1).

50

Treatment Delay (ms)

A 0

B 250

C 500

Table 6.2: Three different conditions were presented to each subject

6.2.4 Experimental Design

In this study, three delay conditions were investigated (summarized in Table 6.2): 0ms

(Treatment A), 250ms (Treatment B), and 500ms (Treatment C). Subjects performed three

repetitions of each of the three treatments for a total of nine trials. The nine trials were

arranged in a pseudo-random fashion. The treatments were grouped into three bins so

that each bin contained one of each of the three treatments. Six possible bins resulted

from the permutations of the three conditions (see Table 6.3). When subjects arrived

they performed an urn sampling without replacement of three numbered balls from a box

originally containing six balls. The order of the balls determined the order of the bins. For

example if the subject drew 3, then 5, then 1, the order of their nine trials would be (B,

A, C), (C, A, B), (A, B, C). The complete order for each of the subjects is show in the

Appendix, Table A.1. Designing the order so that the subject received one of each of the

treatments before receiving a second (and so on) meant that ,if a learning effect was present

during the experiment, the improvement in performance would be distributed more evenly

across the different treatments.

Between each of the trials, the subjects received a short break, and after the third and

sixth trial, the subjects were required to take a minimum 5-minute break to help minimize

fatigue. Before each trial the subject was told which delay treatment they were being given.

Stated Objective

The subjects were told that, although they were being timed, speed should not be their

optimizing factor. They were instructed that the goal of the exercise was to transfer each

51

Table 6.3: Treatment order for each of the six permutations

Number Order

1 A, B, C

2 A, C, B

3 B, A, C

4 B, C, A

5 C, A, B

6 C, B, A

block carefully, so that no errors were made (no dropped blocks) at whatever speed was

necessary to insure successful transfer.

6.3 Results

6.3.1 Training

Of the 15 subjects who started the training, 14 finished. Completion of the training portion

of this experiment was a prerequisite to move to the Block Transfer task. The results of

the training task are plotted in Appendix A. One subject had difficulty with the spatial

relationship between her hands and the tools. This struggled through the non-delayed

training, but was unable to complete Training Task 1A under the 250ms delay. The subject

voluntarily requested to be withdrawn from the study.

6.3.2 Block Transfer under Time Delay

Five surgeons and nine non-surgeons completed the main portion of this experiment for a

total of 14 subjects. Each subject performed three repetitions of each of the three treat-

ments, for a total of 108 block transfers (36 for each treatment) with the time for each

transfer recorded. All nine trials for each of the subjects were illustrated using a boxplot to

see if there was learning throughout experiment. Although the experimental design sought

to minimize the learning effect, some learning was exhibited between the first, second, and

52

third repetition of each treatment. Some subjects exhibited more learning in some condi-

tions than others. The boxplots for each of the 14 subjects who completed the experiment

can be found in Appendix A.

For each subject, the mean block transfer time for each of the three conditions was

calculated. A linear regression was fit to these times and, in all but one case, was fit with

an R2 > 0.969. The slope of the linear fit to block transfer time versus delay represents

the subjects’ sensitivity to delay (how much their completion time increases with increasing

delay) and the y-intercept represents their estimated performance in the nominal (no delay)

condition. The results are listed in Table 6.3.2.

Errors were defined as a block that was dropped. These were further classified as dropped

blocks that were recovered and dropped blocks that were unrecovered. While an aggregate

score or weighting is not given to each of the two types of errors, a block that was not

recovered would be considered a worse error. Table 6.5 lists the total number of errors from

each subject over three repetitions of each of the three treatments and the total number of

errors for the experiment. Figure 6.1 displays the results from Table 6.5 graphically.

Figure 6.1: Two types of errors are shown, recovered and unrecovered, for each of the 15subjects. The number of errors is the total over three trials of each condition.

While the TeleRobotic FLS as described in Chapter 4 reports Block Transfer times and

two types of errors, the RAVEN inherently is able to capture motion data as well. Position

data of each tool was recorded during the trials. For each trial the distance traversed by

53

Table 6.4: Subject by Subject Mean Block Transfer Times reported in seconds for each ofthe three conditions. The mean times for each subject were fitted to a linear regression.The Delay Sensitivity is the slope of the linear fit. The R2 value associated with the linearregression is also listed.

Subject# Surgeon Treatment A Treatment B Treatment C

Delay

Sensitiv-

ity

(ms/ms)

R2

1 N 31.74 43.13 64.63 65.78 0.9695

2 N 33.90 46.98 65.59 63.38 0.9900

3 N 48.8 74.46 90.34 83.08 0.9819

4 N 43.59 66.60 104.98 122.78 0.9795

5 N 22.45 34.65 46.19 47.48 0.9997

6 Y 38.03 51.47 69.34 62.62 0.9934

7 N 32.15 43.93 64.75 65.20 0.9750

8 N 53.84 74.76 103.94 100.20 0.9910

9 Y 40.95 68.61 88.42 94.94 0.9910

10 Y 36.79 60.94 80.03 86.48 0.9955

11 Y NC NC NC NA NA

12 Y 44.18 60.53 86.33 84.30 0.9835

13 Y 23.32 33.08 52.4 58.16 0.9652

14 N 33.96 40.69 66.41 64.90 0.8975

15 N 33.1 48.87 71.38 76.56 0.9898

54

Table 6.5: Errors for each of the three treatments over three trials for each and the totalerrors over all nine trials

Subject# SurgeonTreatment A Treatment B Treatment C

Totalrec unrec rec unrec rec unrec

1 N 0 0 1 0 0 0 1

2 N 1 1 0 2 0 0 4

3 N 2 0 1 1 0 0 4

4 N 0 0 0 1 1 2 4

5 N 0 0 0 0 0 1 1

6 Y 2 0 0 1 1 0 4

7 N 3 0 1 0 2 0 6

8 N 0 0 0 0 0 0 0

9 Y 1 0 0 0 1 1 3

10 Y 1 1 1 0 2 2 7

11 Y NA NA NA NA NA NA NA

12 Y 2 0 1 1 1 1 6

13 Y 0 0 1 0 0 0 1

14 N 2 1 0 0 3 0 6

15 N 0 1 1 0 0 0 2

55

each tool was calculated. The path length is the sum of the distance traversed by the left

and right tools. Table 6.3.2 shows the average path length for each subject for each of the

delay conditions. Path length data from Subjects 1 and 2 were not properly captured and

cannot be reported.

6.3.3 Statistical Analysis

Statistical analysis was performed using R. A two-way analysis of variance (ANOVA) was

used to determine the significant results. Three analyses were performed separately. The

first used block transfer time as the response variable; the second used the number of error;

the third used tool tip path length as the response variable.

The stated objective of the task was to minimize errors. Analysis of number of errors in

response to delay effect and surgeon effect was not significant, but the cause of errors was

noted when possible and is included in the Discussion. It is possible that if the task was

more technically challenging, the frequency or severity of errors would start to differentiate

between subjects with more and less skill.

Delay Effect

The difference in mean block transfer time between each of the three treatments (0ms,

250ms, and 500ms delay) is statistically significant, and is illustrated in Figure 6.2. The dif-

ference in mean path length between each of the three treatments is significant as illustrated

in Figure 6.3.

Surgeon Effect

The difference in mean block transfer between surgeons and non-surgeons is not statistically

significant, and is illustrated in Figure 6.4. The difference in mean path length between

surgeons and non-surgeons is significant at the 0.05 level, and is illustrated in Figure 6.5.

Figure 6.6 and 6.7 illustrate response variables with respect to both Delay Effect and Surgeon

Effect.

56

Table 6.6: Subject by Subject Mean Block Transfer Path Length reported in meters foreach of the three conditions. The mean path length for each subject were fitted to a linearregression. The Delay Sensitivity is the slope of the linear fit in millimeters of increasedpath length per millisecond of increased delay. The R2 value associated with the linearregression is also listed.

