Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A foundation for the design and assessment of improved instruments for minimally invasive surgery Person, John Gunnar 2000

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2000-0257.pdf [ 18.14MB ]
Metadata
JSON: 831-1.0080974.json
JSON-LD: 831-1.0080974-ld.json
RDF/XML (Pretty): 831-1.0080974-rdf.xml
RDF/JSON: 831-1.0080974-rdf.json
Turtle: 831-1.0080974-turtle.txt
N-Triples: 831-1.0080974-rdf-ntriples.txt
Original Record: 831-1.0080974-source.json
Full Text
831-1.0080974-fulltext.txt
Citation
831-1.0080974.ris

Full Text

A FOUNDATION FOR THE DESIGN AND ASSESSMENT OF IMPROVED INSTRUMENTS FOR MINIMALLY INVASIVE SURGERY by JOHN GUNNAR PERSON B.Sc, The University of Manitoba, 1997 A THESIS SUMITTED IN PARTIAL FULFILMENT OF T H E REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Mechanical Engineering) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 2000 © John Gunnar Person, 2000 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of &UHklMirfiL Bfi/GmGRfi-LU^ The University of British Columbia Vancouver, Canada Date APL 2 & 2CQO DE-6 (2/88) Abstract The goals of this project are to establish a foundation for research into the design and assessment of improved instruments for minimally invasive surgery (MIS). Minimally invasive surgery has exploded into general surgical practice in the last decade with the introduction of the laparoscopic cholecystectomy. Though beneficial to patients, the new technique is complicated by physical and mental challenges which make it difficult for surgeons to master. Improved instrumentation and teleoperators have the potential to ease the strain of performing the technique and improve surgeon performance. An experiment on the effects of physical constraints on surgeon performance is presented, in this report suggests that an economical and effective solution is to develop a teleoperator device constrained to 4 degrees of freedom which restores the natural motion mapping between the hand and instrument tip across the fulcrum point in the abdomen wall. A prototype 4 degree of freedom, mechanical teleoperator is constructed to'test this theory. The goal in developing this device is to map motion from the surgeon controlled master handle to the slave instrument tip within the surgical site, using only mechanical force transmission. Such a device has the potential to offer an economic solution to performance problems currently seen in MIS. Pitch and yaw motions are mapped between master and slave by direct mechanical linkage. Fundamental problems with a novel friction drive mechanism, however, prevent the prototype from functioning as anticipated, leaving the project open to future improvements. In order to effectively test for real improvements in surgeon performance, we also proposed a motion analysis system for measuring surgeon performance during actual surgeries. This system will be employed in future studies to validate surgical simulations in the laboratory which may be used to test and develop new instruments. A pilot study is performed to assess the potential of using the system to track the unconstrained motion of the surgeon's dominant arm over the course of a laparoscopic cholecystectomy. An ergonomic posture sampling study is performed to demonstrate the usefulness of the motion analysis system in this application. ii Table of Contents Abstract Table of Contents List of Tables .' v. List of Figures A vi. Preface viii . Acknowledgements :.. ix. Chapter 1 Introduction and Literature Review 1 1.1 Current Status Of Minimally Invasive Surgery (MIS) 1 1.2 Current Solutions to MIS Limitations 4 1.2.1 Robotics and Teleoperation 4 1.2.2 Improved Vision 6 1.2.3 Tool Design 7 1.2.4 Training 9 1.2.5 Assessing MIS Surgical Skills 11 1.2.6 Ergonomics in MIS , 12 1.3 Research Questions , 14 Chapter 2 The Effects of Physical Constraints in Laparoscopic Surgery 15 .2.1 Introduction :.. : 15 2.2 Methods 16 2.3 Results 20 2.3.1 Analysis of the Suturing Task 21 2.3.2 Analysis of the Pick-&-place Task 22 2.4 Discussion 23 2.5 Conclusions • • 27 Chapter 3 A 4 Degree of Freedom Mechanical Teleoperator for Minimally Invasive Surgery... 28 3.1 Introduction 28 3.2 Device '. ; 29 3.2.1 Physical prototype : 35 3.3 Discussion 37 3.4 Conclusion 38 Chapter 4 A Motion Analysis System for Ergonomic and Performance Studies in the OR: A Pilot Study of Ergonomic Posture Sampling During a Laparoscopic Cholecystectomy 39 4.1 Introduction 39 • 4.2 Materials and Methods 40 4.2.1 Motion analysis system and OR setup 40 4.2.2 Polaris Optical Tracking System and Marker Arrays ,42 4.2.3 Ergonomic posture measures , 44 4.2.4 Kinematic model of the upper limb : 46 4.2.5 Kinematic joint centre calibration 47 iii 4.2.6 Joint angles • 50 4.2.7 Validating posture angle measurements 56 4.2.8 Study protocol .- 57 4.3 Results 58 4.3.1 Dealing with corrupted data 61 4.4 Discussion 67 4.5 Conclusions 70 Chapter 5 Conclusions and Future Work 71 5.1 Introduction 71 5.2 Review of present research 71 5.2.1 Effects of physical constraints '. 71 5.2.2 Mechanical teleoperator 72 5.2.3 Motion analysis study 72 5.3 Future research recommendations 73 5.3.1 Mechanical teleoperator 73 5.3.2 Motion analysis system 73 5.4 Future studies: Performance measures and simulation validation 74 5.4.1 Simulation validation 74 5.4.2 Performance and correspondence measures 75 Bibliography 77 -Appendix A Analysis of physical constraints task time data 86 Appendix B Teleoperator shop drawings 95 Appendix C Validation of circle and sphere fitting methods 123 Appendix D Clinical joint angles 132 Appendix E Nonparametric analyses and task time annotations 139 Appendix F Coorinate transforms and nomenclature 145 iv List of Tables Table 1: Tukey confidence, interval analysis for suturing task 22 Table 2: Tukey confidence interval analysis for pick-&-place task... 23 Table 3: R U L A (Rapid Upper Limb Assessment) posture classifications 45 Table 4: Modified R U L A posture scores adopted for current study 46 Table 5: Surgical steps, endpoints, and percentage of overall surgery time for operative tasks examined in posture sampling study 58 Table 6: Breakdown of surgical steps, clock times, and percentage of overall surgery time for operative tasks from the laparoscopic cholecystectomy studied during the motion analysis. .' 59 Table 7: Sample block of joint angle from experimental data illustrating the definitions of missing data, partial data^  and complete data 62 Table 8: Percent missing data during tool change/error tasks vs. normal surgical tasks for cystic duct dissection (CDD) task 62 Table 9: Annotated frequency of data status within each minute of recorded surgery 64 Table 10: Percent frequency of duration for partially corrupted samples in each joint angle from data collected during motion analysis 66 List of Figures Figure 1. Illustration of the minimally invasive surgery (MIS) technique .2 Figure 2. Profile of a pistol-grip MIS instrument for laparoscopic surgery 3 Figure 3. Reduced DOF of motion of the MIS tool tip within the abdomen 3 Figure 4. Relationship between real teleoperator and our emulation of a perfectly transparent teleoperator using constrained open tools 17 Figure 5. Experimental setup emulating transparent teleoperators using constrained open and laparoscopic tools 18 Figure 6. Experimental apparatus 19 Figure 7: Mean task completion time for suturing and pick-&-place experiments 21 Figure 8. Proportion of time spent on three important tasks in laparoscopic surgery under normal, direct vision, and 40 teleoperator & direct vision conditions 25 Figure 9. Percentage of total surgery time for the five tool configurations 26 Figure 10. Isometric view of the UBC Mechanical Teleoperator 30 Figure 11. Illustration of the range of motion of the UBC mechanical teleoperator 31 Figure 12: Conceptual model of a 4 DOF teleoperator by Drown et al. (1998) 32 Figure 13. Illustration of the principle of a friction drive 33 Figure 14: Illustration of a continuously variable transmission (CVT) employed by a cobotic device 33 Figure 15. Illustration of the friction drive function "... 34 Figure 16. Prototype of the mechanical teleoperator with friction drive 36' Figure 17: Close-up of the friction drive mechanism 36 Figure 18: Conceptual illustration of the surgeon handle interface 37 Figure 19: Components and operating room layout of the Polaris Hybrid Optical Tracking System and video recording equipment used in this study 41 Figure 20: Polaris Hybrid Optical Tracking System 42 vi Figure 21: Polaris marker arrays 43 Figure 22: Surgeon arm model 47 Figure 23: Illustration of a cloud of data points describing a 90° arc of a circle 49 Figure 24: Definitions of wrist joint angles '. 53 Figure 25: Clinical joint angles for flexion-extension and radial-ulnar 54 Figure 26: Forearm neutral position and rotation definitions. 55 Figure 27: Example of visual joint angle validation for the wrist 56 Figure 28: Study protocol 57 Figure 29: Cystic duct dissection (CDD) stress levels for each joint angle 60 Figure 30: Gallbladder dissection (GBD) stress levels for each joint angle 60 Figure 31: Gallbladder removal (GBR) stress levels for each joint angle 61 Figure 32: Breakdown of data status showing complete, partial and missing 63 Figure 33: Frame-by-frame data status distribution '. 65 Figure 34: Status of each joint angle sample over the surgery duration 66 Figure 35: Normalized, weighted postural stress contribution of each joint outside of normal, ergonomically safe posture range 67 Figure 36: Time evolution of average normalized stress 68 vii Preface Chapter two of this thesis, "The Effects of Physical Constraints in Laparoscopic Surgery" was previously published as a journal article under the following title: A.J. Hodgson, J.G. Person, S.E. Salcudean & A.G. Nagy (1999) The Effects of Physical Constraints in Laparoscopic Surgery, Medical Imaging Analysis, 3(3), 275-283. As the second author, I contributed to the work by conducting the experiment described in chapter two, preparing the statistical analysis of the data, prepared selected figures and tables for the paper, and co-authored the paper with Dr. Antony Hodgson. Dr. Salcudean, (Department of Electrical Engineering, University of British Columbia) was involved in preliminary work on the project and a pilot study with Dr. Hodgson. Dr. Nagy (Division of General Surgery, University of British Columbia) is a specialist in laparoscopic surgery and acted as the surgical consultant for the project. Confirmation: I am in agreement that the contributions of the thesis author are as stated. Dr. Antony J. Hodgson Department of Mechanical Engineering The University of British Columbia viii Acknowledgements Chapter Two Thanks to the surgeons from the Section of Laparoscopic Surgery, Division of General Surgery, Vancouver Hospital and Health Sciences Centre, and from Burnaby General Hospital who participated in our experiments. We also thank Siobhann Williamson of Karl Stortz Endoscopy Canada Ltd. for loan of the instruments used in this study, and Mana Shirazi-Kia for her assistance in analyzing the videotapes of the experiments. This work is supported by the Canadian N C E IRIS project IS-8. Chapter Three My thanks to the Benny Nimmervol at the UBC Mechanical Engineering machine shop for performing the primary machine work on the teleoperator. Special thanks to Doug Yuen and Dave Camp for their additional assistance on the project. Chapter Four ' My thanks Jeff Stanley, Dave Ristau, and others from Northern Digital Inc. (Waterloo, ON) for their assistance in preparing the Polaris motion analysis system for this experiment. Special thanks to Dr. Kellogg Booth in the UBC Department of Computer Science for lending us the video mixing equipment for this experiment. Thanks to the staff of the Mechanical Engineering machine shop, especially Doug Yuen, for all their help preparing the marker arrays and tools for the experiment. I am also grateful for the assistance of Mona Charman, Marlene Purvey, and all the nurses and staff of the UBC and V G H hospital sites. Thanks also to Barb at Velcro Canada Inc. for the Vel-stretch samples, and Viola Hoo at 3M for the reflective tape samples, and Christine Jansen for preparing the active marker sling and surgeon drawing. I would like to extend special thanks to my advisor, Dr. Antony Hodgson, for his insight and enthusiasm throughout the project. Thanks to Dr. Nagy for his wisdom and insight into the challenging world of minimally invasive surgery. My thanks to Willem Atsma for his assistance compiling this document, and to the other pioneers of Neuromotor Control Lab, Kevin Inkpen and Rich Emrich, for their camaraderie and support over the past years. Special thanks to my wife, Christine, for her love, understanding and support. This work is supported by the Natural Sciences and Engineering Reseach Council of Canada (NSERC) and the Society of American Gastrointestinal and Endoscopic Surgeons (SAGES). Special funding provided by the Flying Flossie scholarship fund. This work is dedicated to my parents for their encouragement in all the interests I have pursued. It has been a learning experience. 1 Chapter 1 Introduction and Literature Review 1.1 Current Status Of Minimally Invasive Surgery (MIS) The interest in minimally invasive procedures, particularly laparoscopic surgery, has increased dramatically over the last decade. The adoption of the minimally invasive surgery (MIS) technique into mainstream general surgery stems from the limitations of conventional open surgery, so it is instructive to compare the two. In conventional open surgery, the surgeon typically exposes the surgical site by making a large incision through the skin with conventional tools such as a scalpel and forceps. The open surgery tools are as easy and intuitive for the surgeon to use as a pen or pencil, so open surgery is typically quite straightforward for a skilled surgeon with a keen knowledge of anatomy and physiology. For the patient, however, the results of open surgery are often more traumatic than the problem the surgical intervention was meant to cure. During an open cholecystectomy (gallbladder removal), for example, the surgeon must create an incision in the abdomen which is typically up to -15-20 cm long to remove an object the size of a kiwi fruit. Though the entire procedure takes less than 60 minutes, the patient may remain in the hospital for weeks recovering from wound trauma associated with the incision, which, when it does heal, leaves a permanent scar. In minimally invasive surgery (MIS), however, the advantage goes to the patent, while the surgeon is left at a disadvantage. Also called 'endoscopic', 'keyhole', or 'minimal access surgery', MIS is a blanket term used to describe any type of surgery where rigid tools, about 30 cm long and -7-15 mm in diameter, are inserted through small ports in the skin and used to manipulate tissue during the operation. The surgeon views the surgical site by inflating the area with CO2 gas and inserting a long, rigid glass lens into the skin, along with MIS surgical tools. A camera attached to the lens allows the surgeon to view the entire procedure on a video monitor placed along side the operating table. Though there are several types of MIS, the most common is the laparoscopic technique (MIS performed within the abdomen) illustrated in Figure 1. Other branches of MIS include thorascopic (MIS in the chest cavity or thorax) and arthoscopic surgery (MIS on joints). Figure 1. Illustration of the minimally invasive surgery (MIS) technique. This example shows laparoscopic surgery, MIS performed in the abdomen. (Left) Long, rigid surgical tools and a rigid glass lens attached to a video camera are inserted into the abdomen, which is inflated with CO2 gas. (Right) The OR layout during a laparoscopic procedure where the assistant (far r.) operates the camera while the surgeon (2nd from r.) manipulates the laparoscopic tools and views the operation on a video monitor. Source: www.laparoscopy.com The advantages to the patient undergoing a MIS procedure rather than a conventional open procedure are (1) minimal scarring, (2) reduced hospital stays, and (3) reduced postoperative pain. In the case of the cholecystectomy operation mentioned previously, patients have been released from the hospital as early as the day following the laparoscopic procedure with a few abdominal scars no larger than the navel. Due to the massive popularity of MIS techniques, this whole area of general surgery "has been patient driven and industry sponsored" (Tendick 1995), an unprecedented event in the history of general surgery. As a result, surgeons had to learn and adapt to the new techniques quickly, and use inferior MIS tools which were hastily adapted from their open counterparts. Surgeons must deal with a number of shortcomings of the MIS techniques. One disadvantage noted regularly is that surgeons lose the direct, 3D view they are accustomed to in open surgery. Due to the nature of MIS surgery, the surgical site is viewed on a 2D video monitor, resulting in a limited, monocular view of the anatomy (Tendick 1995) The tools for MIS are also fraught with both intrinsic and extrinsic problems. Extrinsic problems are caused by the functional requirements of the tools, which must allow the surgeon to manipulate a small grasping jaw within the body using a handle outside the body, approximately 30 cm from the tool tip. Figure 2 illustrates one type of MIS instrument used for laparoscopic surgery. Several tip and handle configurations are available for these type of instruments, including handles in-line with the tool axis, pistol grips as shown in Figure 2 below, and scissor grips. 