Subject# Surgeon Treatment A Treatment B Treatment C

Delay

Sensitiv-

ity

(ms/ms)

R2

1 N NC NC NC NA NA

2 N NC NC NC NA NA

3 N 5.986 7.2858 8.1057 4.24 0.9832

4 N 8.2088 11.7815 16.3581 16.29 0.9950

5 N 5.3507 6.2531 7.3290 3.96 0.9974

6 Y 7.6223 8.8765 9.7528 4.26 0.9896

7 N 6.912 7.7486 9.4067 4.99 0.9652

8 N 8.5130 11.3580 13.4632 9.90 0.9926

9 Y 8.5280 10.2894 11.7445 6.43 0.9970

10 Y 9.4084 15.0260 15.1186 11.42 0.7622

11 Y NC NC NC NA NA

12 Y 8.7550 10.9600 14.0210 10.53 0.9913

13 Y 7.1420 8.6191 10.4480 6.61 0.9962

14 N 7.5452 8.3858 13.3066 11.52 0.8568

15 N 7.7066 10.4009 11.6348 7.86 0.9560

57

Table 6.7: ANOVA Results

Measure Source Sig.

Time

Delay < 2e − 16

Surgeon? 0.8113

Delay*Surgeon? 0.9728

Path Length

Delay 1.978e − 08

Surgeon? 0.01962

Delay*Surgeon? 0.74864

6.4 Discussion

At the outset, there was a hypothesis that surgeons might be more careful and therefore

less prone to errors. Another possibility is that surgeons who perform MIS cases would be

more adapted to the lack of depth perception. It was also suggested that surgeons would

be better at accommodating to delay. The statistical analysis shows that there is no sig-

nificant difference between surgeons and non-surgeons for these telerobotic manipulation

tasks. Clearly, the ability to perform surgery extends far beyond the psycho-motor skill

of moving blocks on a peg board, requires (among many other skills) a high level of cog-

nitive ability, familiarity with anatomy, and the ability to deal with unexpected problems.

However, for the purposes of development on the surgical robot, these results imply that

performing usability or evaluation experiments with non-surgeons may be adequate for fu-

ture simple task experiments such as the Block Transfer. Complex tasks such as suturing,

may require surgeon participation. In addition, the effects of delay may be more dependent

on the individual than on the individual’s surgical expertise.

6.4.1 Errors - Dropped Blocks

Statsical analysis of the errors indicate that the number of errors is not a result of delay

or whether or not the subject was a surgeon. There were a number of different causes of

errors that were observed during the experiment. These are listed below: (Because the task

58

0 250 500

2040

6080

100

120

Delay Effect

Delay (ms)

Tim

e (s

ec)

Figure 6.2: Delay Effect versus block transfer time. 0ms average block transfer time was36.95sec; 250ms average block transfer time was 53.60sec; 500ms average block transfer timewas 75.44sec.

was bimanual, the primary hand is referred to as the hand with which the block is initially

picked up and the transfer hand is the hand to which the block is transferred.)

• The subject bumped the release button inadvertently.

• Mental lapse - The subject pushed the wrong button inadvertently.

• Depth perception error - The subject thinks that (s)he has grasped the block with the

transfer hand but, in fact, has not. Releasing the block from the primary hand then

caused a drop.

• Incomplete grasp - The subject grasps the block with the primary hand from an

awkward angle and, during the transfer, does not have a good grasp with the transfer

hand, resulting in a drop.

59

0 250 500

510

1520

25

Delay Effect

Delay (ms)

Tot

alP

ath

(m)

Figure 6.3: Delay Effect versus path length. 0ms average tool tip path length was 7.629m;250ms average tool tip path length was 9.781m; 500ms average tool tip path length was11.724m.

• Incomplete release - The subject sets the block down on the peg, then drags it off the

peg with the tool, while retracting the transfer hand from the peg.

• Indexing error - In the pedal up state, the subject hits the release button. Upon

transition to pedal down, the block is released and dropped.

Errors related to depth perception are a function of the subjects ability to cognitively

map the 2-D view on the screen to the 3-D world in which their hands and the tools operate

in. In general,some subjects had notably fewer errors than others. Depth errors aside, this

was primarily due to how carefully the subject performed the task, but is not correlated

with average block transfer time. To the extent that the errors were avoidable, the number

of errors may be viewed as a numerical representation that gives some insight into the

subjects demeanor. Average block transfer time, by contrast, is an objective measure of

psychomotor ability.

60

N Y

2040

6080

100

120

Surgeon Effect

Surgeon?

Tim

e (s

ec)

Figure 6.4: Surgeon Effect versus block transfer time (data included from all three delayconditions). The non-surgeon group had an average block transfer time of 55.03sec and thesurgeon group had and average block trasnfer time of 55.79 seconds.

61

N Y

510

1520

25

Surgeon Effect

Surgeon?

Tot

alP

ath

(m)

Figure 6.5: Surgeon Effect versus path length.

0.N 250.N 500.N 0.Y 250.Y 500.Y

2040

6080

100

120

Delay+Surgeon Effect

Delay+Surgeon?

Tim

e (s

ec)

Figure 6.6: Surgeon Effect + Delay Effect versus block transfer time

62

0.N 250.N 500.N 0.Y 250.Y 500.Y

510

1520

25

Delay+Surgeon Effect

Delay+Surgeon?

Tot

alP

ath

(m)

Figure 6.7: Surgeon Effect + Delay Effect versus path length

63

Chapter 7

OTHER TELEROBOTIC FLS EXPERIMENTS

7.1 Introduction

This chapter presents further applications of TeleRobotic FLS. The Intracorporeal Knot

Tying task is performed using the RAVEN under local teleoperation without any delay. A

comparison between RAVEN end-effectors is made using the results from the pilot study

(Chapter 5) and the primary study (Chapter 6). All three of the TeleRobotic FLS tasks

are performed with the ISI da Vinci R© with and without the stereoscopic view to determine

effect of 3D vision on performance.

7.2 Intracorpreal Knot Tying - RAVEN

Knot tying is a relatively complex task that requires both the dexterity and physical ability

to perform the task, and also the cognitive understanding to be able to execute the task

properly. A single surgeon with RAVEN experience performed eight repetitions of the

Intracorporeal Knot Tying task using the RAVEN with two adapted CMI Endo-wrist small

needle drivers.

Methods

The methods used are described in Chapter 4.2.2. Other than the subjects prior experi-

ence on the RAVEN and some informal practice suturing, the subject did not receive any

structured training for this task.

Results

Table 7.1 shows the results of this experiment, which include each of the 8 trials, the

completion time, entrance error (mm), exit error (mm), gap (mm) and if the knot was

secure. Ideally, the surgeon would perform the task in a short amount of time, with the

64

least amount of error, gap, and with a secure knot. On the first day, the surgeon averaged

796 seconds per repetition over three repetitions. A week later, the surgeon averaged 448

seconds over five repetitions. Using a paired t-test, the improvement between the first and

second day’s trials is not statistically significant at the 0.05 level (p = 0.1263) but that is

likely due to the low number of repetitions. Very little knot tying has been performed on

the RAVEN and these limited samplings represent the very steep initial part of the learning

curve for this task on the RAVEN.

Discussion

In order for surgeons to perform telerobotic surgical procedures on animal models, the

RAVEN must be capable of knot tying. This clinically experienced surgeon only tied a

total of eight intracorporeal knots with the RAVEN. For comparison, first year residents

averaged approximately 400 seconds to perform a knot tying task on a porcine model using

regular MIS tools [24]. Additionally from the MISTELS literature [8] the cut-off time for

the knot tying task was 600 seconds. These initial results fall within the range that should

be expected for knot tying performance.