3 Figure 2. Profile of a pistol-grip MIS instrument for laparoscopic surgery. Several handle and tip types are available from a number of manufacturers. Source: Karl Storz Endoscopy The extrinsic problems with the design of a MIS tool are: (1) the surgeon's hands are remote from the surgery site, thereby reducing tactile feedback, (2) the number of degrees of freedom (DOF) of tool motion are reduced, (3) there is a fulcrum effect between the tool handle and tip, and (4) tool handle-to-tip relative motions are scaled across the fulcrum. The fulcrum effect through the entry port into the body is particularly confounding since the handle must be moved in the opposite direction from the desired motion of the tool tip. A simple illustration of these effects is to put a pencil (the MIS tool) through a hole ('port') in a piece of paper (the skin wall) and attempt to write one's name on another sheet while holding the 'port' paper fixed. One will note not only the motion reversal and remote manipulation effects, but also that the number of ways in which the pencil can move has been reduced. In regular manipulation to position the tip, the writer can move the pencil in three translational (X, Y, Z) and three rotational (roll, pitch, yaw) degrees of freedom relative to the axis of the pencil itself. When the pencil is constrained by the hole in the paper (or the MIS tool by the port), the motion of the tip is constrained to pitch, yaw, roll, and plunge (Z) through the entry port fulcrum. Two translational DOF are lost in the plane of the entry port. The number of degrees of freedom of the MIS tool tip are illustrated in Figure 3 below. Figure 3. Reduced DOF of motion of the MIS tool tip within the abdomen. DOF are roll, pitch, and yaw about the fulcrum created by the. entry portal and plunge through the port. Source: (Tendick 1995) 4 Intrinsic problems associated with MIS tools include the reported loss of mechanical force transmission from tip to handle (Gupta 1996). Cases of hand numbness and pain have been reported by MIS surgeons which has been attributed to the intrinsic design of the instrument handle (Neuhaus 1997, Horgan 1997). This problem is particularly acute in ring handled instruments, where the surgeon forces their thumb into the ring, thus constraining themselves to adopt awkward and uncomfortable positions during surgery (Neuhaus 1997, Horgan 1997). 1.2 Current Solutions to MIS Limitations The recent introduction of MIS to the field of general surgery, and the subsequent problems associated with surgeon vision and motor constraints, have prompted many researchers to examine the problems with the surgery and how to improve the surgeon performance during MIS procedures. Early on, unacceptably high rates of complications were associated with the new MIS procedures, and research began to focus on improving training and accrediting for laparoscopic surgeons (Hunter 1997, Pellegrini 1997, Way 1995). Simultaneously, research began into developing improved vision systems to restore depth perception (Tendick 1995), and new tools with force-feedback (Payandeh 1996, van Hemert tot Dingshof 1996, Herder 1997) and articulating end-effectors (Faraz 1997, Dautzenberg 1995, Mukherjee 1996) to restore dexterity in surgery. Shortly thereafter, research commenced into the highly complex technology involved in creating a surgeon-controlled, robot teleoperation system which incorporated the emerging tool and vision improvements (Taylor 1995, Sackier 1994, Finlay 1995, Neisius 1995, Rovetta 1996, Payandeh 1996, Ohgami 1998, Ottensmeyer 1996, Green 1995). In this section, we examine the main fields of research focused on improving surgeon performance in MIS: (1) robotics and teleoperation, (2) improving vision, (3) training and assessment, and (4) tool design. Most of the research is focused on laparoscopic surgery, the branch of MIS performed in the abdomen, therefore the present discussion will focus mainly on that branch. 1.2.1 Robotics and Teleoperation The most technically advanced solution to improving surgeon performance in MIS is the development of surgical robots and teleoperators. Teleoperation, or telemanipulation, is "the extension of a person's sensing and manipulation capability to a remote location" (Tendick 1995). Teleoperation technology has been used in the past for nuclear material processing, undersea work, and space exploration and has recently been the focus of applied research for laparoscopy (Tendick 1995). One group, lead by Russell Taylor at the IBM Thomas J. Watson Research Center in cooperation with the John Hopkins University School of Medicine, has been developing a robotic assistant to augment laparoscopic surgery since 1992 (Taylor 1992). Their research encompasses the design of a redundant, eight-axis Parallel Linkage Remote Center of Motion (PLRCM) robotic manipulator for laparoscopic surgery (Eldridge 1996, Funda 1994), and the accompanying constrained Cartesian motion control algorithm for optimal motion control of the manipulator within the abdominal cavity (Funda 1996). The initial generation of the robot system, called LARS, has been applied as a laparoscopic camera holder during in vivo " 5 clinical trials for both cholecystectomies (gall bladder removals) and nephrectomies (kidney removals) using a small joystick device mounted to the surgical instrument as the controller (Taylor 1996). An interesting innovation of the LARS system is the ability for the surgeon to designate anatomical features by pointing at them. The system will then determine the 3-D position of the feature and move to aim the camera appropriately (Taylor 1995). A system similar to the LARS system has been developed commercially by Computer Motion Inc. of Goleta, California based on the work lead by Y. Wang (Sackier 1994). The AESOP (Automated Endoscope System for Optimal Positioning) is a compact robotic arm designed to hold and position the laparoscopic camera in response to surgeon executed commands (Sackier 1994). The original concept used foot and hand activated control pads to move the AESOP manipulator. The current generation of the manipulator, the AESOP 3000, responds to voice commands given by the surgeon ['AESOP 3000': Brochure, Computer Motion Inc.]. The AESOP robot has been adapted to the clinical setting, and results from reported studies have found the device to be an effective assistant for laparoscopic surgery (Hoening 1997). Computer Motion is currently developing the AESOP arm into a remote access telemanipulator named Zeus (source: www.computermotion.com). The Zeus system allows the surgeon to manipulate MIS tool handles attached to master arms and will scale down and filter the manipulations to produce smoothed motion of robotically controlled instrument tips within the surgical lumen. The Zeus system was recently used to perform the first minimal access, robotically assisted coronary artery bypass in North America (source: www.computermotion.com). In a controlled study, however, surgeons were found to perform simulated surgical tasks better with manual laparoscopic tools than with the Zeus system, with the exception of tying small sutures (Garcia-Ruiz 1998). The authors speculate that performance using the robot will improve with practice and improved robot interfaces (Garcia-Ruiz 1998). EndoSista, developed in Beaconsfield, England by Armstrong Projects Limited, is another commercially available system which is functionally similar to the AESOP system. The EndoSista is a four degree of freedom, surgeon-controlled laparoscope manipulator which senses head gestures of the surgeon and translates these into image-based movements of the laparoscopic camera (Finlay 1995). Several other academic research groups are interested in developing teleoperation systems for laparoscopic surgery. The Human Machine Systems Lab at the Massachusetts Institute of Technology developed a platform for telesurgery which used laparoscopic surgery as a test model for telerobotic surgery with stable force feedback under time delay constraints (Hu 1996, Ottensmeyer 1996). Alberto Rovetta and associates from Milan, Italy developed a telerobotic manipulator for laparoscopy based on the IBM Scara 7565 (Rovetta 1996). The focus of their research deals with the complications of teleoperation, and the successful trial of a Trans-Atlantic satellite link for laparoscopic telesurgery (Rovetta 1995, Rovetta 1996). Payandeh and associates at Simon Fraser University have developed several stand-alone components for a laparoscopic telesurgery system and are currently in the process of designing an integrated system for a master/slave telemanipulator (Shell 1996). Neisius and associates in Karlsruhe, Germany are also currently working on a telemanipulator for laparoscopic surgery based on a six DOF dexterous endoscopic instrument they had developed previously (Dautzenberg 1995, Neisius 1995). A group in Japan is also developing a teleoperation system for laparoscopic surgery with haptic force feedback (Ohgami 1998). The four stages of development include an LED motion analysis, design of a master-slave manipulator, development of a system for tactile sensation, and future work on a telepresence device for laparoscopic surgery (Ohgami 1998). 6 Surely the most ambitious project undertaken in the field of telepresence and teleoperation in laparoscopic operative surgery is that of SRI International for the United States military (Green 1995, Satava 1994a, Satava 1994b). The telepresence surgery system developed by SRI integrates vision, hearing, and manipulation in a teleoperation system which is designed to allow the military surgeon to assist casualties in the battlefield while themselves remaining several hundred kilometers away from the frontline (Green 1995). The system consists of two main components: the master input tool handles and viewing platform in the surgeon's console, and a robotic surgical unit located at the surgical table (Green 1995). To date, telepresence surgery has been demonstrated using the SRI system over a distance of 160 meters using direct links (Green 1995). The proliferation of research into robotics applied to minimally invasive surgery has prompted the Society for Minimally Invasive Therapy (SMIT) to form a task-force, SMITROB, to coordinate medical and engineering research efforts in the fields of robotics and microsystems in minimally invasive therapy (Buess 1997, Schurr 1997). Currently, robotic camera and instrument holding arm show the clearest advantage of robotic intervention in minimally invasive surgery (Dunlap 1998), though the articles reviewed in this section indicate that surgeon controlled, instrument positioning robots are improving rapidly. Clearly, considerable research and development has been focused on the areas of telerobotics and telepresence surgery, in both the academic and commercial arenas. 1.2.2 Improved Vision As the name "laparoscopic" suggests, minimally invasive surgery is highly dependent on the video endoscopes used to visualize the surgical site within the abdominal cavity. Thus, it is hardly surprising to find that considerable research has been devoted to improving the camera and video system employed in laparoscopic surgical procedures. As the focus of this thesis is not in the area of vision, only a brief outline of the topic is presented in this section. Four important cues that humans normally rely upon to gauge depth are motion, pictorial (e.g. edges, contrast), monocular (e.g. focus), and binocular cues (Reinhardt-Rutland 1995). In MIS, most of the depth perception cues can be restored by improved image clarity and resolution, though monocular cues remain restricted (Patkin 1995, Reinhardt-Rutland 1995, Tendick 1996). The remaining cue that significantly effects depth perception in MIS is the loss of binocular, or stereoscopic, vision which people normally experience as a disparity in retinal images due to eye separation (Reinhardt-Rutland 1995). The loss of binocular and monocular depth cues forces surgeons to "grope forward and backward with instruments to gauge the relative depths of objects by touching them" (Tendick 1996). Significant improvements in task performance have been demonstrated using 3-D laparoscopic systems over the conventional 2-D scopes, though the clear advantage is seen when complex tasks, such as suturing, are attempted (Tendick 1996, van Bergen 1998). Although several manufacturers have developed stereoscopic (three-dimensional) laparoscopes in an effort to restore stereoscopic vision during laparoscopic surgical procedures, van Bergen found that only one-third of surgeons would prefer to work with 3-D systems (van Bergen 1998). In recent study, Taffinder and associates repeated previous attempts to show improvement using 3D scopes with a second generation system that uses a smaller screen closer to the surgeon. They did find a significant improvement in performance with both novices performing grasping and cutting tasks, and surgeons performing suturing tasks (Taffinder 1999). With video technology constantly improving, surgeons may one day experience MIS viewing with full restoration of visual depth cues. 1.2.3 Tool Design The third major area of interest in MIS technology is the design of tools with force feedback, articulating end effectors, improved handles, or other innovations to assist in surgery. In most cases, laparoscopic tool and end-effector design are closely coupled to the design of robotic telemanipulators. As mentioned in section 1.2.1 Robotics and Teleoperation, Payandeh and associates from Simon Fraser University are developing stand alone tools and devices which will eventually be incorporated into a surgical telemanipulator (Shell 1996). One such design is a thumb controlled dexterous arm designed to articulate within the abdominal cavity, thus providing two additional DOF to the end-effector (Shell 1996). Optimal design of the arm was based on work to formulate the dexterous workspace for laparoscopic extenders with articulating stems (Faraz 1997). Dautzenberg and associates have also developed a dexterous instrument for the University of Tubingen-Karlsruhe teleoperation system mentioned in section 1.2.1 Robotics and Teleoperation. The powered, dexterous instrument which they have developed boasts a quick-change, modular design which is adapted for telemanipulator control and may be easily cleaned and sterilized (Dautzenberg 1995). An Articulated Manipulator for Minimally Invasive Surgery (AMMIS) has also been developed at the US Naval Postgraduate school by Mukherjee and Song in cooperation with Dr. R. Satava (Mukherjee 1996). The geared manipulating end can be designed to bend 180° or more, can be miniaturized to fit a 5mm cannula port, and is ideally suited for surgical telemanipulator applications (Mukherjee 1996). There have also been innovations in end-effector technology to improve standard, hand-held MIS instruments. Curved instruments were introduced commercially to increase the working space within the abdominal cavity, but they did not actually increase the number of degrees of internal freedom of the instrument tip (Melzer 1997). Melzer and associates have developed a flexible end-effector to increase the internal DOF of hand held MIS instruments which they call the Deflectable Endoscopic Instruments System - DENIS (EPflex, Dettingen, Germany) (Melzer 1997). Two wheels on the tool handle operate +/-1200 variable axial deflection of the instrument tip, and full axial rotation, respectively (Melzer 1997). In a different vein of end-effector innovation, Frank and Cuschieri developed a prehensile MIS grasper which essentially allows the instruments jaws to articulate like a "thumb and finger" pair to capture tissue, rather than pinch it (Frank 1997). Balazs and associates also developed an alternative to the standard mechanical articular joint by replacing the joint with elastic beams to improve grip and responsiveness of the instrument (Balazs 1998). On the opposite end of the instrument, researchers are also working to improve the usability of MIS instruments by redesigning the handle. A number of handle'designs are currently available, including pistol grip handles with a flat shank or rings for the thumb and fingers, and axial or in-line handles with trigger or syringe like rings (Matern 1999). All handles use some kind of trigger mechanism to activate the jaws at the instrument tip. Surgeons have found various problems with the instruments which lead to fatigue and numbness in the hands and arms, especially with ring handled instruments (Neuhaus 1997, Horgan 1997). In an effort to improve surgeon comfort and instrument usability, several improved handle designs have been proposed. Mueller suggests that the instrument can be customized by adding finger rings to the handle, but this system constrains the fingers and sometimes leads the surgeon to adopt awkward and 8 uncomfortable postures (Mueller 1993). Taking greater consideration for surgeon comfort, Amaral and Levine developed a dual position MIS instrument handle which can be rotated to accommodate a pistol-grip or an in-line (axial) orientation (Amaral 1994). The ability to pivot the handle allows the surgeon to chose the most appropriate orientation for a particular task, thus reducing hand fatigue and improving control (Amaral 1994). Hasson presents an axial handle with a trigger mechanism similar to a spring loaded syringe, which allows the surgeon to open the instrument jaws by pulling a sliding ring backwards between two fingers towards the thumb (Hasson 1993). The advantage of this design is that the surgeon is free to choose their hand orientation when activating the tool, and it may be used as a probe since the jaws remained closed when not activated (Hasson 1993). Others have worked to completely rethink designs to make the more handles accommodating. A group from Dundee, Scotland, including Emam, Frank, Hanna, Stockham and Cuschieri, have developed a novel spring activated rocker handle for a needle driver which they have found improves quality of task performance over conventional tools during simulated MIS tasks (Emam 1999). In their design, a movable arm pivots in a fixed shank held against the palm such that the index and middle finger close the instrument jaws while the two smaller fingers open the jaws (Emam 1999). Matern, Waller, and associates developed 14 function and comfort criterion in order to design novel, ergonomically correct handles for both a multifunction instrument and a variable task instrument (Matern 1999). Currently, their experiments comparing the new design against conventional handles show no clear advantage to any handle type, but show promise for the multifunction handle design (Matern 1999). Another important area in MIS tool design is the development of force feedback technology. Recent research has shown that the force produced at the handle of a standard laparoscopic grasper is approximately twice that of the resultant force seen at the tip while the opposite is true for open instruments (Gupta 1996). The implication