7.3 Comparison of End Effectors - RAVEN

In the Pilot Study (Chapter 5), three subjects performed the Block Transfer task with the

RAVEN using adapted Computer Motion Zeus Micro-assist graspers. For the main study,

subjects performed the same tasks with the RAVEN using adapted Covidien Maryland

Graspers (Chapter 6).

In the Pilot study, Subject 3 was the same subject as Subject 15 in this study. The

difference between the pilot study and this study is a change from a tool with a wrist to

a tool without a wrist. While a sample size of 1 can hardly be considered sufficient for

comparison, this subject showed dramatic improvements in completion time with the non-

wristed tools. Conventional wisdom in robotics is that the increase in dexterity due to the

wrist would make performing a task easier (and therefore faster). For the task of moving

the FLS blocks around the peg board, the lack of a wrist made it easier for the subject to

manipulate the tool tips. This may be due in large part to a non-optimal input device that

65

Table 7.1: Intracorporeal Knot Tying Task, RAVEN Surgical Robot

Date Trail Number Completion Time (s) entrance error (mm) exit error (mm) Gap (mm) Knot Secure

Day 1 1 1074 0.2 0.6 0.0 yes

2 640 0.0 0.4 0.0 yes

3 673 0.0 0.7 0.0 yes

Day 2 1 505 0.0 2.6 0.0 yes

2 380 0.8 0.1 0.0 yes

3 459 0.0 0.8 0.0 yes

4 406 0.0 1.5 0.0 yes

5 490 1.0 3.0 0.0 yes

66

was mapping 6-DOF at the hands of the operator (3 position, 3 orientation) to 5-DOF at

the tool tips (3 position, 2 orientation). In order to articulate the wrist, the operator had

to pitch the stylus. This non-intuitive mapping of orientation made it difficult for subjects

to perform the manipulation tasks they wanted. The simpler non-wristed tool eliminated

this complication, making the manipulation much more straightforward.

In a study that would track changes to the system it would be difficult to maintain precise

control over all the potentially confounding variables and simply isolate one (in this case End

Effectors). However, with the assumption that the performance of the subjects is normally

distributed, this section compares the results of the non-delayed average block transfer

times, with the Computer Motion Zeus single wrist graspers to the Covidien Maryland

graspers. In the pilot study, using the wristed tools, the average no-delay block transfer

time for the three subjects was 56.68 seconds. In the main study, using the non-wristed

tools, the average no-delay block transfer time for all 14 subjects was 36.95 seconds. Using

a paired T-test the difference in means is statistically significant (p = 1.382e − 06). This

implies that for subjects performing the block transfer task with the RAVEN in its current

master/slave configuration, using a non-wristed tool presents a performance advantage.

7.4 TeleRobotic FLS - Effect of Sterescopic View, ISI da Vinci R©

In this set of experiments the ISI da Vinci R© was configured for dry-lab tasks, not operating

through MIS ports. The operative site was viewed through a 12mm 3D 0◦ Scholly endoscope.

For the Block Transfer and Knot Tying tasks two 8mm needle drivers were used. For pattern

cutting an 8mm needle driver was used for the left tool and a Potts scissors was used for

the right tool.

7.4.1 Block Transfer

Subject Population

Four surgeons participated in this set of experiments to determine the effect of 3D on com-

pletion time. The population consisted of two practicing urologists with clinical experience

on the da Vinci R©, and one general surgeon and one urology resident, both with limited

67

experience on the da Vinci R©.

Methods

The methods used are described in Chapter 4.2.1. Subject 1 used the “ultra fine” motion

scaling which is a mapping of 5:1 ratio of master to slave motion. Subjects 2, 3, and 4 used

the “fine” motion scaling which is a mapping of 3:1. Each subject was allowed to warm

up for 5 minutes by practicing the block transfer task with the 3D turned on. Subject 1

performed the experiment first without 3D, then with 3D. Subjects 2, 3, and 4 performed

the experiment first with 3D, then without 3D.

Results

The Block Transfer task performed with the da Vinci R© was accomplished much more quickly

than with the RAVEN, so times are reported in Table 7.2 for moving all six blocks left to

right, then right to left. The mean one-direction transfer time without 3D was 62.93 sec and

with 3D was 48.53 seconds. The results for mean block transfer time were 10.49 seconds

without 3D and 8.09 seconds with 3D. Using a paired t-test, the difference in mean times

is statistically significant (p = 1.769e − 4 < 0.05).

Observations

Subjects 1, 2, and 3 participated in both this study and the study using the RAVEN

described in Chapter 6. A strikingly different approach was taken by all surgeons using the

da Vinci R©. With the RAVEN and with standard MIS graspers, subjects pick up the block

and transfer it to the other hand on the same side of the block (the “top”; the most readily

viewable in the operative scene). With the da Vinci R©, the surgeons transferred the blocks

between the two hands in a “top to bottom” fashion by articulating the wrist in much the

way persons might naturally perform the task if doing it directly with their hands.

Subject 3 has limited experience on the da Vinci R© and had not used it at all in over

a year. Subject 4 had only gone through an hour of dry lab training and had one hour of

clinical time at the controls. It is a testament to the da Vinci R© that these two surgeons

68

were able to adapt to it and use it so readily.

7.4.2 Intracorporeal Knot Tying

Subject Population:

A single surgeon, Subject 1 from the Block Transfer experiment in the previous section,

participated in this study.

Methods

The methods used are described in Chapter 4.2.2. The subject used the “ultra fine” motion

scaling which is a mapping of 5:1 ratio of master to slave motion. The subject performed

the experiment described in Section 7.4.1 immediately prior to this set of tasks. Three

repetitions of the knot tying task with 2D, then three repetitions with 3D, were performed.

Results

The surgeon was able to suture the rubber (Penrose) drain with a secure knot and no gap

with and without 3D. The mean time to complete the knot tying task without 3D was 169.3

seconds and with 3D was 85.3 seconds. A paired t-test to compare the difference in means

resulted in p = 0.08591. The result is not statistically significant at the 0.05, level which is

probably due to the limited number of samples taken for each of the two conditions.

By comparison, the same surgeon had a fastest knot tying time of 380 seconds with the

RAVEN under no delay. The surgeon had much more experience on the ISI da Vinci R©.

The user interface of the da Vinci R© as well as the fully articulated wrists lend themselves

to dexterous tasks such as knot tying.

7.4.3 Pattern Cutting

Subject Population:

A single surgeon, Subject 1 from the Block Transfer experiment in the previous section,

participated in this study.

69

Table 7.2: Block Transfer Task, ISI da Vinci R©

Sub Vision Trial Number L-R Time (s) R-L Time Errors

1 2D 1 84 78 0

2 62 65 0

3 66 78 0

3D 1 69 64 0

2 49 51 0

3 63 51 0

2 2D 1 56.10 37.09 0

2 45.15 50.38 0

3D 1 35.12 32.31 0

2 38.31 35.28 2

3 37.50 37.18 0

3 2D 1 68.53 70.19 0

2 57.91 70.47 0

3 60.34 49.72 0

3D 1 63.43 52.93 1

2 53.47 48.66 0

3 54.97 48.18 0

4 2D 1 59.88 95.59 0

2 63.72 57.88 0

3 55.06 53.56 0

3D 1 46.69 49.87 0

2 47.40 46.91 0

3 50.13 41.40 0

70

Table 7.3: Intracorporeal Knot Tying Task, ISI da Vinci R©

Vision Trial Number Completion Time (s) Entrance error (mm) Exit error (mm) Gap (mm) Knot Secure

2D 1 225 0.0 1.0 0.0 yes

2 146 0.0 0.5 0.0 yes

3 137 1.8 0.0 0.0 yes

3D 1 95 0.0 0.8 0.0 yes

2 70 0.3 0.8 0.0 yes

3 91 0.0 0.0 0.0 yes

71

Table 7.4: Pattern Cutting, ISI da Vinci R©

Vision Trial Number Completion Time Errors

2D 1 224 0

2 166 0

3 234 0

3D 1 216 0

2 179 0

Methods

The methods used are described in Chapter 4.2.3. The subject used the “ultra fine” motion

scaling which is a mapping of 5:1 ratio of master to slave motion. The subject performed

the experiments described in section 7.4.1 and section 7.4.2 immediately prior to this set of

tasks. Three repetitions of the pattern cutting task with 2D, then two repetitions with 3D,

were performed. The subject was right handed and used a fine grasper in the left hand and

a Potts scissors (not curved) in the right hand.