of this is that the laparoscopic surgeon loses the tactile sensation normally afforded by open instruments. Another study of commercially available laparoscopic instruments revealed that variations in transmission force and high friction losses in existing instruments prevent surgeons from sensing the pressure applied to grasped tissues (Sjoerdsma 1997). In order to contend with the disparity in force perception, many researchers are currently engaged in development of haptic or force-reflecting devices to return tactile sensations to the surgeon, particularly in magnifying mechanical force at the instrument tip. Payandeh and colleagues again have a hand in the development of force reflecting devices. A micromachined pressure sensor has been developed and incorporated into a laparoscopic grasper (Shell 1996, Unpublished demonstration at IRIS/PRECARN June 15-17, 1998). The device will eventually give surgeons variable pressure signals through the grasper handle depending on the type of tissue being manipulated (Shell 1996). A group from the Electrical Engineering Department at the Delft University of Technology in The Netherlands is also developing a force reflecting system for surgical telemanipulation (van Hemert tot Dingshof 1996). In this reflective system, the slave jaws are position controlled by input from the master instrument handle, and the master is force controlled by input from the slave (van Hemert tot Dingshof 1996). Another group from the Laboratory of Measurement and Control also at the Delft University of Technology has developed a mechanical force feedback device for hand-held laparoscopic tools as an alternative to the electro-mechanical devices (Herder 1997). A group from Harvard University is more interested in restoring tactile sense than amplifying mechanical force, and has developed a system which uses a capacitance tactile sensor and a display grid of small pins (Howe 1995). When the sensor is activated by pressing it against a small object, the pins in the display grid press against the fingers to reveal a shaped relief of the sensed object (Howe 1995). As a final note on innovations to improve tool functionality of MIS instruments, there are researchers currently working on special purpose instruments dedicated to specific operative tasks or functions. Wallwiener and colleagues have presented a multifunctional instrument which has been shown to reduce operative time by up to 20% by minimizing the number of tool changes during a MIS procedure (Wallwiener 1995). The instrument functions include pneumatically retractable cutting electrodes, mechanically operated coagulation forceps, and irrigation and suction, as wellas provide a channel for another conventional 5mm MIS instrument (Wallwiener 1995). Others, such as Nagai and Araki (Nagai 1999), Swain (Swain 1997), and Payandeh (Shell 1996), are developing instruments dedicated to MIS suturing. An MIS suturing instrument called Endo-Stitch is commercially available from U.S. Surgical Corp. (Norwalk, CT). 1.2.4 T r a i n i n g Training for laparoscopic surgery is another area presently under study in academic and commercial research. With the rapid adoption of the laparoscopic cholecystectomy procedure between 1988 and 1991 as the new standard for gallbladder removal surgery, many surgeons trained in brief 1-2 day courses to learn the new techniques (Hunter 1997, Pellegrini 1997). Simultaneously, a sharp increase in major bile-duct injuries was observed during laparoscopic procedures which was later correlated to inadequate training and lack of experience in laparoscopic techniques (Hunter 1997, Pellegrini 1997, Way 1995). The results of these findings, compiled in a review by Pellegrini and Sinanan, is that "numerous organizations and individuals have proposed or ratified a structured approach to training and credentialing [laparoscopic surgeons]" to complement improved tool design (Pellegrini 1997). Traditionally, surgical training is based on apprenticeship and proctoring, where students learn by watching, and eventually participating in laparoscopic procedures (Tendick 1997). The competency of the newly trained surgeon may then be evaluated by an independent, skilled observer: a proctor (Pellegrini 1997). There is, however, an inherent risk to the patient in allowing a surgeon to learn laparoscopic techniques during an operations. As Way, Bhoyrul and Mori comment, "Essentially, in present practice, the surgeon's first 10 patients become part of [ones] training program" (Way 1995). Several options to traditional training have been adopted, such as using bench top trainers and cadaveric or animal specimens. In training for laparoscopic surgery, students can learn basic skills and movement coordination on bench-top, or in vitro, trainers (Sackier 1991). These trainers are generally a box with ports on one side to insert the video laparoscope and instruments where students train by operating on cadaveric animal tissue or plastic anatomical models placed in the trainer while viewing the operative site on a monitor (Sackier 1991, Zucker 1993). Sackier was the first to introduce the in vitro training box in 1991 (Sackier 1991), which is now commercially available from Karl Storz Endoscopy (Tuttlingen, Germany). Other bench-top trainers are commercially available, such as the Simuview Suture Trainer from Simulab Corporation (Seattle, WA). As noted by Way, Bhoyrul and Mori, "Most of the students can acquire the basic skills (i.e., placing and tying a suture) in about 2 hours" (Way 1995). In vitro trainers are useful in teaching the basic skills of dealing with limited vision and tool constraint problems, but are limited in that the dissections are not close to reality since there is no simulation of bleeding (Buess 1997), and it would not be practical to keep a library of synthetic models to cover all "important pathologies and anatomical variations" (Tendick 1997). 10 Animal specimens offer a training situation closer to actual surgery than in vitro trainers. Animal training is generally performed on anethesized pigs since "the pig is a good model for the surgeon to practice dissecting the gallbladder from the liver bed during cholecystectomy" and "an excellent model for operations on the upper gastrointestinal tract" (Way 1995). The cost, however, of one operation on a pig, exclusive of instruments, is around $500 US (Way 1995). Combined with large differences between pig and human anatomy, the requirement for special facilities to hold the animals and the need for anesthesia and preparation for surgery, animal specimens become less attractive for a long term, dedicated laparoscopic training method (Way 1995, Buess 1997). The emerging alternative to both animal specimens and bench-top training boxes is simulated surgery on virtual reality, computer-based training consoles. Tendick, Mori, and Way comment that the advantages of training with a surgical simulator include accurate human anatomy, realistic interaction, the ability to practice a technique or procedure repeatedly until skills are perfected, and the ability to record performance (Tendick 1995). The current disadvantage with virtual reality simulators is the large amount of computer power needed to perform convincing 3-D rendering of anatomical tissues and the cost associated with providing the necessary equipment (Tendick 1995, Buess 1997). Nevertheless, several groups are currently working towards development of virtual reality trainers for surgery. Satava had the first demonstration of a virtual surgical trainer in 1993 with a system allowing the user to take a virtual tour of the abdomen (Satava 1993). Tendick and Cavusoglu from the University of California, Berkley, are currently "developing algorithms for multi-scale modeling and tissue interactions, and integrating anatomical data into demonstration systems" for use in a laparoscopic simulator (Tendick 1997). Boston Dynamics Inc. has also developed a virtual reality simulator which currently simulates open surgery anastomosis (suturing tube-like organs together) with a graphical interface and haptic force feedback (Playter 1997). Another commercially available trainer is the MIST VR (Minimally Invasive Surgery Trainer - Virtual Reality), which uses conventional laparoscopic tool handles as the interface to a computer generated training environment (Wilson 1997). Opposed to the most other VR trainers, the MIST VR focuses on motor skills development by using simple geometric shapes for targets rather than realistic anatomy models, and incorporates measures accuracy, error and task completion time to facilitate skill assessment (Wilson 1997). One of the most advanced virtual reality trainers is under development by the Karlsruhe Research Center in Karlsruhe, Germany. This system employs a 3-D graphical simulation program (named KISMET) and a surgical training box to create a simulated laparoscopic surgery environment (Kuhnaphel 1997). In all respects, the 'Karlsruhe Endoscopic Surgery Trainer' behaves as the tools would in conventional laparoscopic surgery, except that the tools are instrumented with potentiometers to transfer position data to the KISMET system in order to operate in a virtual abdomen (Kuhnaphel 1997). The system has already been used in conjunction with development of the ARTEMIS telesurgery system (covered in the section 1.2.1) and is itself currently under modification to include force feedback to improve the simulated training conditions (Kuhnaphel 1997). The Training Center of the Section for Minimally Invasive Surgery, Department of General Surgery at the Eberhard-Karls University in Tuebingen, Germany is working closely with Karlsruhe group to further develop the 'Karlsruhe Endoscopic Surgery Trainer', and eventually implement the system into the training program (Buess 1997). 11 1.2.5 Assessing MIS Surgical Skills Through technical improvements and training, surgeons hope to improve their performance during MIS procedure. The question then becomes, how does one assess performance and skill in MIS procedures to see if improvements are occurring. Surgeon's want to standardize and benchmark performance in order to provide a means of quantitatively and reliably assessing the training of new surgical residents. The added advantage is the opportunity of assess the impact of new instruments or techniques on performance. Assessment methods can be grossly divided into outcome measures on simulated tasks, and passive assessment of performance in actual surgery. The line between training technology discussed in the previous section, and assessment techniques in simulated tasks blurs since developers of MIS trainers have their own ideas about what types of skills are important to develop to prepare a resident surgeon for the MIS operating room. In general, the outcome measure of choice for in vitro trainers is completion time on simulated tasks. Chung and Sackier use completion time on six simulated tasks to show improved performance in surgical residents taking a laparoscopic workshop (Chung 1998). In the more complex virtual reality systems, the surgeon has the opportunity to assess their performance on more diverse measures such as tool motion and accuracy (Willson 1997). Hanna, Cuschieri and associates were the first to introduce a computer-controlled, bimanual performance assessment system for endoscopic surgery which used standard instruments; the Advanced Dundee Endoscopic Psychomotor Tester (ADEPT) (Hanna 1998). Their system automatically prompts and records surgeon tool motion in performing five tasks: moving a slider and a dial, toggling a switch and manipulating a joystick. The ADEPT system offers the potential to assess performance by recording tool movement, error and task completion time. Hanna has also worked towards defining a quantifiable measure for performance based on endoscopic knot quality (Hanna 1997). They use a tensile testing machine to assess the ratio of ultimate knot strength and integrated tensile force up to rupture against the ultimate suture material strength and integrated force to produce a Knot Quality Score (KQS). From their study, they propose a KQS of 25% could be used as a benchmark level for performance in laparoscopic surgery. Derossis and her associates at McGill University in Montreal were the first group to develop an in vitro assessment method which measure skill based on both speed and precision (Derossis 1998b). The system, named the McGill Inanimate System for Training and Evaluation of Laparoscopic Skills (MISTELS), includes seven simulated tasks that closely approximate laparoscopic tasks performed in fundamental laparoscopic procedures. They went on to prove that their training system helped improve performance of surgical residents with practice (Derossis 1998a), and showed a high correlation in measured performance on both in vitro simulations and in vivo on a porcine model (Fried 1999). Another take on skill assessment is from the perspective of passive measurement of actual surgery. The concept behind this approach is that by measuring and decomposing laparoscopic surgery into its constituent tasks, surgeons and researchers will be able to evaluate which particular tasks are the most problematic and require attention. Claus proposed a quantitative analysis method in which they video tape the surgeon and laparoscope view during an operation and afterwards analyze the operation using a spreadsheet and a limited set of descriptive terms to decompose the procedure into its constituent surgical tasks (Claus 1995). The information recorded includes actions performed, instruments used, surgeon eye direction, and "effort" estimated subjectively on the evaluator's assessment of the surgeon's posture, verbal comments, 12 and degree of concentration. In general, they are concerned with the time and effort for each action. Cao takes this decomposition a step further by performing a task analysis that breaks the laparoscopic procedure down from surgical steps and sub-steps, down to tasks, sub-tasks, and motions (Cao 1996, Cao 1999). Cao promotes applying the decomposition technique to a training system which would allow the student to select the level of skill and knowledge they wish to concentrate on, such that a novice would concentrate on developing fine motion skills, while an expert would select a higher level to develop an understanding of the procedure as a whole. Cao finds that the surgical steps (and times) required for an expert to perform a laparoscopic cholecystectomy, for example, include preparing the patient (~10 mins.), isolating the gall bladder (~19 mins.), removing the gall bladder (~16 mins.). A further decomposition of "isolating the gall bladder" can be broken down into the tasks of "locate", "grasp & elevate", and "dissect surrounding tissue", each of which in turn can break down into fine motions: "reach & orient", "grasp & hold/cut", "push", "pull", and "release". Traverso prepared a multicenter trial with 10 other surgeons to determine a standardized procedure time and time for component tasks of the laparoscopic cholecystectomy operation (Traverso 1997). Out of 359 cases, they found that the mean completion time was 73 +/- 28 mins., which broke down into: place and remove trocars (25 mins./ 34% operation time), dissection (29 mins./ 40%), intraoperative cholangiogram (11 mins. / 15%), and remove gall bladder (5 mins. / 7 %). They found that the total time dissecting is most likely the cause of prolonged operative time, and that surgeon experience had a significant effect on operative time, while patient age and gender did not. Traverso proposes that surgeons can use these data to compare their own performance during laparoscopic cholecystectomies. 