Results

The surgeon was able to cut between the concentric circles (no errors) with and without

3D. The mean time to complete the pattern cutting without 3D was 208.0 sec and with 3D

was 197.5. Using a paired t-test, the difference in mean is not statistically significant at the

0.05 level (p = 0.7348). The result is not surprising because the pattern cutting task has

the subject approach a flat gauze pad and cut on this planar surface. The subject did not

need to move in depth once contact with the gauze pad was established.

7.5 Discussion

While the SAGE FLS scoring system is a proprietary formula, the McGill MISTELS liter-

ature [11] describes the tasks from which the FLS tasks were derived. This literature lists

the cut-off time for each of the tasks, which is included in Table 7.5. In comparison to the

72

Table 7.5: MISTELS task cutoff times

Task Cutoff Time (s)

Block Transfer 300

Intracorporeal Knot 600

Pattern Cutting 300

TeleRobotic FLS Block Transfer study (Chapter 6), which reports average block transfer

time, the 300 second cutoff for MISTELS is the total time for all 12 transfers (6 left-to-right

then back right-to-left).

Comparison between the UW RAVEN and ISI da Vinci R© can be made for the Block

Transfer and Intracorporeal Knot Tying tasks. The most fair comparison is the RAVEN

under no delay to the da Vinci R© without 3D. In the Block Transfer, the da Vinci R© no-3D

mean block transfer time (12.0 sec) compares to the RAVEN no-delay mean block transfer

time (23.3 sec). For knot tying, the mean completion time was 169.3 seconds on the da

Vinci R© with 2D vision. By comparison the same surgeon had a fastest knot tying time of

380 seconds with the RAVEN under no delay. The surgeon had much more experience on

the ISI da Vinci R© but also the user interface of the da Vinci R© as well as the fully articulated

wrists lend themselves to dexterous tasks such as knot tying.

Beyond the pure performance results, which might lead one to think the UW RAVEN is

simply an inferior system, there are many contrasts to draw between the two systems. First,

and most obviously, is that the da Vinci R© is a commercially available system in relatively

widespread use in operating rooms across the world. Significant engineering, clinical, and

financial resources have been put into the development of the da Vinci R©. It is able to

address the needs of clinicians today. The RAVEN, in contrast, was designed for the needs

of the future. It is significantly more compact and lightweight, making it portable and

deployable. It is inherently designed for telesurgery across nearly any terrestrial distance.

These features were highlighted during the HAPs/MRT (Chapter 2.5.1) and NEEMO XII

(Chapter 2.5.4) missions.

73

7.6 Conclusions

TeleRobotic FLS prvovides a set of tasks with quantifiable metrics by which to evaluate

surgical robot performance. It has been used in this thesis to study the effect of time delay

on a bimanual gross manipulation task using the RAVEN, as well as studying the effect of

stereoscopic vision on the ISI da Vinci R©. It will be a useful tool, providing an empirical

measurement, for tracking changes or improvements to this or other systems, as well as

improvements in subject performance. TeleRobotic FLS represents a common language by

which new robotic platforms can be evaluated.

74

BIBLIOGRAPHY

[1] J. Arata, H. Takahashi, P. Pitakwatchara, S. Warisawa, K. Konishi, K. Tanoue, S. Ieiri,S. Shimizu, N. Nakashima, K. Okamura, Young Soo Kim, Sung Min Kim, Joon-SooHahm, M. Hashizume, and M. Mitsuishi. A remote surgery experiment between japan-korea using the minimally invasive surgical system. In Robotics and Automation, 2003.Proceedings. ICRA ’06. IEEE International Conference on, pages 257–262, 2006.

[2] J. Arata, H. Takahashi, P. Pitakwatchara, S. Warisawa, K. Tanoue, K. Konishi, S. Ieiri,S. Shimizu, N. Nakashima, K. Okamura, Y. Fujino, Y. Ueda, P. Chotiwan, M. Mitsuishi,and M. Hashizume. A remote surgery experiment between japan and thailand overinternet using a low latency codec system. Robotics and Automation, 2007 IEEEInternational Conference on, pages 953–959, April 2007.

[3] P. Berkelman and Ji Ma. The university of hawaii teleoperated robotic surgery system.Intelligent Robots and Systems, 2007. IROS 2007. IEEE/RSJ International Conferenceon, pages 2565–2566, 29 2007-Nov. 2 2007.

[4] Peter Berkelman and Ji Ma. A compact, modular, teleoperated robotic minimallyinvasive surgery system. In Biomedical Robotics and Biomechatronics, 2006. BioRob2006. The First IEEE/RAS-EMBS International Conference on, 2006.

[5] M.C. Cavusoglu, F. Tendick, M. Cohn, and S.S. Sastry. A laparoscopic telesurgicalworkstation. IEEE Transactions on Robotics and Automation, 15(4):728–739, August1999.

[6] Kevin Clearly and Charles Nguyen. State of the art in surgical robotics: Clinicalapplications and technology challenges. Computer Aided Surgery, 6:312–328, 2001.

[7] B. Davies. A review of robotics in surgery. Proc. Instn Mech Engrs, 214(H), 2000.

[8] A.M Derosis, G. Fried, M. Abrahamowicz, H.H. Sigman, J.S Barkun, and J.L. Meakins.Development of a model for training and evaluation of laparoscopic skills. AmericanJournal of Surgery, 2175, 1998.

[9] A.M Derosis, G. Fried, M. Abrahamowicz, H.H. Sigman, J.S Barkun, and J.L. Meakins.Development of a model for training and evaluation of laparoscopic skills. AmericanJournal of Surgery, 2175, 1998.

[10] S.A. Fraser, D.R. Klassen, L.S. Feldman, G.A. Ghitulescu, D. Stanbridge, and G.M.Fried. Evaluating laparoscopic skills. Surgical Endoscopy, 17, 2003.

75

[11] G. Fried, L. Feldman, M.Vassiliou, S. Fraser, D. Standbridge, G. Ghitulescu, and C. An-drew. Proving the value of simulation in laparoscopic surgery. Annals of Surgery,240(3), September 2004.

[12] Scott William Gunther. The Red DRAGON: a spherical mechanism tracking device forminimally invasive surgical training. Master’s thesis, University of Washington, 2006.

[13] J. Rosen J., M. MacFarlane, C. Richards, B. Hannaford, C. Pellegrini, and M. Sinanan.Surgeon/endoscopic tool force-torque signatures in the evaluation of surgical skills dur-ing minimally invasive surgery. In Proceedings, MMVR-99 (Medicine Meets VirtualReality), San Francisco, January 1999.

[14] Y.S Kwoh, J.L. Hou, E.A. Jonckheere, and S. Hayall. A robot with improved absolutepositioning accuracy for ct guided stereotactic brain surgery. IEEE Tran. Biomed.Engr, 35(2):153–161, February 1988.

[15] MJH Lum. Kinematic Optimization of a 2-DOF Spherical Mechanism for a MinimallyInvasive Surgical Robot. Master’s thesis, University of Washington, 2004.

[16] M.J.H. Lum, D.C.W. Friedman, G. Sankaranarayanan, H. King, A. Wright,M. Sinanan, T. Lendvay, J. Rosen, and B. Hannaford. Objective assessment of telesur-gical robot systems: Telerobotic FLS. In Proceedings, Medicine Meets Virtual Reality(MMVR), Long Beach, CA, 29-Jan — 1-Feb 2008.

[17] M.J.H Lum, J. Rosen, H. King, D.C.W. Friedman, G. Donlin, G. Sankaranarayanan,B. Harnett, L. Huffnam, C. Doarn, T. Broderick, and B. Hannaford. Telesurgery viaunmanned aerial vehicle (UAV) with a field deployable surgical robot. In Proceedings,Medicine Meets Virtual Reality (MMVR), Long Beach, CA, 2007.

[18] M.J.H. Lum, J. Rosen, M.N. Sinanan, and B. Hannaford. Optimization of a SphericalMechanism for a Minimally Invasive Surgical Robot: Theoretical and ExperimentalApproaches. IEEE Transactions on Biomedical Engineering, 53(7):1440–1445, Jul 2006.

[19] A.J. Madhani, G. Niemeyer, and Jr. Salisbury, J.K. The black falcon: A teleoperatedsurgical instrument for minimally invasive surgery. In Proceedings of the IntelligentRobots and Systems, volume 2, pages 936–944, 1998.

[20] J. Marescaux. Transatlantic robot-assisted telesurgery. Nature, 413, Sept. 27.

[21] J. Marescaux, J. Leroy, F. Rubino, M. Smith, M. Vix, M. Simone, and D. Mutter.Transcontinental robot-assisted remote telesurgery: Feasibility and potential applica-tions advances in surgical technique. Annals of Surgery, 235(4), April 2002.

76

[22] M. Mitsuishi, J. Arata, K. Tanaka, M. Miyamoto amd T. Yoshidome, S. Iwata,M. Hashizume, and S. Warisawa. Development of a remote minimally-invasive surgicalsystem with operational environment transmission capability. In Robotics and Automa-tion, 2003. Proceedings. ICRA ’03. IEEE International Conference on, volume 2, pages2663–2670, 2003.

[23] J. Peters, G. Fried, L. Swanstrom, N. Soper, L. Sillin, B. Schirmer, K. Hoffman, andthe SAGES FLS Committee. Development and validation of a comprehensive programof education and assessment of the basic fundamentals of laparoscopic surgery. Surgery,135, 2003.

[24] J. Rosen, J.D. Brown, L. Chang, M. Sinanan, and B. Hannaford. Generalized approachfor modeling minimally invasive surgery as a stochastic process using a discrete markovmodel. IEEE Transactions on Biomedical Engineering, 53(3):399–413, March 2006.

[25] J. Rosen, JD Brown, L. Chang, MN Sinanan, and B. Hannaford. Generalized Approachfor Modeling Minimally Invasive Surgery as a Stochastic Process Using a DiscreteMarkov Model. IEEE Transactions on Biomedical Engineering, 53(3):399–413, 2006.

[26] G. Sankaranarayanan and B. Hannaford. Comparison of performance of virtual cou-pling schemes for haptic collaboration using real and emulated internet connections.In proceedings of ROBOCOMM, the first International Conference on Robot Commu-nication and Coordination, Athens, Greece, 2007.

[27] Richard Satava. History of robotic surgery: The early chronicles: a personal historicalperspective.

[28] H. Takahashi, S. Warisawa, M. Mitsuishi, J. Arata, and M. Hashizume. Developmentof high dexterity minimally invasive surgical system with augmented force feedbackcapability. Biomedical Robotics and Biomechatronics, 2006. BioRob 2006. The FirstIEEE/RAS-EMBS International Conference on, pages 284–289, 0-0 2006.

[29] R.H. Taylor. Robotics hip replacement surgery in dogs. 1989.

[30] Nabil Zemiti, Tobias Ortmaier, and Guillaume Morel. A new robot for force control inminimally invasive surgery. 2004.

77

Appendix A

PLOTS FROM THE EFFECT OF TIME DELAY STUDY WITHRAVEN

The appendix contains plots of the results from the Effect of Time Delay study using

the UW RAVEN.

A.1 Training

This section contains the plots for each of the six training tasks performed by the subjects,

first without delay, and then under a 250ms delay.

Training Task 1A - 0ms Delay

0

10

20

30

40

50

60

70

80

90

0 5 10 15 20 25

Repetition

Tim

e(s)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)

Figure A.1: Training Task 1A - 0ms delay

A.2 Time Delayed Block Transfer Plots

This section contains the results from each of the 14 subjects who completed the time

delayed block transfer experiment. Each figure is the result of the 9 trials performed by the

subject. The data is presented in the form a set of 9 boxplots. Each box represents the

inner quartile range (IQR), and the whiskers extend to the furthest data point within 1.5

IQR. Outliers are points that fall beyon 1.5 IQR.

The order of the trials for each subject are summarized below in Table A.1.

78

Training Task 1B - 0ms Delay

0

10

20

30

40

50

60

0 5 10 15 20 25

Repetition

Tim

e(s)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)

Figure A.2: Training Task 1B - 0ms delay


0

10

20

30

40

50

60

70

80

0 2 4 6 8 10 12 14 16

Repetition

Tim

e(s)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)



0

10

20

30

40

50

60

70

0 2 4 6 8 10 12 14 16

Repetition

Tim

e (s

)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)


79


0

10

20

30

40

50

60

70

80

90

100

0 2 4 6 8 10 12 14

Repetition

Tim

e (s

)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)



0

20

40

60

80

100

120

140

0 2 4 6 8 10 12 14

Repetition

Tim

e (s

)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)



0

10

20

30

40

50

60

70

80

0 2 4 6 8 10 12

Repetition

Tim

e(s)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub11*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub11*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub04)

Log. (Sub15)


80


0

10

20

30

40

50

60

70

0 2 4 6 8 10 12

Repetition

Tim

e(s)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)



0

10

20

30

40

50

60

70

80

90

0 2 4 6 8 10 12

Repetition

Tim

e(s)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)



0

10

20

30

40

50

60

70

80

90

0 2 4 6 8 10 12

Repetition

Tim

e (s

)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub04)


81


0

20

40

60

80

100

120

140

0 2 4 6 8 10 12

Repetition

Tim

e (s

)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)



0

20

40

60

80

100

120

140

0 2 4 6 8 10 12

Repetition

Tim

e (s

)

Sub02

Sub03

Sub04

Sub06*

Sub07

Sub08

Sub09*

Sub10*

Sub12*

Sub13*

Sub14

Sub15

Log. (Sub02)

Log. (Sub03)

Log. (Sub04)

Log. (Sub06*)

Log. (Sub07)

Log. (Sub08)

Log. (Sub09*)

Log. (Sub10*)

Log. (Sub12*)

Log. (Sub13*)

Log. (Sub14)

Log. (Sub15)


Subject Trial 1 Trial 2 Trial 3 Trial 4 Trial 5 Trial 6 Trial 7 Trial 8 Trial 9

1 A C B A C B B A C

2 B A C A B C A C B

3 B C A B A C C A B

4 A C B C A B C B A

5 C A B B A C C B A

6 A B C A C B B C A

7 C B A B A C A C B

8 A C B C A B B C A

9 A B C B A C A C B

10 B A C B C A C A B

12 B A C C B A A C B

13 C A B B A C C B A

14 B A C A B C B C A

15 B C A A C B A B C

Table A.1: Order of the nine trials

82

A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

Block Transfer times Sub01

Com

plet

ion

Tim

e (s

ec)