1.2.6 Ergonomics in MIS Ergonomics, from Greek words ergo and nomos meaning "Laws of Work", is the applied study of human limitations and patterns of work in an effort to reduce work' related physical trauma and suggest system changes to improve the quality of working conditions and performance. Simply put, it is the study of making the system suit the worker, rather than trying to make the worker fit the system. Patkin and Isabel note that the mismatch between technology currently available for MIS and normal human capacity suggests and important opportunity of the application of ergonomics (Patkin 1993). As discussed previously, specific issues related to the physical constraints of the instruments and patterns of tool use lead to work related trauma. Several reports have noted that often surgeons are forced to adopt awkward postures for extended periods of time, and that pressure points on the thumb, palm and fingers cause injury to the surgeons (Neuhaus 1997, Horgan 1997, Crombie 1996, Kano 1995). Berguer suggests that there are five important areas of consideration for ergonomic research in MIS: visualization, manipulation, posture, mental and physical workload, and OR environment (Berguer 1997). As discussed in previous sections, the problems of visualization and manipulation constraints have been widely addressed. Indeed, some researchers have applied an ergonomic approach to the design or evaluation of their systems by intentionally designing within the bounds of human limitations (Frank 1997, Amaral 1994, Emam 1999, Matern 1999). Whether they are aware of it or not, most researchers working to improve vision systems and instruments for MIS are involved in improving the ergonomics of the technique. Few, however, have openly addressed the need to study MIS from an ergonomic perspective in an effort to determine where specific problems lie and suggest what measures could be applied to correct them. A pioneer in ergonomics applied to laparoscopic surgery is Dr. Ramon Berguer (MD) of the University of California, Davis. His research has focused on comparing open and laparoscopic techniques and determining some of the risk factors for stress and fatigue related to minimally invasive techniques. His research collaborations include issuing an ergonomic questionnaire and measuring E M G activity of the thumb, finger and arm muscles on a laparoscopic trainer at the Society of American Gastrointestinal Endoscopic Surgeons (SAGES) conferences from 1996-1998 (Berguer 1998). Berguer, Remler and Beckley used a combined subjective ranking system and E M G comparison of open and laparoscopic techniques to illustrate the need for design optimization in MIS instruments as they discover greater forearm discomfort using MIS techniques (Berguer 1997). Alarcon and Berguer performed a comparison of OR crowding due to equipment and furniture between open and minimally invasive surgery (Alarcon 1996). They found 5% greater crowding in MIS and suggest that the increase could affect the performance of the surgical team unless addressed by innovative technology (Alarcon 1996). This problem was previously addressed by Schurr and Buess in presenting the OREST system of surgeon controlled, integrated endoscopic equipment, which was primarily designed to solve ergonomic and functional problems associated with the need for multiple devices in endoscopic surgery (Schurr 1995). Berguer also collaborated on a comparison of head, trunk and pelvis posture during open and laparoscopic surgery (Berguer 1997). By collecting data using a motion analysis system and a floor-mounted forceplate, they conclude that posture is more restricted during laparoscopic procedures and suggest that this may induce fatigue more readily than in open surgery (Berguer 1997). In another collaboration, Berguer and associates compared in-line (axial) and pistol-grip MIS instrument handles to determine which configuration was better for surgeons to use, from an ergonomic stance (Berguer 1998). They found no advantage to using one MIS instrument over the other, save the situation where task requirements demand the surgeon to work at 90° to the torso. In this instance, the pistol-grip forced the surgeons to bend their wrists back towards themselves, so the in-line handle showed a clear advantage in this situation (Berguer 1998). Berguer, Forkey and Smith collaborated in an experiment to study subjective and objective ergonomic problems associated with using laparoscopic instruments to conclude that MIS surgery is more taxing the surgeon than conventional open surgery (Berguer 1999). Through a questionnaire, they found that 12% of surgeons experience frequent pain in the neck and upper arms associated with MIS instrument use. In a study of muscle activity comparing a standard open instrument to a MIS instrument, they found that the MIS instrument requires higher forearm and thumb muscle contractions to operate (Berguer 1999). Other researchers have studied posture in surgery. Crombie proposed that repetition, extended static muscle loading, and extreme postures are likely the cause of strain injuries resulting from performing MIS (Crombie 1996). Rau, Radermacher, Thull and von Pilchler used the Ovako Working-posture Analysis Systesm (OWAS) method to study 14 types of laparoscopic operations and found that a high contribution of static working posture in the trunk, head, arms and hands are caused by a poor arrangement of equipment (Rau 1996)t extent can new instruments benefit surgeon performance in the operating room? Consequently, how can one 14 define and measure performance in a controlled setting which still has relevant implications to actual operations. 1.3 Research Questions Under closer scrutiny, one could argue that all the areas discussed in current literature, whether dealing with training, improved instrumentation, robotic assistance, ergonomic or skill assessment, are linked, to the common goal of improving surgeon motor performance. Our general research interest is to address the question of improving surgeon motor performance through engineering solutions. The goal of this thesis in particular is to establish a foundation for research into the design and assessment of improved instruments for minimally invasive surgery. The proliferation of research in the various areas of interest covered in the previous section raises a number of research questions. Especially considering the efforts to develop robotic assistants and new instruments to augment surgeon performance, a number of questions are raised: 1. What are the key physical limitations of the surgery which need to be addressed, by any system, robotic or mechanical? Can some of these limitations be ignored to simplify instrument design without compromising surgical task performance? 2. Is a complex robot the most practical solution for solving the motor control problems found in MIS? Could a practical and affordable mechanical solution be developed to overcome the same limitations currently addressed by complex electro-mechanical robotic systems? 3. To what extent can new instruments benefit surgeon performance in the operating room? Consequently, how can one define and measure performance in a controlled setting which still has relevant implications to actual operations? This thesis endeavors to address these questions. Chapter 2 presents an examination of the effects of physical constraints on surgeon performance, in order to determine the extent to which reversed motion and restricted movement of the instrument tip contribute to task performance. The results of the experiment suggest a 4 degree of freedom teleoperator device, which restores the natural motion mapping between the hand and instrument tip across the fulcrum point in the abdomen wall, provides an economical solution to performance problems. Chapter 3 presents a prototype 4 degree of freedom mechanical teleoperator designed to test this theory on simulated tasks. Such a device has the potential to address the same constraint problems currently addressed by robotic teleoperators, yet provide an affordable solution for improving performance in common surgeries. Chapter 4 discusses the need for developing well validated simulations in order to test the potential for new instruments or toolsets to demonstrate improvements in surgeon performance. An ergonomic posture analysis of a live laparoscopic cholecystectomy is presented to illustrate the feasibility of using a combined optoelectronic / video motion analysis system for simulation validation. 15 Chapter 2 The Effects of Physical Constraints in Laparoscopic Surgery 2.1 Introduction Although laparoscopic surgery has significantly reduced patient morbidity and overall length of stay in the hospital, the procedures are more technically demanding, primarily because of the awkwardness of the surgical instruments themselves (Treat 1996). In particular, because the instruments are introduced through "keyholes", surgeons must deal with three important limitations: (1) their instruments must be long (up to ~50 cm), (2) rigid instruments lose two translational degrees of freedom (x- and y-axis in the plane of the abdominal surface) due to constraints associated with the entry portal, and (3) the transverse translational motions of the tip of their instruments and their hands must be reversed, again because of the keyhole constraint. In combination, the loss of feel, the added motion constraints and the cognitive remapping dramatically increase the difficulty of surgical manipulations. A variety of devices and systems have been proposed and developed to deal with these problems; most can be classified either as advanced passive instrumentation aimed at increasing the internal degrees of freedom of the instruments (Melzer 1993, Mueglitz 1993), or as teleoperators, which use a master input device to drive a powered slave instrument which is inserted into the body (Hill 1994). While there is anecdotal evidence and a few comparative studies (Tendick 1996, Hill 1994) which show that these systems allow surgeons to perform various surgical tasks somewhat faster or more comfortably than when they use conventional laparoscopic instrumentation, none of the teleoperators are perfectly transparent, so it remains unclear how much room for improvement exists. The purpose of this study is to assess the ultimate potential performance of advanced laparoscopic teleoperators by studying how well surgeons perform simulated surgical tasks when using an emulated teleoperator which is perfectly transparent. We have chosen standardized surgical tasks most relevant to a variety of laparoscopic procedures (Claus 1995), and report here on two of the tasks we are currently studying: a pick-&-place task and a suturing task. We present experimental results showing that only reversed motions significantly extend the completion time of the pick-&-place task, whereas both reversed motion and limited degrees of freedom significantly extend the completion time of the more complex suturing task. Since these results represent the best possible performance of any conceivable laparoscopic teleoperator, they can serve as a basis for a performance index against which existing or future teleoperators can be evaluated. Such an index, with its quantified performance bounds, would be a valuable guide for a teleoperator designer trying to decide whether or not the potential benefits of a proposed device will be sufficient to offset its expected cost and complexity. 2.2 Methods When using a laparoscopic teleoperator, surgeons manipulate a master interface located outside the body whose motions are echoed by a slave end-effector inside the body (see Figure 4). Ideally, surgeons would have the impression that they are performing the surgery directly; in reality, however, they will be aware of the presence of the intervening systems of sensors and actuators which make up the teleoperator, and the dynamics associated with these components will degrade the surgeons' performance. Since an ideal teleoperator would exhibit no dynamics except the kinematic constraints inherent in the device, we can emulate a perfectly transparent teleoperator by directly constraining the master interface and asking the surgeon to perform simulated tasks while so constrained (see Figure 4). The performance of surgeons using directly constrained tools therefore represents an upper bound on their performance using any possible teleoperator. In this study, we assessed five different constraint conditions ranging from the existing 4 degree of freedom (DOF) laparoscopic tools (condition 4L) to 6-DOF open tools (condition 60). The other conditions we studied included 6-DOF laparoscopic tools (6L), and 4- and 5-DOF open tools (40 and 50). 17 Figure 4. Relationship between real teleoperator and our emulation of a perfectly transparent teleoperator using constrained open tools. The surgical tools consisted of a set of two conventional laparoscopic tools (Storz needle driver with in-line grip and grasper with pistol grip) and a set of two open tools (forceps and needle driver) modified to accept extension rods (described later). For the 6 DOF tasks, the subjects used either the laparoscopic or open instruments directly on the simulated surgical task; no additional constraints were imposed. For the 4 DOF tasks, subjects inserted their instruments through an "entry portal" consisting of a 5 mm high-density polyethylene (HDPE) bushing mounted in a self-aligning spherical bushing, as illustrated in Figure 5, and in Figure 6, panel 1. This bushing pair allowed all three rotations plus a plunge motion, but restricted motions in the transverse plane, much the way a trocar does in a standard laparoscopic procedure. When subjects used the open instrumentation in a 4 DOF task, we restricted 2 DOF by clamping an extension rod to one of the arms of the instrument and passing this rod through the bushing pair. In the 5 DOF open task, we used the same.extension rod clamped to the instrument, but we allowed the instrument to rotate about an axis perpendicular to the rod. 18 1 (1) (2) Figure 5. Experimental setup emulating transparent teleoperators using constrained open (1) and laparoscopic (2) tools. Figure 5 illustrates our experimental test setup for the open and laparoscopic tool configurations. Note that in the open configuration the surgeon grasps the instrument at a point between the constraint and instrument tip, while in the laparoscopic configuration the constraint lies between the surgeon's hand and the instrument tip. When using the 4 DOF open instruments, we allowed the surgeons to position the long axis of the grasper at an angle to the axis of the constraint rod, and most surgeons opted for an angle of approximately 30° (see Figure 5, Panel 1 and, Figure 6, Panel 2). One might initially suspect that this will make the 4 DOF open task significantly easier since the open instrument could approach a given point from an arbitrary angle, whereas the laparoscopic tool can only approach directly along the axis between the point and the spherical bushing. However, in the tests described here, the orientation of the grasper tip is determined by the orientation of the test piece (the object in the pick-and-place task or the suture in the suturing task), and since the degree-of-freedom limitations make the approach angle dependent on the tool tip orientation, the surgeon cannot independently choose their approach angle, so there is no effective advantage. Furthermore, both tasks are done in a plane which is approximately normal to the approach angle of the instrument, so there is no significant advantage to using anything other than a direct approach. The 40 and 4L configurations are therefore functionally equivalent in the context of this experiment. In all tests, the subjects could view their hands, the instruments and the surgical task directly, rather than through a laparoscopic camera (see Figure 6); this ensured that none of the results we report include any effects due to the restricted or distorted view presented by laparoscopic cameras. 19 Figure 6. Experimental apparatus. From left to right: (1) test stand with spherical bushings (circled) on support arms above the jig for the simulated suturing task, (2) open instrument (needle driver) with clamp and extension rod, (3) tray for pick-&-place task, and (4) suturing The simulated surgical tasks were mounted on a Plexiglas sheet (-50x80 cm), to which were attached two posts and arms fitted with the self-aligning spherical bushings described earlier (Figure 6, panel 1). For the pick-&-place task, we made a dish 15 cm in diameter with six semi-spherical holes of 12.5 mm radius drilled 6.25 mm deep with centres at a radius of 4 cm at 60° intervals around the circle, along with a similar hole in the centre (Figure 6, panel 3). For the suturing task, we used a crocheting frame fitted with medium-weight latex rubber (dental dam) with two incisions, approximately 3 cm long, in the centre of the rubber. The frame was held with a hobbyist's "third hand" which was secured to the Plexiglas base with rubber vacuum grippers (Figure 6, panel 4). The sutures were #3 silk sutures with ski tip needles. For the pick-&-place task, subjects began with a small cylinder (an electrical resistor) in one of the outer holes and their grasper above the centre hole. Subjects were asked to touch the centre of the centre hole, pick up the resistor, transfer it to another hole 120° away, and return to the centre. They then carried out five more transfers in various directions, ending up at the original hole, and repeated the entire sequence twice for a total of twelve transfers. We timed each transfer individually from one contact with the resistor to the next. For the suturing task, subjects picked up the needle with a needle driver, passed the needle through one side of the incision in the latex rubber, passed it through the other, and tied a simple square knot with the suture. Each subject repeated this task four times, and we timed each attempt from first contact with the needle until the second knot loop was pulled tight. 20 This experimental setup follows a previous pilot study performed with three surgeons and three graduate students (Hodgson 1997). For the present study, we recruited fifteen surgeons from two local hospitals to perform the pick-&-place and suturing tasks. There were five attending surgeons and ten residents drawn approximately uniformly from the six years of the surgical residency program. First year residents had no prior laparoscopic experience (beyond assisting with procedures), whereas the attending surgeons had performed upwards of 2000 laparoscopic cholecystectomies and other laparoscopic procedures. In all cases, the subjects performed the suturing task first and the pick-&-place task second. Within each task, they used the open instruments first (60, 50 and 40 in sequence) and the laparoscopic instruments second (6L followed by 4L). This sequence was chosen in order to have any learning effect benefits from practicing the tasks favour the presumably more difficult laparoscopic instrumentation. Subjects were allowed to make practice movements or knots under each experimental condition until they felt comfortable performing the task. Subjects were also allowed to redo any task where they felt their performance could be improved. The quality of the task performance was not assessed directly in this study. We observed, however, that most surgeons strove to perform the tasks at a quality level consistent with their level of skill and experience. In the case of suturing, for example, the surgeon would choose to redo a suture if they found it to be too loosely tied. 2.3 Results Figure 7 shows the average completion time for the suturing (15 surgeons) and pick-&-place (14 surgeons) tasks. In the case of pick-&-place, one surgeon did not complete the laparoscopic test, so that information is not included in this analysis. The trend observed in both tasks is that the mean completion time of the 40 condition is comparable to the 60, 50, and 6L conditions, and approximately 40% faster than with the 4L condition. In order to test the significance of differences between these mean completion times, a two factor analysis of variance (ANOVA) with m replications per cell is performed. In this case, we examine a Model III A N O V A , or Randomized Block A N O V A (Zar 1984). In this analysis, the surgeon is considered the random factor (taken from a normally distributed population of surgeons), and tool configuration is the fixed factor. There are 15 levels of surgeon factor in the suturing task, 14 levels of surgeon factor in pick-&-place, and 5 levels of tool factor in both tasks. 21 120 100 _ 80 6 0) — 60 a> E p 40 20 0 Mean Completion Times for Suturing Task (Bars: std. dev.; n=15) 65.87 " 3 T T O 18.07 36.21 34.54 . 60 50 40 6L 4L Mean Completion Times for Pick-&-Place Task (Bars: std. dev.; n=14) o 3 .1 2 2.61 1.78 1.75 1.85 60 50 40 6L 3.11 4L Figure 7. Mean task completion time for suturing and pick-&-place experiments. Error bars show standard deviation. The A N O V A analysis is used to test the following null and alternative hypotheses on both tasks: Ho: There is no effect on mean completion time due to tool configuration. H A : There is an effect on mean completion time due to tool configuration. One could also test for the effect of surgeon on the mean completion time due to surgeon. This is, however, not important for our study, assuming that the pool of surgeons we have tested is reasonably representative of those practicing laparoscopic surgery. Further, this analysis would require knowledge about the interaction effect (Zar 1984). This allows us to proceed with the analysis as in the case of a Randomized Block A N O V A , where surgeons are considered blocks and tools are the fixed effect tested on each block. 2.3.1 Analysis of the Suturing Task Based on an A N O V A test (m=4 measurements per cell), we reject the null hypothesis that there is no difference in completion time for the suturing task when using different tool configurations (F-41.43 > Fcrit=7.57, a=0.00001). Multiple comparisons between tool effects for a randomized block A N O V A is performed by the Tukey multiple comparison test (a=0.001) to test the null hypothesis (Ho) that means are equal. The alternative hypothesis (HA) in this study is that no differences exist between mean task completion times. The comparison test and subsequent Tukey confidence interval analysis are summarized in Table 1 below. Here, a significant difference in mean completion time for the suturing task is shown between the 4L tool condition and the other four tool conditions (60, 50, 40, and 6L). An important observation is that we see a significant difference between 60 vs. 40 and 40 vs. 4L. The implication here is that one would realize a time savings advantage by moving from a 4L to a 40 configuration without adding DOF. Perhaps not surprisingly, we realize an additional time savings moving from 40 to 50 to 60, since reorientation of the tool tip is important in suturing tasks. However, to take advantage of these time savings, one would have to add complexity and cost to the teleoperator. Note that we cannot reject the hypotheses that the means of 50 and 40 are equal. . 22 Table 1: Tukey confidence interval analysis for suturing task (a=0.001, n=15), Star (*) indicates significant difference between means (P < 0.001) Comparison Difference Conclusions CI Bounds between (F c r i t = 7.57) means Lower Upper 60 vs. 50 * 13.80 Reject Ho 3.86 23.74 ' vs. 40 * 18.13 Reject Ho 8.19 28.08 vs. 6 L * 16.46 Reject Ho 6.52 26.40 vs. 4L * 47.79 Reject Ho 37.85 57.73 50 vs. 40 4.33 Accept Ho - -vs. 6L 2.66 Accept Ho - -vs. 4L * 33.99 Reject Ho 24.05 43.93 40 vs. 6L 1.67 • Accept Ho - -. vs. 4L * 29.66 Reject Ho 19.72 39.60 6Lvs. 4 L * 31.33 Reject Ho 21.39 41.27 The power of this analysis is calculated to be greater than 0.99. That is, there is a less than 1% chance of having committed a Type II error (not rejecting a false positive, 'Ho: equal means') in the suturing task A N O V A . The power of detecting an interaction is also calculated, and found to be 0.76. This indicates a 24% chance of not detecting a true interaction between tool and surgeon. 2.3.2 Analysis of the Pick-&-place Task The results of the A N O V A for the pick-&-place task (m=\2 measurements per cell) suggest that there is strong evidence to reject the null hypothesis related to tool effect (F=47.41 > Fcrjt=7.25, a=0.00001). In the pick-&-place analysis, 16 of the 840 time measurements were corrupted by surgeon error in performing the task (e.g. dropping the cylinder during transfer from one cup to the next). These measurements were taken as 'missing data', and calculations were performed to compensate for their absence in the A N O V A tableau, in accordance with methods suggested by Zar(1984). As in the suturing task analysis, Tukey multiple comparison tests are performed to determine a significant difference in means for tool conditions. The results of this analysis are summarized in Table 2 below. The power of this test is also calculated to be greater than 0.99, (i.e. a less'than 1% chance of having committed a Type II error in this analysis). The power of detecting an interaction is calculated, and found to be 0.88, indicating a 12% chance of not detecting a true interaction between tool and surgeon. 23 Table 2: Tukey confidence interval analysis for pick-&-place task (a=0.001, n=14), Star (*) . indicates significant difference between means (P < 0.00J) Comparison Difference Conclusions CI Bounds between (F c„t = 8.38) means Lower Upper 60 vs. 50 0.03 Accept Ho - -vs. 40 0.07 Accept Ho -vs. 6L* 0.83 Reject Ho 0.55 1.10 vs. 4L* 1.33 Reject Ho 1.05 1.61 50 vs. 40 0.10 Accept Ho - -vs. 6L* 0.86 Reject Ho 0.58 1.14 vs. 4L* 1.36 Reject Ho 1.08 1.64 40 vs. 6L* 0.76 Reject Ho 0.48 1.03 vs. 4L* 1.26 Reject Ho 0.98 1.54 6L vs. 4L* 0.50 Reject Ho 0.22 0.78 Table 2 shows there is a significant difference in mean completion time for the pick-&-place task between the 4L tool condition and the open tool configurations (60, 50, and 40). In this task, we do not see a significant difference between the open tool conditions. The implications of these observations is that we achieve a significant advantage in moving from laparoscopic tools to open tools, but that there is no advantage to adding DOF to an open tool manipulator for surgical tasks which are similar to our pick-&-place task. The detailed times and statistical analysis of both suturing and pick-and-place task comparisons are detailed in Appendix A. 2.4 Discussion Our results confirm a number of previous studies which show that surgical tasks take considerably longer when performed laparoscopically rather than with an open procedure. For example, Cao (1996) analyzed surgeons performing laparoscopic appendectomies, cholecystectomies and fundoplications, and found that they performed tool orientation motions and end-effector operation motions serially, in contrast to the parallel organization of these motions in open procedures. This leads to significant increases in surgical times, particularly for procedures such as bowel resections which require much suturing. Although it is currently unclear whether this performance gap is primarily due to visualization or manipulation issues, a number of systems are available which boast either enhanced vision (Tendick 1995) or enhanced dexterity (Melzer 1993, Mueglitz 1993, Hill 1994). Depth-enhancement systems have progressed to the point where they are both commercially . available and considered by surgeons to be very useful; for example, our consulting surgeon has told us that there have been occasions where he would have had to convert from a laparoscopic to an open procedure had he not been using his Zeiss 3D system. Cao substantiated this opinion through a province-wide survey sent to 252 surgeons practicing laparoscopic procedures in British Columbia, Canada (response rate -30%) in which she found that many of them attributed the bulk of their difficulties to visual perception and hand-eye coordination issues. 24 This impression has been substantiated by Tendick et al. (1996), who performed a "marker plucking" test (similar to our pick-&-place task) under a variety of vision conditions ranging from full binocular direct vision to monocular endoscopic vision, using both open and laparoscopic instruments. They showed significant differences in task execution time between binocular and monocular direct vision (~50%> less with binocular), and between monocular direct vision and viewing through an endoscopic camera (-35% less with monocular). However, in all cases, using open instruments had only a minor effect (<15%) on the task performance times; in fact, when looking through the endoscopic camera, subjects actually took -10%) longer with open instruments. Because they had not expected that the kind of instrumentation would have had virtually no effect in this task, Tendick et al: also studied a more challenging knot-tying task using both kinds of instruments under both binocular direct and endoscopic viewing conditions. Again, they found that binocular direct vision reduced the task completion times by roughly 50%). Significantly, they found (under both vision conditions) that using open instruments reduced knot-tying times by 40-60%). Our results show similar improvements when using open instruments. Interestingly, while we have demonstrated the importance of manipulation limitations in this more complex task of knot-tying, the vision limitations may be much less significant. In a subsequent study, Tendick et al. (1995) revisited the knot-tying task and assessed the effects of different camera systems on performance using endoscopic instruments. Although they confirmed the superiority of direct vision (decreasing task time by ~25%>), they somewhat surprisingly found that neither stereo vision nor enhanced-contrast vision conferred any significant benefit over standard 2D laparoscopes for either experienced or novice surgeons. The most advanced 4-DOF laparoscopic telemanipulator to date has been developed by Hill et al. (1994). They tested the performance of two engineers and one laparoscopic surgeon on two simulated surgical tasks - a bead transfer task and a cannulation task. They found that for both tasks their subjects took roughly twice as long when using the telemanipulator as when using open instruments; this agrees well with our suturing results. Although they regarded the bead transfer as primarily a positioning task, the fact that it took the same relative amount of time as the more complex cannulation task suggests that it is more akin to the suturing task than to our pick-&-place task; in the latter, our subjects did not show a significant difference in completion time between using the 40 or 60 configuration. In order for advanced mechanical and optical instruments to be adopted in operating theatres, their makers will have to convince the hospitals that their systems will contribute to decreased costs, complication rates.or injuries to the surgeon. Since a primary determinant of operating room costs is the total time of a surgery, and since we and others have demonstrated the potential for reducing execution times on tasks which simulate various aspects of surgery, it is reasonable to suppose that we could estimate the impact on surgical times of enhanced instrumentation. To do so, we must first understand how much time is spent performing actions which could be sped up using a proposed device. Claus et al. (1995) defined laparoscopic surgeons' actions in terms of their functions during surgery (reallocation or exchange of tools, dissection, suturing, etc.) and timed how long they spent performing each of these functions in a variety of advanced laparoscopic procedures (including hernia repair, fundoplication, etc.). They found that tool positioning (i.e., reallocation or exchange of instruments), took -22% of the procedure time, dissection took another ~22%>, and suturing took only -5% (although suturing was also perceived to be the most difficult task to 25 perform laparoscopically). Considering positioning and dissection to be analogous to the pick-&-place task, then -44% of total surgery time is devoted to tasks similar to our simulated pick-&-place task, and - 5 % to tasks similar to our simulated suturing task. The first bar in Figure 8 illustrates the relative contributions of suturing, dissection and positioning tasks in laparoscopic procedures performed according to current practice (Claus 1995). If dissection is more like the pick-&-place task than the suturing one, then Tendick's results imply that a vision system which provides direct binocular vision could potentially achieve a 50% reduction in tasks which take 44% of a procedure's time (see the middle bar of Figure 8), while our results suggest that we can reduce that 22% of the current procedure's time by a further 40.5% (1.855 vs 3.115- or 8.9% of the total procedure time) by using instruments which behave like 4-DOF open instruments (the third bar of Figure 8). Hill 's results suggest that much of the total 33% reduction could be achieved by improved instrumentation alone. Despite the fact that suturing takes only 5% of the total time of current procedures, we presume that the tool behaviour itself is the primary determinant of the suturing time because Tendick et al. showed that enhanced vision systems provide little benefit, at least with endoscopic tools. In contrast, our results predict that we could save approximately 45.0% to 72.6% of suturing time by using tools which behave like 4- or 6-DOF open instruments, respectively. The major benefit of decreasing the suturing time and increasing surgeon comfort may well not be time savings per se, so much as increasing surgeons' willingness to attempt more complex surgeries than they have done so far using the conventional laparoscopic instrumentation. 50% 2- 40% a> E D) w 20% c o o I— 0_ 30% 10% 0% (1) Current Laparoscopic Procedure 22% 22% - -33% -22% 11% 11% 7% 5% 1 5% —— M-wfrn (2) With Direct Vision only Surgical Mode (3) With 40 Teleoperator & Direct Vision • Time Saved Q Positioning • Dissection • Suturing Figure 8. Proportion of time spent on three important tasks in laparoscopic surgery under: (1) normal (current practice), (2) direct vision, and (3) 40 teleoperator & direct vision conditions. Estimate of surgery time based on results of Claus (1995). 26 Considering all three tasks together, Figure 8 shows that we might achieve a 22% reduction in the time needed to perform these tasks if we could provide the equivalent of direct binocular vision. If we then added a 4-DOF teleoperator which performed as well as the constrained open tools (40 condition), we could reduce the time spent on these three tasks by a total of 33% of the total surgery time (~38% reduction over binocular direct vision alone). By adding a fifth or sixth degree of freedom to the teleoperator, we could reach total reductions of 34%> or 35% of total surgery time, respectively, but these small marginal improvements probably would not warrant the extra expense and complexity of the higher DOF teleoperators. Figure 9 illustrates the overall time savings for surgeries performed using the five different teleoperator configurations coupled with direct vision. Here, 22% of the total surgery time under the 4L condition is allotted to the pick-&-place task, and 5% to the suturing task, as discussed above (we neglect the dissection task). The four other conditions illustrate the reduction in total surgery time relative to the 4L case. We would expect to reduce surgical time by a comparatively large amount (~11.2%) if we could implement, in conjunction with a direct vision system, a teleoperator which behaved as the 40 tools did. Adding two additional degrees of freedom would only reduce surgical time by another 1.8%, although it may enable surgeons to attempt more complex procedures. We conclude, therefore, that we can achieve the greatest reductions in overall surgical time by eliminating the reversed motions instead of adding degrees of freedom. 30% E £ 2 0 % 0) O) i _ to o ^10% .*-» c CD U L_ <D 0% a Pick-&-Place • Suturing 60 50 14.0% 14.8% 15.8% 126 12.4% 13.1% 2.4% 2.7% 1.4% i 40 21.1% 18.5% 6L 27.0% 22.0% 4L Figure 9. Percentage of total surgery time for the five tool configurations. 27 Although these surgical time reduction estimates are somewhat speculative, they represent a reasonable extrapolation from our current knowledge of the distribution of motor activities during laparoscopic surgery (Cao 1996, Claus 1995) and the influence of motor constraints during representative motor tasks (Hill 1994, Tendick 1995, this paper). We are planning to validate these extrapolations by comparing the motion patterns surgeons use in conventional laparoscopic surgery and when using a simulated surgical setup. If we could directly compare times for equivalent open and laparoscopic procedures, then the extrapolations we have presented here and the validation study we are currently planning would not be necessary, but unfortunately differences between the two versions of a procedure make that impossible. Nonetheless, it is instructive to see whether or not the results of the several hundred clinical papers which compare equivalent open and laparoscopic procedures on the basis of operation time (among other things) are consistent with our results. We examined 25 references (totaling 29 comparisons) from this literature in more detail [example citations: (Cox 1996, Fisher 1991, Ortega 1995, Smith 1992, Stringer 1997)] and found that open procedures take approximately 74% ± 17%> of the operation time required for equivalent laparoscopic procedures. A complete list of these references and the procedure times quoted is available in Appendix A. In the in vitro test example we present in Figure 9, the results show that task completion time under the 60 configuration is approximately 52% of the 4L configuration. Our result is consistent with the 74%o figure mentioned above, in that it shows a reduction in operative time using open instruments, but it cannot be directly compared with that number because the elements comprising the 50% of a laparoscopic procedure which cannot be classified as positioning, dissection and suturing (e.g., trocar insertion, tool changes, incisions, etc.) are not equivalent to the other elements of open surgery. 