Mean = 39.49

Mean = 32.5

Mean = 23.23

Mean = 49.95

Mean = 41.63 Mean =

37.8

Mean = 74.4

Mean = 53.28

Figure A.13: Time Delayed Block Transfer Task, Subject 1, non-surgeon

A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

100

120


Com

plet

ion

Tim

e (s

ec)

Mean = 32.92

Mean = 33.78

Mean = 51.28

Mean = 41.52

Mean = 64.92

Mean = 65.29

Mean = 66.56


83

A1 A2 A3 B1 B2 B3 C1 C2 C3

5010

015

0


Com

plet

ion

Tim

e (s

ec)

Mean = 54.86 Mean =

47.98 Mean = 43.57

Mean = 93.81

Mean = 75.96

Mean = 53.61

Mean = 113.89

Mean = 81.25 Mean =

75.87


A1 A2 A3 B1 B2 B3 C1 C2 C3

5010

015

020

0


Com

plet

ion

Tim

e (s

ec)

Mean = 50.79 Mean =

43.5 Mean = 36.47

Mean = 73.41

Mean = 69.93

Mean = 56.46

Mean = 112.61 Mean =

108.5Mean = 93.83


84

A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080


Com

plet

ion

Tim

e (s

ec)

Mean = 21.85

Mean = 23.21

Mean = 22.31

Mean = 34.57

Mean = 36.2 Mean =

33.18

Mean = 42.69

Mean = 50.06 Mean =

45.84


A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

100

120


Com

plet

ion

Tim

e (s

ec)

Mean = 51.13

Mean = 36.17

Mean = 26.78

Mean = 51.23

Mean = 50.56

Mean = 52.6

Mean = 80.12

Mean = 65.75 Mean =

62.15

Figure A.18: Time Delayed Block Transfer Task, Subject 6, surgeon

85

A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

100

120


Com

plet

ion

Tim

e (s

ec)

Mean = 35.58

Mean = 25.96

Mean = 34.9

Mean = 44.51

Mean = 47.12

Mean = 77.91

Mean = 61.91

Mean = 54.45


A1 A2 A3 B1 B2 B3 C1 C2 C3

5010

015

0


Com

plet

ion

Tim

e (s

ec)

Mean = 54.43

Mean = 56.69 Mean =

50.4

Mean = 83.22

Mean = 66.09

Mean = 74.99

Mean = 119.22

Mean = 98.76 Mean =

93.84


86

A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

100

120

140

160


Com

plet

ion

Tim

e (s

ec)

Mean = 40.23

Mean = 46.83

Mean = 35.8

Mean = 75.6

Mean = 72.65

Mean = 57.59

Mean = 103.54

Mean = 87.62

Mean = 74.12


A1 A2 A3 A4 B1 B2 B3 C1 C2 C3

5010

015

0


Com

plet

ion

Tim

e (s

ec)

Mean = 56.23

Mean = 25.05

Mean = 33.87

Mean = 38.9

Mean = 85.16

Mean = 81.93 Mean =

73.02


87

A1 A2 A3 B1 B2 B3 C1 C2 C3

5010

015

020

025

0


Com

plet

ion

Tim

e (s

ec)

Mean = 57.06 Mean =

46.63Mean = 28.83

Mean = 79.22

Mean = 60.4

Mean = 41.97

Mean = 122.27

Mean = 79.95

Mean = 56.79


A1 A2 A3 B1 B2 B3 C1 C2 C3

2030

4050

6070

80


Com

plet

ion

Tim

e (s

ec)

Mean = 26.7

Mean = 21.73

Mean = 21.54

Mean = 32.53

Mean = 35.19

Mean = 51.2 Mean =

48.52

Mean = 57.49


88

A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

100

120

140


Com

plet

ion

Tim

e (s

ec)

Mean = 37.92 Mean =

31.98Mean = 31.98

Mean = 49.98

Mean = 36.71

Mean = 35.39

Mean = 82.69

Mean = 54.88

Mean = 61.66


A1 A2 A3 B1 B2 B3 C1 C2 C3

2040

6080

100

120


Com

plet

ion

Tim

e (s

ec)

Mean = 36.55

Mean = 29.04

Mean = 33.72

Mean = 56.88

Mean = 45.69

Mean = 44.07

Mean = 86.4

Mean = 66.34 Mean =

61.4


89

Appendix B

SUBJECT FEEDBACK

B.1 Introduction

The development of the RAVEN has been a collaborative effort between engineers and

surgeons. Without this relationship, engineers would not be able to design a system suited

specifically for the end user. Much of the TeleRobotic FLS effort focuses on objective

measures of performance on the surgical robot system. However, a key part of system

development and improvement is usability feedback. At the end of the experiment subjects

were asked to provide written feedback in terms of the RAVEN’s usability and how delay

effected their performance. All of the responses were submitted by email. In this chapter,

subject feedback from the experiment in Chapter 6 is included.

B.2 Feedback from Surgeons

B.2.1 Subject 10

Subject 10 is a general surgeon.

Impressions of the robot:

I thought the robot was very interesting as a starting point, huge potential.

Absolutely needs a wrist for suturing if going to be competitive with any true

surgical robot. The unintentional motion needs to be stopped, maybe with a

gyroscope or something that corrects or stops the instrument at that height?

Easy usage, very easy to get used to, I was flying with it by my second go. I

think a circular grasper or similar to daVinci finger actuator should be used

and something with gel pads is a must to allow the surgeon to know how much

grasp pressure he or she is exerting on the tissue is also a must that’s next gen.

product and should be investigated. I would love to work on that with you at

90

your next job, being a surgeon is fun, but this would be cool too! Obviously a

3d screen with feedback or teleprompter teaching screen so a proctor could write

on a screen elsewhere for teaching would be great. The instruments are basic,

so knocking those is unnecessary, you guys would have an entire assortment of

instruments, but I always wanted an instrument changer, like a modern gatlin

gun that the surgeon would just push a button and a new instrument would be

loaded down the shaft or a multi-tip instrument would rotate and that grasper

tip would reset into the applier. Anyway, lots of ways to skin that cat, those are

just two of my many ideas, can’t give you all my cool ones.

First day you came in and today:

I cleary thought I did better today, not sure why, I learned to relax today and

I was already prepared for what to expect. I also wasn’t excited about the

prospects, I was ready to go and it wasn’t new. I’m sure my brain learned while

I was sleeping too, you know the theories, but I was also away, went to Chicago

and was gone for over a week, so some extinguishing had to occur as well. Who

knows. I liked it better today regardless.

The delay:

First, the delay truly makes movements seem as if they are occurring with a

gel around the instruments. The instruments seem harder to move, they don’t

respond or seem “stuck” to put it in simplistic terms. Of course I know why they

won’t respond, but the brain won’t listen to reason. I have to consciously slow

down my movement and of course at 250 ms delay this is much less pronounced

and I can make the adjustments because I think I can predict much better where

the instrument will go when I move it in my mind after I’ve learned the robot.

That’s why I think the times improved on the delay today anyway. At 500 ms

delay, that prediction is worse because the delay impedes the time between the

movement of the action and when it occurs and micro-adjustments can’t occur,

91

or when they do it’s far too late. Just a theory. Also, once you’ve made a

movement, you can’t adjust, it’s too far or too little and you are forced to make

a series of smaller movements when you need delicate moves because errors are

more pronounced due to the delay and overshooting the target. You probably

know all this already.

What I learned was the spatial orientation and I was more comfortable with

the peg placement, you saw I just knew where the pegs were and I knew how to

move the arms quickly in space without looking around, also I learned how to

toggle to avoid the self movement high up. I just felt I was going up and toggled

immediately before I started losing control of the tips of the instruments. I also

prophylactically opened and closed quite a bit, may be more than I needed. The

back of the table is the hardest, but my strategy was to move two arms at a

time, once my block was dropped I was already moving that other arm, and that

comes from my laparoscopic experience and using two hands in surgery all the

time and being cognizant of both my hands in the operating room during the

cases. Maybe? Anyway, I think that’s what saved me some time.

I also think by taking some risks I dropped more blocks but eventually saved

time.

B.2.2 Subject 13

Subject 13 is a pediatric urologist with prior RAVEN experience and extensive clinical

experience with the ISI da Vinci

During the experiment, especially during the tasks that were not delayed,

I could sense that I reached an optimum performance time and economy of

motion where I could achieve the task successfully with minimal inappropriate

instrument tracking such that I could feel concentration centers of my brain

shutting down. As if I had developed a rote muscle memory for the tasks and

no longer needed to concentrate on moving through space. The downside is that

with this diminished sense of performance anxiety, I plateau in my performance

92

times. It was if I developed a balance between speed, economy of motion, chance

for success, and minimal brain effort.

As far as managing delay, I initially slowed my arm/hand movements down to

accommodate for the delay. This helped prevent me from overshooting targets.

However, after familiarity with the task was established, I began to accelerate

”non-critical” moves and decelerate for critical moves. For example, the speed

at which I traveled between pegs was fast because I had developed a memory to

the distance I need to move the styli in space to get from one peg to another.

But when I needed to actually grasp a block and dock a block, I slowed down

as my movements demanded more accuracy. I believe you had mentioned that

this was also observed by another subject.

B.3 Feedback from non-Surgeons

B.3.1 Subject 3

Subject 3 is a graduate level engineering student

As a first-time user of the Raven I found it easy to learn how to maneuver the

robot. The user interface was intuitive, as my hand movements corresponded

well to what I saw on the screen, and the only additional commands that had

to be dealt with were the indexing foot pedal and the gripper buttons. The self-

closing artifact of the gripper and the slight unreliability of the foot pedal were

somewhat disturbing, but certainly not to a degree where they would impact the

outcome of the experiment. I think the division of the experiment into two parts

was important, since doing the whole experiment in one go would be exhausting,

and that might very well have impacted the outcome had it not been done. The

introductory video tutorials were a bit on the long side, but at the same time

they were informative in demonstrating the tasks to be performed.

B.3.2 Subject 14

Subject 14 is a graduate level engineering student.

93

I think the experiment may have been better if a third round of training, at

500ms delay, was included. I felt fairly confident when I started the experiment

at the 0ms and 250ms delays, having already controlled the RAVEN at both of

those levels. Controlling the system at 500ms was significantly more difficult,

especially for the first trial. I spent a fair amount of time learning how the system

performed and how fast I could move without overshooting. I also dropped a

few blocks during the first experiment before I realized how important it was

to wait for visual feedback that a grasp had been completed before releasing

the first hand. If I had performed the training tasks at 500ms, I feel like my

performance during the experiment would have improved.

I also think think the experiment would have been better if the system was

able to run without crashing or breaking. It would have also been somewhat less

emotionally taxing if the C-Arms were not repeatedly initialized and E-stopped

during the experiments.

I was able to notice the difference between 0ms and 250ms delay, but it did

not seem as significant as the difference between 250ms and 500ms. This may

or may not have been a result of the training. I think an interesting extension

to this experiment would be to add additional delays and see if the relationship

between delay and performance is nonlinear, or if it is approximately linear for

values above a certain threshhold.

Other experiments that might be interesting would be the effect of alcohol,

caffeine, sleep deprivation, etc. on performance.

B.3.3 Subject 15

Subject 15 is a graduate student in the MD/PhD program. Subject 15 participated in the

Pilot Study.

In comparing the old tools with the new tools for the RAVEN TeleRobotic

FLS, I found it much easier to use the newer tools. I believe I was using the

old set of tools at the end of its usefulness. The wrists on the old tools would

94

often turn unexpectedly and it would be very difficult to reorient them. This

caused me a lot of frustration. I did not experience this phenomenon with the

new tools. The only unexpected thing that happened was that occasionally

the graspers would become closed on their own, but closing and reopening the

graspers immediately solved this problem. I was surprised that it was easier to

do the FLS tasks using tools with fewer degrees of freedom.

With regards to the time off before the last set of trials using the new tools,

I did not feel that this made a difference for me. I did not notice any particular

improvement or loss in skill during the last set of trials. If there was a difference,

it may have been more to do with fatigue rather than skill. I may have felt more

energetic during that last set.

95

Appendix C

RAVEN OPERATION TIPS

This appendix contains the “cheat sheet” that subjects where given for the experiments

in Chapter 6. These tips helped the subjects learn to telemanipulate the RAVEN as well

as provided them with insight into the Block Transfer task.

C.1 Procedural Tips

Throughout the experiment, you may refer back to these procedural tips or ask the experi-

menter to explain the task you are about to or are trying to accomplish.

C.1.1 Grasping

Grasp, then release to reopen the jaws

When the jaws are in the grasp state, the jaws are held closed by the actuator. When the

jaws are in the open state, the jaws are not actively held open. In the open state, the jaws

may open and close slightly as you move the tools around. In order to “reopen” the jaws

just before grasping an object, grasp, then release to reopen the jaws.

Visually confirm the grasp

Occasionally, the grasper will not fully close when the grasp button is pressed. In order to

minimize errors, visually confirm that you actually have a solid grasp on the object you are

grasping.

Orient the tips of the Maryland graspers so they point down

The Maryland graspers have curved jaws. Generally, it is easier to grab the FLS blocks

with the tool tips pointed down.

96

Grasp with 50% of the jaws

If you attempt to grasp with just the very tips of the jaws, the block might get dropped. If

you attempt to grasp with too much of the jaws (take too big a bite), the jaws may not be

able to close properly and the block might get dropped.

C.1.2 Tool Roll

Tool Roll Limits

The RAVEN manipulators limit tool roll to one full rotation in each direction. The Omni is

limited to 1/3 of a turn in each direction. If you hit the physical limit of the Omni, lift the

foot pedal and index as you would for positioning. If the RAVEN tool runs out of travel,

“unroll” it in the opposite direction.

Omni Kinematic Singularity

In some positions and orientations of the stylus, the Omni has self-induced motion which

will cause the surgical tool to roll on its own. Please have the experimenter explain the

kinematic singularity of the Omni input device.

C.1.3 Foot Pedal

Sticky Pedal

In order to connect the surgeon site to the patient site, the foot pedal must be pressed. When

you press the pedal you will hear the brakes disengage and, when you release the pedal, you

will hear the brakes reengage. During indexing, you release the pedal before repositioning

your hands. Sometimes the foot pedal “sticks” and releasing the pedal doesn’t actually

disengage the connection between the surgeon and patient sites. The first clue of a sticky

foot pedal is that, when you released your foot, you did not hear the brake engage; the

second clue is that, when you move your hands to index, you see the tools still moving.

In order to correct a stuck pedal condition, simply press and release the pedal once more,

listening for the brakes to disengage. ***If you have delay, the brakes will not engage or

97

disengage immediately***

C.1.4 TeleRobotic FLS Block Transfer

Optimizing your task - NO ERRORS

The results are reported in terms of number of errors and completion time. There is no

trade-off of errors for time. In surgery, errors are unacceptable. Move as efficiently as you

can, but be deliberate and minimize potential errors.

Grab left side of blocks with left tool, right side of blocks with right tool

Grasping the left side of the block with the left tool and the right side of the block with the

right tool makes the transfer step easier, because the tools don’t need to cross each other.

Don’t chase your hands around

If you are having trouble during the transfer step, keep the hand holding the block fixed

then use the other hand to approach the block and grasp it. If you are moving both hands

you will risk getting spatially lost.

Visually confirm the block is on the peg before releasing it

Depth perception in this experimental set-up is difficult. In order to reduce errors visually

confirm that the block is on the peg before releasing it.