2.5 Conclusions The type of instrumentation used by surgeons has a significant effect on how long it takes them to perform two simulated surgical tasks: suturing and pick-&-place. In particular, we have quantified the contributions of the fulcrum effect and of additional degrees of freedom to surgeons' performance on these tasks. Those results place an upper bound on the performance of any possible teleoperator, and we can use proximity to this bound as a performance index for both existing and future teleoperators. If either mechanical or electromechanical teleoperators could be developed which enable surgeons to move a master input device in a manner analogous to an open instrument, one could potentially reduce the time devoted to these tasks in current laparoscopic procedures by ~38%o over the gains which'could be achieved by implementing direct binocular vision alone.. Improved teleoperators which emulate the operation of open tools also have the potential to enable surgeons to perform procedures which are currently too difficult to attempt. The marginal benefit of providing 5- or 6-DOF in the 'open-tool' teleoperator, as opposed to 4-DOF, will depend on the extent to which dissection techniques are most like the pick-&-place or suturing task. 28 Chapter 3 A 4 Degree of Freedom Mechanical Teleoperator for Minimally Invasive Surgery 3.1 Introduction The consensus among surgeons and researchers alike is that instruments currently available for minimally invasive surgery (MIS) are cumbersome and difficult to use. Many solutions have been proposed to improve the functionality of current instruments, including improved, dexterous end-effectors (Mukherjee 1996, Balazs 1998, Frank 1997), improved handle design (Hasson 1993, Matern 1999, Amaral 1994, Emam 1999) and specialized tools for specific operative tasks such as suturing (Nagai 1999, Swain 1997). Though improvements to instrument functionality have the potential to ease the burden of MIS tool use, the operation of these devices is still complicated by the 'fulcrum effect' through the patient's abdominal wall. Two complications of the fulcrum effect are that the instrument handle moves opposite to the desired direction of instrument tip motion, and the handle-to-tip relative motions are scaled across the fulcrum. In a recent study, we show that by overcoming the problem of motion reversal and scaling alone, even without increasing the number of degrees of freedom (DOF) of instrument motion inside the patient, we can decrease task completion time in simulated MIS tasks significantly (Hodgson 1999, Chapter 2). Therefore, a sensible improvement to MIS instruments would be to overcome the fulcrum effect altogether in order to restore the natural, intuitive hand motions and make the sensation of performing MIS feel more like open surgery. One solution adopted by several groups to overcome the fulcrum effect of the MIS technique is to use robotic teleoperation technology. Teleoperation has been used in the past for nuclear material processing, undersea work, space exploration, and other applications where the work site is too remote, inaccessible or dangerous for a human to be physically present (Tendick 1995). The enticing potential of teleoperation systems, or teleoperators, for MIS is to place the surgeon's hands on the other side of the fulcrum, as if inside of the patient, and thereby restore the natural, intuitive motion mapping between the instrument tip and the surgeon's hands. The most progressive teleoperator systems seen today are surgeon controlled robots such as the SRI International telepresence system, which virtually immerses the surgeon in the surgical site (Green 1995), the Zeus system developed by Computer Motion Inc. (Goleta, Calif) based on the AESOP camera holding robot (www.computermotion.com), and the Karlsruhe telemanipulator system (Neisius 1995). These systems have proven their usefulness, as illustrated in recent reports of robotically assisted heart surgery performed using the Computer Motion Inc. Zeus system (Boehm 1999), though clearly superior basic performance results using this system over 29 conventional MIS tools remains unproven (Garcia-Ruiz 1998). The question is, however, are such systems cost-effective surgical tools for routine general surgery even given the potential for enhanced performance? Assuming that a given hospital would invest in only one large-scale robotic device, and that it would be used for extremely complex cases, what option is left for the surgeon who wishes to perform a routine operations such as a cholecystectomies? To address the need for a low cost MIS instrument which overcomes the fulcrum effect and restores the intuitive mapping between hand and instrument motions, we propose a prototype 4 degree-of-freedom (DOF) mechanical teleoperator with a unique friction drive mechanism. 3.2 Device The UBC mechanical teleoperator is illustrated in Figure 10. Based on the findings of Chapter 2, the teleoperator allows 4 degrees of freedom (DOF) of tool tip motion within the body (three rotation and one translation), the same degrees of motion which are currently available using conventional MIS instruments. Mechanically, the UBC teleoperator is designed to conform to the following functional requirements: • Resist a 15 N force applied perpendicular to the tool tip. • Resist a maximum moment of 1 N-cm applied to the tool. • Provide a stiffness of 70 N/cm in each direction. • Allow a range of motion of at least 45° from vertical, 15 cm plunge, and infinite rotation of the tool about its vertical axis. • Resist a maximum torque of 5 N-cm applied to the tool. • Resist a maximum plunge load of 10 N. • Align the slave tool such that its axis always passes through the entry port fulcrum in the skin. These functional requirements were laid out in the original concept report by Davis and George (1996) based on approximations of tissue stiffness and forces exerted by the surgeon. The advantage of this teleoperator design over conventional instruments is that the teleoperator restores the natural, intuitive motion mapping between the tool tip (slave, 1.) and the handle the surgeon manipulates (master, r.) by placing the surgeon's hand on the correct side of the fulcrum point. In this design, the surgeon moves the handle outside of the patient and the tool tip inside of the patient follows that motion exactly. The four-bar, parallel linkage provides a mechanical coupling to maintain the 1:1 movement mapping between the handle (master) and tool (slave) in the pitch and yaw DOF when moving the instrument from side to side or front to back. This parallelogram concept for providing pitch and yaw motions is based on an earlier 2 DOF teleoperator prototype designed in this laboratory (Davis 1998). 30 Figure 10. Isometric view of the UBC Mechanical Teleoperator. The surgeon manipulates the master tool handle (r.) in order to move the slave handle (I.) within the body. The four-bar, parallel linkage design maintains 1:1 movement mapping between the master and slave in the pitch and yaw DOF, while a unique friction drive mechanism translates the tool roll and plunge motions from master to slave. An inherent advantage of the parallel linkage arrangement is that it creates a fixed fulcrum point along the tool axis which remains stationary through all degrees and ranges of motion of the teleoperator, as illustrated in Figure 11. This fixed fulcrum is aligned with the point of entry into the abdomen during surgery to provide the passive safety feature of preventing lateral strain on the abdomen wall when using the device. 31 Handle (master) Figure 11. Illustration of the cone shaped range of motion of the UBC mechanical teleoperator about the fixed fulcrum point created by the parallel linkage mechanism. The dashed line represents the projection of the base rotation axis to the fulcrum point on the tool axis. The surgeon manipulates the handle (r.) and the tool (1.) follows the motion exactly. The unique work presented in this thesis is an original friction drive mechanism translates the tool roll and plunge motions. While the original parallelogram design provides a base for coupling the instrument tip pitch and yaw motions to the handle, a method of producing roll and plunge motions is required to achieve the 4 DOF required for full functionality. A concept for producing these motions was presented previously in the work of Drown et al. (1998) in which they used a gimbal mechanism to provide roll and a moving slider to produce plunge. An illustration of this conceptual prototype is presented in Figure 12. Figure 12: Conceptual model of a 4 DOF teleoperator by Drown et al. (1998). A slider mechanism provides plunge motion and gimbals couple instrument roll. Though functional, it was later decided that this design had too much moving inertia due to the plunge and roll mechanisms. The effect of inertia would be difficult to compensate and would adversely affect the smooth motion desired of the device. In order to compensate for this, an approach was adopted to design a drive mechanism which would translate the plunge and roll motions from a fixed position on the teleoperator base frame. The challenge with the teleoperator design was to devise a method of transferring both the rotation and plunge motions while keeping the drive mechanism as low as possible in order to reduce the moving mass of the teleoperator about the pivoting axis (illustrated in Figure 11). The solution we ultimately adopted was based on the work of Colgate and Peshkin who have designed a spherical friction drive mechanism which uses contact friction between a common driving sphere and individual drive axles. The general principle of a friction drive mechanism is to use a rotating friction contact in order to drive the tool. In the case of the teleoperator, a sphere pressed against the tool shaft is used as the friction drive. By rotating the sphere about the vertical axis, the tool will rotate, and by rotating the sphere about the horizontal axis, the tool will plunge. This concept is illustrated in Figure 13. 33 Figure 13. Illustration of the principle of a friction drive mechanism employed in the UBC teleoperator design. Rotating the horizontal drive while the tool is vertical causes the tool to plunge, while rotating the vertical drive causes the tool to roll. The concept of the friction drive mechanism employed here is based, in part, on the work of cobotic devices by Peshkin and Colgate (Peshkin 1996). A cobot ("Collaborative roBOT") is a passive robot which is powered by human input forces, but controls the direction in which forces can be applied. Such devices employ continuously variable transmissions (CVTs) in order to control force input. A C V T is a device which couples the angular velocities of drive rollers according to an adjustable ratio. In the C V T illustrated in Figure 14, the rotation of the two shafts on the left are constrained to be in a proportion dictated by the angle set on the steering rollers on the right. The large ball in the centre is the mechanism drive ball. Figure 14: Illustration of a continuously variable transmission (CVT) employed by a cobotic device. Rotation of the two shafts on the left are constrained to be in a proportion dictated by the angle set on the steering rollers on the right. (Source: Peshkin 1996) 34 The current prototype of the mechanical teleoperator uses a standard mouse ball (a rubber coated steel ball) as the drive mechanism. The mouse ball was selected as the friction drive device based on qualitative tests on a single sided prototype and friction tests performed on the rubber coating (Dussaults 1998). The two drive shafts pressed against the sphere transfer tool plunge and rotation between the master and slave, as illustrated in Figure 13. In Figure 10, the horizontal shaft near the top of the device drives tool plunge, and the drive belt and pulley system facilitates tool rotation. It is interesting to note that activation of the horizontal and vertical drive shafts are coupled when the tool is pivoted out of vertical alignment. Consider, for example, moving the handle in a pure plunge motion (without rotation) when the tool is at a 45° angle to the vertical orientation. As illustrated in Figure 15 below, even though tool roll was not desired, the vertical drives are active. Though this is an interesting point to mention about the friction drive, it does not cause unintentional tool motion or effect the function of the teleoperator on the whole. In early design concepts we considered a decoupled friction drive system, but abandoned it when we discovered that it would continually change the tool position at the fulcrum point, which is to remain invariant. (a) (b) (c) Plunge • > • • Drive shaft Mouse ball Tool rotation rotation motion Figure 15. Illustration of the friction drive function, (a) Only the horizontal drive shaft rotates, in turn rotating the mouse ball and causing the tool to plunge when in the vertical position, (b) Only the vertical drive shaft rotates, in turn rotating the mouse ball and causing the tool to roll (rotate about its axis) when in the vertical position, (a) Both the horizontal and vertical drive shafts rotate equally; the coupled rotation causes the tool to plunge when angled at 45 °from the vertical position. 35 3.2.1 Physical prototype A physical prototype teleoperator was designed using SolidWorks 98 Plus, and constructed locally in the Mechanical Engineering Machine shop based on drawings and materials presented in Appendix B. Figure 16 below shows the completed teleoperator. The teleoperator weighs ~1.1 kg, most of which lies above the main rotational axis between the fulcrum points illustrated in Figure 11. In order to compensate for the inertia caused by this 'top-heavy' design, brass counterweighs are mounted below the main rotation axis (Figure 16). Counterbalancing the device inertia reduces the strain on the surgeon and improves tool responsiveness. For future generations of the device, we are considering replacing these counterweights with a system of linear springs described by Rahman et al. (1995) which would reduce the overall weight of the teleoperator and allow for fine tuning the counterbalance. The entire teleoperator is designed to mount to a professional studio light stand (Manfrotto, Italy; Serial no. 143/IA08) which may be clamped to the operating room table, or to a separate floor mount. We envision that a surgeon would use the mechanical teleoperator while seated in the operating room beside the operation table with a thin screen or mirror projection of the laparoscope image placed between the surgeon's eyes and hands. Positioned in this manner, the surgeon would likely be more comfortable over the duration of long procedures and the monitor view combines the display space with hand workspace to restore a natural and intuitive mapping between the hands and the remote tool tip. MacKenzie has demonstrated improved performance in pointing tasks using this direct monitor arrangement compared to using a remote monitor placement as is commonly used in MIS operations (MacKenzie 1999). Materials for the laboratory prototype were selected based on availability, but the surgical version of the teleoperator could be easily converted to surgical grade metals to allow sterilization. The mechanical teleoperator has no electrical parts, which offers the advantage of being subjected to less stringent device approval regulations, which in turn offers the potential for more rapid adoption into clinical settings. 36 Figure 16. Prototype of the mechanical teleoperator with friction drive. Shown here mounted on the photo stand mounting arm (Manfrotto, Italy) Major components of the friction drive mechanism are shown in Figure 17. Figure 17: Close-up of the friction drive mechanism. (A) Drive ball (1) held in position against the tool (2) by a ball castor on a pivoting yolk (3). (B) Back view showing the pivoting yolk (3) and adjustable seating force screw (4). The horizontal drive shaft for tool plunge (5), the tool (2), and the vertical drive shaft and pulley for tool roll are also shown (6). Note that the drive belt has been removed from the pulley (6) for image clarity. 37 Ultimately, the design will incorporate a forceps-like handle for the surgeon to manipulate the instrument tip within the abdomen, as illustrated in Figure 18. A preliminary design was developed by Brown et al. (1998) in cooperation with the author. The design is modular, so a surgeon may select another tool head and exchange it interoperatively without removing the . entire teleoperator from the patient or changing the master handle. Solutions discussed for providing tool jaw functioning include a direct cable link between the master and slave, or pneumatic or hydraulic actuation. Figure 18: Conceptual illustration of the surgeon handle interface (r.) which controls motions of the MIS instrument tip within the abdomen (I.). The black point along each instrument shaft represent the location of the fulcrum point of the mechanical teleoperator. 3.3 D iscuss ion Initial qualitative tests have shown the 4-bar linkage system to be successful in producing the pitch and yaw motions relative to the fixed fulcrum point shown in Figure 11. This motion, however, has been successfully implemented in an earlier prototype so the results are not surprising (Davis 1998). The plunge and roll motions, however, could not be produced by the friction drive mechanism as expected. In examining the device response to increased seating force, it was discovered that the rubber coating on the mouse ball deforms around the drive shafts and tool, becoming increasingly difficult to move, up to the point where the system seizes. Up to that critical point, the coulomb (dry) frictional force is hot great enough to prevent the slave tool from slipping during master tool activation, despite high seating forces. The C V T drive illustrated in Figure 14 is not limited in size or materials as our teleoperator friction drive is, and has thus been optimally configured for successful implementation. Though our friction drive can transmit motion intermittently, the mouse ball does not provide effective, consistent force transmission and should therefore not be relied upon. Trials with solid stainless steel balls have proven to be more successful, but the resulting motion is not as free and transparent as we hope to achieve with this device. Consequently, continued evaluation of a variety of friction surfaces is recommended. Currently, alternative designs for plunge-roll motion transmission which do not rely on friction drive are also being considered. For plunge motion, we are considering a direct linkage mechanism which would couple the motion of slave and master across a double, mirrored pivot point set between the two ends. Rotation may be achieved by a bevel gear type arrangement connected to a drive shaft between the master and slave. Since the current teleoperator is effective for transmitting pitch-yaw, we plan to use the base of this prototype to construct a second generation prototype teleoperator with an improved plunge-roll motion transmission mechanism. Once we have a functioning prototype, we plan to use this device in laboratory trials to test the potential performance improvement over conventional MIS instruments on tasks representative of actual surgery. We are currently developing standardized surgical task simulations and^ a performance index based on motion analysis studies of expert surgeons performing actual MIS procedures in the OR. We hope to demonstrate the usefulness of a mechanical teleoperator on these simulated tasks and, based on the laboratory studies, design an improved teleoperator appropriate for in vivo clinical trials. 3.4 Conclusion This chapter presented the original design of a prototype 4 degree of freedom mechanical teleoperator for minimally invasive surgery. The. pitch-yaw motion transmission is effective, but the friction drive mechanism for plunge-roll is not reliable using a standard mouse ball. Recommendations for future research includes design of an alternative mechanism for plunge-roll motion transmission and continued testing of friction materials on the friction drive. 39 Chapter 4 A motion-analysis system for ergonomic and performance studies in the OR: A pilot study of ergonomic posture sampling during laparoscopic cholecystectomy 4.1 Introduction The need for well designed and accurate simulations of minimally invasive surgery (MIS) is critical to the continued advancement of the technique over the next ten years. Predictions indicate that 70% of surgeries in North America will be performed laparoscopically by 2008 (Globe and Mail, March 1, 2000). Given this rate of acceptance, the need for well trained laparoscopic surgeons will become acute within the next decade. The challenge to hospitals and accreditation bodies is to train and evaluate new surgeons objectively before they perform their first live operations. Furthermore, laparoscopic instrument designs are found to cause serious fatigue and strain related injuries and current efforts are focused on assessing and improving designs (Emam 1999, Berguer 1997, Radermacher 1996). All of these areas would benefit from well validated simulations of MIS techniques which would allow for controlled studies of the effectiveness and usefulness of training methods and new instruments, as well as the objective assessment of surgeon skill and performance. The following chapter discusses our initial efforts at validating surgical task simulations through a "proof of concept" pilot study using a combined optoelectronic / video motion analysis system to assess the ergonomic postural stress of performing a laparoscopic cholecystectomy (gall bladder removal) using the minimally invasive technique. The cholecystectomy was selected as the baseline surgery for study since it is widely accepted as the most common and well defined laparoscopic procedure (Traverso 2000). Though our long term research interest is the design and assessment of new surgical instruments, accurate simulations of discrete surgical tasks would prove invaluable in assessing the potential for new instruments or toolsets to improve surgeon performance in actual surgery. The ergonomic assessment of surgery is itself a topic of interest in the area of minimally invasive surgery. A valuable element of applied ergonomics is the study of workers in their natural work environment in an effort to determine which tasks or task components may be risk factors for developing work related strain injuries. Dr. Ramon Berguer is a pioneer in illustrating that 40; laparoscopic surgery is more economically stressing than conventional open surgery, in , comparisons of fatigue due to instruments (Berguer 1999), postures (Berguer 1997) and OR crowding (Alarcon 1996). He has also performed comparisons showing equal ergonomic stress between in-line and pistol-grip handles on laparoscopic instruments (Berguer 1998b). Other researchers have also employed ergonomic assessment techniques in evaluating laparoscopic surgery or instruments. Radermacher et al. (1996) used ergonomic techniques to assess static working postures and critical tasks in an effort to improve workplace design through virtual reality simulations. Emam (1999) used an optical motion analysis system in tracking the angular velocity of the shoulder and elbow in simulated tasks designed to compare laparoscopic instrument handles. Previous ergonomic posture analyses in the operating room relied on direct or recorded observations of the surgery (Radermacher 1996) at sampling frequencies ranging from 1 min. (Kant 1992) to 5 seconds (Berguer 1997). In this study, we propose to use a combined optoelectronic / video motion analysis system to perform a continuous ergonomic posture analysis of a laparoscopic cholecystectomy. Automating the posture sampling process has the advantages of higher sampling rates and reducing tedium of high frequency manual assessments. Also, automating the assessment eliminates the need for expertise in subjective posture ratings. Bhattacharya (1999) showed the usefulness of an automated, portable dosimeter based system for free, continuous (1 Hz) ergonomic posture measurement of carpentry work. In this pilot study, we hope to show that our optical motion analysis system proves similarly useful for continuous posture measurement during surgical tasks in the operating field. 4.2 Materials and Methods 4.2.1 Motion analysis system and OR setup The motion analysis system used in this experiment consists of an optoelectronic motion tracking system package combinded with standard video analysis. The optoelectronic system was chosen for the experiment because it allows for rapid and accurate tracking of passive markers in three dimensional space. The video system was added to augment the motion analysis with information about the actual surgical task corresponding to the raw motion tracked by the optoelectronic system. The components of the motion analysis.system used in this study include the following: • Polaris Hybrid Optical Tracking System (Northern Digital Inc., Waterloo, Ontario, Canada) • 10mm 0° surgical laparoscope, camera and illuminator (Stryker Endoscopy, Santa Clara, CA) • Laptop computer (PC, Pentium II, 233 MHz) • Video camera (Panasonic OmniMovie S-VHS PV-S770-K) • Digital video mixer (Videonics MX-1, Campbell, CA) Video cassette recorder (General Electric Pro-Feet QP VG4025) 41 • Color video monitor (Sony PVM-1271Q) • Camera stand tripod (Manfrotto, Italy) • Camera stand mounting arm (Manfrotto, Italy) These components are arranged and connected in the OR as illustrated in Figure 19. Note that the laparoscope system is a standard component of the OR video cart provided by the hospital and does not represent special experiment equipment: Figure 19: Components and operating room layout of the Polaris Hybrid Optical Tracking System and video recording equipment used in this study. The Polaris system tracks both active markers (infra-red light emitting diodes, IREDs) and passive retro-reflective marker spheres (both shown in Figure 21) to determine the location of the surgeon's upper limbs in 3D space (A more detailed description of the Polaris system is presented in the following section). The Polaris system consists of two separate components; the Position Sensor (2kg; 60cm x 10cm x 12cm) and the Tool Interface Unit (5kg; 32cm x 13cm x 30cm). The Position Sensor is mounted on an adjustable camera stand arm near off of the OR video cart in a position near the laparoscope image monitor in order to afford an unobstructed view of the surgeon. The Tool Interface Unit is placed in a convenient location outside the sterile field and controls the Position Sensor and active IRED markers. A computer connected to the Position Sensor controls the Polaris system and stores raw limb position data collected during the surgery. Images from the laparoscope and the OR video camera are mixed and recorded on standard VHS tape. Following the analysis method of Cao (1996), the digital mixer is used to create a ' picture -in-picture' image with the smaller laparoscope view in the foreground superimposed on the 42 background view of the external OR video camera trained on the surgeon. The date and time are displayed and recorded on the OR camera view. The video monitor connected to the recording V C R allows the experimenter to view the mixed images and ensure a clear view of both are recorded. 4.2.2 Polaris Optical Tracking System and Marker Arrays The main component of the motion analysis system is a Polaris Hybrid Optical Tracking System (Northern Digital Inc., Waterloo, Ontario, Canada), as shown in Figure 20. The Polaris system tracks the 3D coordinates of both active infra-red light emitting diodes (IREDs) and passive retro-reflective markers mounted on custom made 'L-shaped' aluminum marker arrays, also shown in Figure 20. The Polaris dynamically tracks and calculates the coordinate transformations describing the location (X,Y,Z spatial coordinates) and orientation (roll, pitch, and yaw angles) of marker arrays at a sampling rate of 15 Hz and returns the information to the host computer. Figure 20: Polaris Hybrid Optical Tracking System (Northern Digital Inc., Waterloo, Ontario, Canada). Components of the Polaris include the Position Sensor (top) and the Tool Interface Unit (bottom). Active and passive marker arrays are also shown, with a 6 inch ruler included for scale. The active marker arrays (Figure 21, L. panel) are made from 1/8" plate aluminum machined into an 'L-shaped' base supporting three lAn diameter active infra-red LEDs (Northern Digital Inc., Waterloo, Ontario, Canada). Active marker array activation is controlled directly by the Polaris during sampling periods. An elastic sling fitting around the surgeon's chest and shoulders firmly secures the torso active marker array in position over the sternum. The sling 43 supports the active marker array and prevents the surgeon from becoming entangled in the marker electrical control cable, as well as prevents folds in the surgical gown from obscuring the markers during surgery. Adhesive Velcro® hook tape attached to the back of the marker array locks with a patch of Velcro® loop material sewn into the elastic sling to hold the marker array in place. Figure 21: Polaris marker arrays: (L.) active IRED marker array, (R.) passive retro-reflective marker array. The Polaris system tracks up to three (3) active marker arrays and six (6) passive marker arrays, sequentially sampling each of the three active arrays followed by all passive arrays, at a rate of 15 Hz. Passive marker arrays (Figure 21, R. panel) are also made from 1/8" plate aluminum machined into an 'L-shaped' base. The passive array base supports three W diameter polypropylene balls (Part no. O-BBP-8, Small Parts Inc., Miami Lakes, FL) covered with overlapping discs of retro-reflective tape (3M, St. Paul, MN). The reflective marker balls are threaded and mounted to the array plate with machine screws. Two parallel slots machined into the wide section of the marker plate accommodate an elastic Velcro® strap (Vel-stretch, Velcro Canada Inc., Surrey, BC, Canada) which is used to fasten the marker arrays onto the surgeon's proximal and distal forearm. Adhesive Velcro® hook-and-loop tape is used to fasten one marker array to the back of the surgeon's dominant hand. Sterilized Velcro® loop tape is attached to the glove on the back of the surgeon's hand during the procedure to hold the marker array in place without encumbering the surgeon's hand motion. The torso marker sling, passive marker arm cuffs, and adhesive attachments to the hands were found to be acceptably immobile relative to the torso, arms, and hands in subjective tests. Future tests may quantify any motion artifact by objective measures. The instrument in the surgeon's dominant hand is tracked with a passive marker array on a 3" post fastened to a 3A" metal binder clip (ACCO, Lincolnshire, IL). The binder clip securely attaches the marker array to the ring handle or shaft of the instrument, yet is easily removed and replaced to facilitate instrument changes during surgery. Two instrument tracking marker arrays are provided to allow the assisting nurse to attach the array on the next instrument before the surgeon requires it. The binder ring attachment system is designed to fit instruments used at the Vancouver hospital (Storz endoscopic graspers 26175 M E , Karl Storz, Tuttlingen, Germany), and therefore may not be appropriate for all instrument designs. A more robust design is required to ensure all instrument designs can be tracked. 44 In total, four (4) passive arrays and one (1) active array are used to describe the motion of the surgeon's dominant arm and instrument motion in this pilot study. All equipment within the OR is tested and approved by Vancouver Hospital Biomedical Engineering, cleaned, and Ethylene Oxide (EtO) sterilized before entering the sterile field. 4.2.3 Ergonomic posture measures In human factors and ergonomics literature, attention is given to assessing poor working posture as both an indicator of the potential for developing work-related musculoskeletal disorders and for improving working conditions by identifying problems with workstation design (Karhu 1977, Aar J 3 s 1988, Keyserling 1986, Bhattacharya 1999). Methods of postural sampling are commonly used to identify potentially harmful postures in manual labor tasks, either by direct observation or instrument-based continuous sampling techniques (Genaidy 1994). In postural sampling, the severity of ergonomic stress correlates to the degree of joint angle deviation from neutral joint positions, thus ergonomic stress level can be ranked based on zones of joint angles observed during a given task cycle. One posture analysis method which has been widely adopted for its simplicity and rapid application is the Ovako Working Posture Analysis System (OWAS), developed at the Ovako steel company in Helsinki, Finland, to assess poor working postures seen in their factory (Karhu 1977). The technique is attractive because it simplifies posture analysis by providing diagrams to assist observers in classifying poor postures. Berguer et al. (1997) and Radermacher et al. (1996) have used the OWAS technique for evaluating surgeon posture during minimally invasive surgery. One concern that should be considered when applying OWAS to surgery is that it was designed based on assessments of working in heavy industry. Consequently, the posture limits and scores in OWAS are based on the subjective opinion of steel workers of their working conditions and concentrate primarily on the torso and lower limbs. Also, the OWAS method is considered a macropostural classification system as it groups non-neutral postures around a joint into one category, and is thus less descriptive (Genaidy 1994). For example, OWAS posture zones for the upper arm indicate only if none, one, or both of the arms are above shoulder level (Karhu 1977). Since it is unlikely that a surgeon would be working with elbows at or above the shoulder level, a more specific measure is required to describe upper arm postures. As an alternative to the OWAS method for ergonomic posture classification, we propose to use the Rapid Upper Limb Assessment (RULA) technique, a posture sampling method designed specifically for assessing upper limb postures during light manual work (McAtamney 1993). Like the OWAS method, the RULA method evaluates ergonomic stress exposure by using body posture diagrams and scoring tables to specify posture zones. In the RULA method, the posture limit guidelines and scores are based on the combined findings of several ergonomic studies (McAtamney 1993). Considering the light work context of minimally invasive surgery, it is more appropriate to use the posture zones specified by the R U L A method of posture sampling. Table 3 below details the posture zones used in the RULA technique. ' 4 5 Table 3: Original RULA (Rapid Upper Limb Assessment) posture classifications (McAtamney 1993). Upper arm scores Flexion/extension angle 1 -20° - 20° 2 < -20°; 20° - 45° 3 , 45° -90° 4 > 90° +1 if upper arm abducted Lower arm scores elbow flexion angle 1 60°-100° 2 0°-60°;>100° +1 if arms cross body midline or are out to side Hand scores wrist flexion 1 0° (neutral) 2 -15°-15° 3 <-15°;>15° +1 hand in radial or ulnar deviation Wrist pronation/supination 1 mainly in mid-range of pronation/supination 2 near or at range limits One problem with adopting the RULA posture zones to our free-motion posture analysis, however, is the confusion over shoulder joint 'abduction-adduction' and 'flexion-extension'.' In biomechanics, these clinical definitions of joint angles have not been well defined from a calculation perspective since the axis of rotation changes depending on the order of rotations (An 1991). Thus, 'flexion' followed by 'abduction' does not result in the same position as 'abduction' followed by 'flexion'; a phenomenon known as "Codman's Paradox" (An 1991). In ergonomics literature, however, shoulder disorders are commonly associated with postures where the elbow remains elevated, in any combination of flexion and abduction (Kroemer 1989). In order to avoid confusion over shoulder rotations, we adopt the convention used by Keyserling (1986) where shoulder flexion/abduction is classified by identifying elbow elevation below 45° from the neutral (anatomical) position as neutral, 45°-90° as mild elevation, and above 90° as severe elevation. A score similar to the R U L A method is applied to indicate the posture level: neutral (1), mild elevation (2), severe elevation (3). A further modification is presented to account for potential motion artifact and measurement error in the wrist angle measurement such that the flexion-extension and radial-ulnar deviation penalty zone is increased from 0° deviation from the neutral position to ±5°. Finally, in order to numerically classify the range of wrist pronation/supination, we adopt the posture classification presented by Genaidy (1994), where pronation or supination outside a ±15° range from neutral is considered poor posture. The lower arm classification of 'arms crossing the midline of the body or working out to the sides' is assessed by considering the relative position of the wrist and elbow joint centres with respect to the torso marker array. Finally, additive scores such as midline crossing or wrist deviation, as seen in the R U L A analysis (Table 3), are separated in our analysis in order to identify the contribution of individual postures to ergonomic stress. The posture scoring strategy adopted in this study is presented in Table 4 below. 