98

Appendix D

SUGGESTED IMPROVEMENTS TO THE RAVEN

D.1 Introduction

During the pilot study (Chapter 5) seven subjects completed the training portion and three

of those finished the time delay TeleRobotic FLS Block Transfer experiment. The pilot

study, including training for subjects who were not able to complete the study, represented

approximately 50 hours of operational time for the RAVEN. Tools breaking ultimately

ended the pilot study. It was clear that a large scale experiment drawing from “RAVEN

novices” pushed a system more than we previously had. This resulted in the wrist joints of

the adapted Computer Motion micro-wrist tools fatiguing and breaking.

Over the course of the main study, 15 subjects ran the RAVEN for between 1.5-6 hours

for a total time of approximately 70 hours. During these experiments many shortcomings

of the RAVEN were exposed and will be discussed in this chapter.

D.2 Issues and Potential Solutions

D.2.1 Depth Perception

Lighting

A typical MIS environment is generally dark with a single point light source located in the

endoscope. Surgeons view the operative site on a monitor similar to that used in these

experiments. The use of shadows, focus, occlusion, and knowledge of the anatomy of the

operative site give the surgeon the ability to navigate. With the RAVEN set-up on the FLS

box trainer with no lid but a single light, some shadows were cast, but it was a generally

flat lighting condition which made depth perception much harder for both surgeons and

non-surgeons. By moving toward a closed box trainer and endoscope with light source, the

lighting conditions would be more similar to minimally invasive surgery and would make

99

depth perception easier. Further, operating through MIS ports in a closed box trainer would

reduce the effective length of the tools resulting in less tool flex and vibration.

3D Display

One of the surgeon subjects, a urologist with extensive da Vinci experience, commented

that he had become a “3D baby” in response to missing a block due to depth perception

error. Another surgeon subject, a neurosurgeon, said that he would have had a much easier

time with better depth cues and, ideally, a stereo vision system. Section 7.4.1 showed a

statistically significant improvement in completion time for the Block Transfer Task using

the ISI da Vinci with the 3D vision system enabled. A well implemented 3D vision system

for the RAVEN should enhance performance.

D.2.2 Accidental press of ungrasp button

Currently, the forward-most button on the Omni stylus is for grasping and the rear most

button is for ungrasping. Multiple dropped blocks occurred because the subjects’ finger(s)

accidentally bumped the ungrasp button, resulting in an error. In order to make this more

failsafe, make the forward most button ungrasp (less likely to get bumped). Alternately, a

better grasp interface that allows for proportional grasping could be implemented.

D.2.3 Proportional grasping

Current grasping on the RAVEN is a binary degree of freedom. The tool tips are either

closed (held closed with a constant torque command to the actuator) or are “opened” (a

torque command is given to the actuator to open the grasper for a few seconds, then no

torque command is given). When the tool tips are closed, there is a constant load on the

actuator. In order to reduce the overall load on the actuator, no torque is given in the

nominal open state. However, because of cable coupling as the tools move throughout the

workspace, the tool tips tend to open and close on their own. When moving up to an object

to grasp, one will often have to grasp, then ungrasp, in order to reopen the tool tips, then

pick up the object. Position-controlled grasping would eliminate the “grasp, then ungrasp,

100

to reopen the tools” problem, as well as result in lower power dissipation by the actuator.

This modification would require new grasper hardware and software on the Surgeon site and

modification to the software on the Patient Site (specifically in the kinematics and cable

coupling equations).

D.2.4 Pedal Up grasp state transition

When the system is in the pedal up state, the operator’s motions are not transmitted

between the Surgeon Site and Patient Site. However, a few dropped blocks occurred because,

while in pedal up, the white button on the Omni stylus (ungrasp) was pressed, then, when

the footpedal was pressed, the tool immediately ungrasped. A test was performed to verify

this phenomena. While in pedal up, the last button on the Omni stylus will be transmitted

upon transition to pedal down. A simple check of the SUI code to ignore stylus button

while in pedal up would solve this problem.

D.2.5 “Bump the encoders”

When the USB Board is powered on, the encoder chips contain a random value. Once the

encoder chips see a transition from the encoders, they become properly initialized. We have

come up with a hardware fix to this initialization problem; it is to “bump the encoders”

after power up. If one attempts to run the RAVEN without bumping the encoders after

power up, the random/garbage value is sent to the control software as the starting point.

Then as soon as the manipulator moves, the encoder chips correctly initialize and send the

“correct” value to the control software, causing a perceived large change in position and, in

response, a large current to the motors. The result is usually an e-stop in the best case but,

occasionally, the large transient in actuator response is great enough to break a cable. This

problem can be fixed by investigating the initialization of the encoder chips on the USB

board.

101

D.2.6 Data Collection

Data Collection on the RAVEN Patient Site is fairly straightforward. There are currently

114 parameters collected at 100Hz (data generated at a 1kHz, but only one sample for every

ten saved to disk for an effect zero-order hold). The large amount of data generated makes

the files somewhat cumbersome to parse down into meaningful subsections of experiments.

A useful improvement to the Toolkit would be a button to auto-increment the data file name

and start a new data collection file. Thus, rather than a single or a few relatively large files

from a robot run, one could have many shorter files that would be easier to manipulate and

associate meaning to (for example, one file for each of the training task repetitions).

Currently, pedal state is the only way to know if there was something happening when

performing post-hoc analysis of the data. For experiments where there is an experimenter at

the Patient Site, an “experimenter input” to the USB Board(s) would allow the experimenter

to encode more information into the data stream. A generic multi-contact socket could be

interfaced to each controller box. A generic experimenter box with switches could be the

simplest level of encoding points of interest into the data. The input could be tied to the

task board for a more automated encoding of experiment information into the data stream.

D.2.7 Sticky Footpedal

The footpedal that enables clutching of the operator’s hands to the RAVEN’s tool tips is a

simple USB footpedal designed for gaming. The Surgical User Interface software is set up to

monitor key presses for transitioning between “pedal up” and “pedal down” states. Pressing

the right-most pedal down is mapped to the “d” key and releasing the pedal is mapped to

the “e” key. However, sometimes the footpedal or software associated with the pedal misses

that transition. The result is that, when a user lifts his foot off the pedal, the system is

still in the pedal down state. Some of the subjects learned to work around the sticky pedal

problem, while others struggled. In the worst cases subjects who were attempting to clutch

to reposition their hands, let off the footpedal and moved their hands before verifying the

state transition. Under a sticky pedal condition the (usually) large and fast motions of the

hands caused and emergency stop of the RAVEN. In order to overcome this condition the

102

software should be modified to not just monitor the transition of the footpedal but it’s state

(up or down).

D.2.8 Omni Kinematic Singularity and Tool Roll

The Omni contains a kinematic singularity within its workspace that results in a self-induced

motion which causes the surgical tool to roll on its own. When moving the stylus laterally

there is a coupling between the stylus and the proximal joints which induces a roll of the

RAVEN tool. If this coupling were accounted for on the SUI end, it would make the user

interaction more intuitive.

D.2.9 RTAI Stability/USB Driver

The RTAI Linux PC that runs Patient Site is much more stable than it was initially. However

in a full day of operation, the USB driver seems to freeze and cannot be unloaded. The

result is either a soft or hard restart of the PC. This is an issue that should be investigated

further.

103

VITA

Mitchell Jay Hiroshi Lum received his Bachelor of Science in Electrical Engineering from

the University of Washington in 2002. His experience in undergraduate research with the

BioRobotics Lab led him to remain at the University of Washington for his graduate studies.

In 2004, he received his Master of Science in Electrical Engineering. He most recently

completed his Doctorate in Electrical Engineering also from the University of Washington.

quantitative performance assessment of surgical robot...

Documents