46 Table 4: Modified RULA posture scores adopted for current study of continuous ergonomic posture analysis during minimally invasive surgery. Upper arm scores Elevation angle 1 0 ° - 45° < l i 2 4 5 ° - 9 0 o ( 1 ) 3 > 90° ( 1 ) Lower arm scores elbow flexion angle 1 60° -100° 2 and arms cross body midline or are out to side 3 0°-60°; >100° 4 and arms cross body midline or are out to side Hand scores Wrist flexion 1 ;. -5 - 5° (neutral) 2 -15°-15° 3 <-15°;>15° Wrist deviation 1 - 5 - 5 ° (neutral) 2 < -5°; >5° ( 2 ) Wrist pronation/supination (+9 = pronation) 1 -15°-15 o ( 3 ) (neutral) 2. < -15°; >15° ( 3 ) arm elevation posture zones from Keyserling (1986) ( 2 > range added to RULA score to accommodate potential motion artifact and measurement error wrist pronation/supination posture zones from Genaidy (1984) 4.2.4 Kinematic model of the upper limb In order to examine the posture of the surgeon's upper limbs in free motion, a method is required to determine the approximate joint angles from raw marker position data. To simplify motion analysis, a rigid-body model of the human upper body is required to relate the tracked limb motions to the joint angles of interest. The arm model employed in this study is a ball joint at the shoulder, a rotary (hinge) joint at the elbow, and a two rotary joints at the wrist describing the motions upper arm abduction-adduction and flexion-extension, elbow flexion-extension, and wrist flexion-extension and radial-ulnar deviation, respectively. A further axial-rotary joint oriented along the forearm axis from the elbow joint centre to the wrist joint centre describes forearm pronation-supination. The functional arm model is illustrated in Figure 22 below (R.). Placement of the four marker arrays used to capture the arm motion is also illustrated in Figure 22. 47 • Figure 22: Surgeon arm model: (L.) Marker placement to capture limb motion during surgery and (R.) corresponding functional arm model of the upper limb. Numbers indicate reference frame number associated with the tracked marker arrays. V • 4.2.5 Kinematic joint centre calibration To dynamically track the joint centres and angles from raw marker position data, we employ methods of classical dynamics (i.e. homogeneous transformation matrices) familiar to , biomechanics and robotics literature. Details of the homogeneous transform, coordinate frame nomenclature, and some selected topics related to coordinate transforms are compiled in Appendix F. Schmidt et al. (1999) propose a method of upper limb motion tracking using triads " of retro-reflective markers rigidly supported by cuffs on the upper arms, lower arms, and hands. The design of the Polaris system constrains us to track rigid marker arrays attached to the surgeon's limbs, so the method proposed by Schmidt et al. (1999) is a reasonable for our application. In our model of the upper limbs, the active marker array on the torso describes a rigid frame of reference for the shoulder joint. Similarly, relationships are described between the proximal 48 forearm array and the elbow joint centre, and the distal forearm array and the wrist joint centre. Thus, knowing the relationship between the calculated joint centres and the tracked marker arrays, one can calculate the joint centres locations at any point in time. For example, the global elbow joint centre location relative to a tracked marker on the forearm may be determined mathematically with the following equation. c / ( 0 = r 0 2 (0C e 2 (O (i) Here, T02 is the homogeneous transformation matrix representing the forearm coordinate frame {frame 2) with respect to the global base frame (frame 0 - the Polaris measurement frame), and Ce represents the fixed location of the elbow joint centre relative to frame 2. Knowing T02, which is evaluated from the tracked marker array coordinate and orientation data, one can describe the location of the elbow joint centre in frame 0, Ce°, at any posture where the forearm marker is visible. The tracked marker frames are numbered as follows: (7) torso; (2) proximal forearm; (3) distal forearm; (4) hand (see Figure 22). The global joint centre locations of the shoulder (C/), elbow (Ce°) and wrist (Cw°) are then described by the following equations. . C/(O=T 0 1 (OC/(0 Ce°(t) = T02(t)Ce\tc) (2) Cw°(0 = T0i(t)Cw\tc) Where t represents any sample time where the measured frame (7,2, or 3) is visible, and Cs , Ce and Cj are constant at tc, the calibration time. The next step in developing the above equations is to determine the constants Cs'(tc), Ce2(tc) and Cw3(tc), the constant joint centre location relative to the respective tracked marker fixed to that segment. This process is known as joint calibration. Schmidt et al. (1999) propose a method of joint calibration in which they locate anatomical landmarks which approximately lie on the axis of rotation of a joint (e.g. the lateral and medial epicondyles at the elbow) and estimate the joint center as the midpoint of the line between those landmarks. The Polaris system, however, does not allow for accurate tracking of individual markers (only arrays of markers), so anatomical landmark based localization of joint centers, as proposed by Schmidt et al. (1999), is not a feasible solution for locating joint centres in this study. Though a probe could be constructed to replace the function individual markers as proposed by Schmidt et al. (1999), the required knowledge of appropriate anatomical landmarks and the skill in locating them greatly effects the accuracy and efficiency of the calibration method. Alternatively, we employ a kinematic (dynamic) calibration method using circle and sphere fitting techniques to mathematically locate the approximate centre of each of the shoulder, elbow, and wrist joints. The basis of kinematic limb calibration is to locate of the joint axis or centre of rotation by fitting a circle (or sphere) to a cloud of raw data points collected from limb markers, as illustrated in Figure 23. If each point in the data cloud is assumed to be a 'snapshot' of a single point fixed to a rotating body, then the distance from the centre of rotation to each point describes the radius of a circle. To illustrate this point in a clinical application, consider collecting a circular arc of data points from a marker on the distal forearm while flexing and extending at the elbow. By fitting a circle to the arc of points, and assuming the elbow acts as a rotary (hinge) joint, one can approximate the axis of rotation of the forearm as the circle centre and thus locate the elbow 49 joint axis kinematically. Similarly, sphere fitting methods are used to locate the shoulder and wrist joint centres. 1 1 / / " ^ L P B \ V LLS \ TLS J • • , i L _ J i i -100 -50 0 50 100 Figure 23: Illustration of a cloud of data points describing a 90° arc of a circle (R = 100, noise = ±5%). The LLS (Linearized Least Squares; Santo-Munne 1996) and TLS (True Least Squares; Zill 1992) methods fit an optimal circle to the data, while the LPB (Linearized Perpendicular Bisector; Halvorsen 1999) method solves the intersection of all perpendicular bisectors of sets of two points in the data cloud. Three kinematic calibration circle/sphere fitting methods are available for joint centre calibration in this experiment: (1) True Least Squares (TLS) (Zill 1992); (2) Linearized Least Squares ; (LLS) (Santo-Munne 1996); (3) Linearized Perpendicular Bisector (LPB) (Halvorsen 1999). The basis of the LLS and TLS methods is to find the optimal base circle which fits the circular arc of raw data. The LPB method, on the other hand, localizes the centre by solving the intersection of the set of perpendicular bisectors generated between two points. These methods are illustrated graphically in Figure 23 and presented in greater detail in Appendix C. The three kinematic calibration methods were compared for accuracy and precision on simulated data at various noise levels and circular arc sizes (detailed in Appendix C). Simulation results show that the TLS method is the most accurate, but has the tendency to break down dramatically at high noise levels and small arc sizes, while the LPB and LLS methods are less accurate but more robust to noise. The order of preference for applying circle fitting methods based on ' accuracy is: (1) TLS, (2) LPB, (3) LLS. The joint calibration is performed on the shoulder, elbow, and wrist joints respectively. The shoulder and wrist joints are modeled as spherical joints, thus the joint centre may be determined 50 by sphere fitting. During shoulder joint calibration, the subject moves their arm in two crossed arcs (flexion-extension and abduction-adduction at the shoulder) in the approximate shape of a sphere. The subject holds their forearm in extension and data is collected from and distal forearm frame near the wrist {F3) relative to the body frame {FI), such that the resulting shoulder joint location is calibrated relative to the body frame, Cs'. The wrist joint is similarly located by moving the hand frame {F4) in two crossed arcs approximating a sphere fit (radial-ulnar deviation and flexion-extension) relative to the distal forearm frame {F3) at the wrist. The calibrated wrist joint centre is then Cw3. The elbow joint is modeled as a hinge joint and thus the joint rotation axis is determined by circle fitting. The subject performs flexion-extension of the elbow and data is collected from and distal forearm frame near the wrist (F3) relative to the body frame {FI). It is important for the subject to hold the upper arm vertical during this calibration by locking the elbow to the waist in order to minimize extraneous forearm motion relative to the body frame {FI). Note that this flexion-extension calibration only determines the axis of rotation of the elbow, and the joint centre could lie anywhere on that axis. In order to locate the centre of rotation at the elbow, a second circle fit is performed between the proximal {F2) and distal forearm frames {F3) to locate the pronation-supination axis of the forearm. The intersection of flexion-extension and pronation-supination rotation axes is taken as the elbow joint centre relative to F2, Ce2. 4.2.6 Joint angles •\ Knowing the location of each joint centre, one may determine the joint angles for the posture analysis. The joint angles of interest, as illustrated in Table 4, are shoulder elevation, elbow flexion-extension, forearm pronation-supination, wrist flexion-extension and wrist radial-ulnar deviation. Traditionally, in motion analysis studies, angular limb position is expressed either by sequential Cardan (XYZ, Y Z X , etc.) or Euler (ZYZ, Y X Y , etc.) rotation angles taken directly from the rotation matrix between coordinate frames on the joint centre (Schmidt 1999, Peterson 1996), or by directly solving for the joint angle by measuring the angle of the vector between joint centers from a given coordinate frame axis (Anglin 1993). An illustration of both the Eulerian and Direct methods of joint angle calculation is provided in Appendix D. The Direct joint angle calculation method is adopted for this study because calculating the angle based on the joint center alone is easier to visualize and simplifies calculations. Also, the dependence of Eulerian solution on sequential rotations can lead to erroneous joint angle solutions in compound joint rotations. This point is elaborated in the discussion on wrist joint angles calculations. 4.2.6.1 Shoulder Joint The shoulder joint elevation angle calculation follows the approach of Anglin (1993), taking the angle from the vertical (Y) axis at the shoulder joint to the vector described between the shoulder and elbow joint centres, calculated as follows: u = V *e ye Z c J 0 -1 0] (3) 51 e d s = cos" u • V where u represents the components of the elbow joint centre relative to the shoulder frame (Ce), determined by the following equation. Cf=(T0ITisy]T02C> (4) where Ce is known from kinematic calibration of the elbow joint centre, and TAB represents the homogeneous transform from frame A to frame B. There are some conditions to using this calculation on the shoulder joint angle. The first is that the Y-axis of the tracked torso array (7), which provides reference orientation for the shoulder joint frame (T/s), should be nominally aligned with the torso vertical plane. Second, one must accept inaccuracy in the joint angle measurement as the true shoulder joint centre shifts relative to the fixed, calibrated location during arm motion, as previously noted. Since, however, the ranges of interest for joint angle elevation noted in Table 4 cover a broad range of shoulder movement, we expect to capture the representative range of postures, if not the precise elevation angle. 4.2.6.2 Elbow Joint The elbow joint angle is a straight forward calculation, given the location of the shoulder, elbow, and wrist joint centre at a given time during motion capture in the global reference frame, 0. From basic linear algebra, one can calculate the joint angle of the elbow as follows. u~Ls -Ce v = C„,°-C„° . (5) ' U • V 9 = ;r-cos" In this calculation, the joint angle ranges from 0° at full extension to nearing 180° at full flexion. This format is adopted to conform with the joint angles expressed in the R U L A posture notations. 4.2.6.3 Forearm pronation-supination The forearm pronation-supination angle is determined from the relative rotation of the marker arrays on the proximal and distal forearm, frames 2 and 3 respectively. Here, pronation is defined as the positive rotation and supination as negative rotation of the frame 3 origin marker about the forearm axis, between the elbow and wrist joint centres. Begin by described the joint centres and the origin of frame 3 relative to the fixed frame on the proximal forearm, frame 2. 52 Cc2{t) = T;2\t)T03{t)Ce\tc) Cw\t) = T-2\t)Tm{t)Cw\tL) C2{t) = T02\t)T03{t)C,\t) (6) Here C 3 3 is the origin of frame 5: (0,0,0). Next, define the forearm rotation axis between the elbow and wrist joint centres and determine the angle required to align the axis with the x-axis of frame 2. From this we determine the homogeneous transform T2E between frame 2 and an elbow joint frame with x-axis oriented towards the wrist joint centre. u = X2 = C2 - C2 v = X 2 2 = [ l 0 0] 9 = cos u • v (7) r = u x v The homogeneous transform matrix between frame 2 and frame E is then the rotation matrix defined by rotation by 0 about the vector, r, Rr(0). The calculation of rotation about an arbitrary axis is detailed in Appendix F. From this, determine the position of C j £ from the following equations. K{O) = TE2 C 3 E = TE2C. (8) Where C / is the known position of the frame 3 origin relative to frame 2. Now determine the angle offset between the z-axis of frame E and the y-z components of C j £ . u = v = 0 y , 2 3 0 0 l] + C O S - C O S f \ U • V VII IIH 1  y f \ u • V v r t l r l i y if (v x u) • -X > 0 {pronation) if (u x v) • -Xs < 0 {supination) (9) o p s { t ) = e p s { t l ) - e p s { Q Where dps(tj) represents the current measurement and 6ps(tc) represents the measurement of the neutral posture at calbration time, tc, by the same method. The procedure for neutral calibration will be detailed later. 53 4.2.6.4 Wrist Joint The definition for wrist joint angles measured in this study are flexion-extension and radial-ulnar deviation, as illustrated in Figure 24 below. Figure 24: Definitions of wrist joint angles: (I.) radial (+) and ulnar (-) deviation; (r.) flexion (+) and extension (-). Neutral posture at 0° (Source: Anglin 1993) A problem inherent in the measuring of joint angles by either Eulerian or Direct calculation methods is that the representation of compound postures depends on sequential rotations about various axes. The problem is pronounced in this study in measurements of the wrist joint angles. As an illustrative example, imagine rotating your hand from a neutral position to 30° radial deviation, followed by flexion from 0° through to 90°. At the extreme range of flexion, both the Eulerian and Direct methods would predict the radial deviation angle to still be 30°, as it was the first rotation in the sequence. Based on a visual examination of the hand posture, however, the value one likely would estimate for radial deviation would be closer to 0°. This example is expanded in Appendix D. From a physical interpretation of the joint angle, one is interested in the rotation away from a neutral plane of interest, independent of the other joint angles. This is consistent with the visual nature of posture sampling techniques, such as the modified RULA evaluation used here. Thus, a different type of joint angle measurement is proposed for this study, herein referred to as clinical joint angles. Returning to the hand rotation example noted above, let us define the Clinical joint angle for radial-ulnar (RU) deviation as the elevation angle of the hand marker from the x-z plane at the wrist, and flexion-extension (FE) as the elevation angle of the hand marker from the x-y plane at the wrist, as illustrated in Figure 25 below. Now, the measures of RU and FE are independent of each other, and more accurately represent the physical joint angles seen clinically. o° + 54 Figure 25: Clinical joint angles for flexion-extension (Qfs) and radial-ulnar deviation (Q-RU) of the hand in the wrist frame, C/, . 6FE is the elevation angle ofpoint C™from the X-Y plane, and 9RU is the elevation of point C & f r o m the X-Z plane. Flexion is in the -Z axis direction, and radial deviation is in the +Y axis direction. iv In order to calculate the Clinical joint angles for the wrist, take the projection of the marker (Cj, in Figure 25) into the neutral plane of interest and determine the angle between the vectors described by these points. The following equations illustrate the calculation of the wrist flexion-extension angle, 9FE-u V -w w h y h W W e F F w { t , ) = + cos u • V \\\u\lKs -I —cos u • V vinllhi/ if zf < 0 (flexion) if zh =0 (neutral) if zhw > 0 (extension) (10) 0FEW{t) = 6FEw(tl)-9FEw(tc) Similarly, for wrist radial-ulnar deviation angle off of the X - Z plane, ORU-u = 'RU w w w A w Xu 0 zh + COS -1 -cos u-v vinllhiy u-v vnwllllviiy 55 if yhw > 0 (radial deviation) if yh =0 {neutral) if yhw < 0 (ulnar deviation) (11) QRUW (() = O R I / ^ , ) - 6 R U W ({c ) As with the forearm pronation-supination angle, (t,) represents the current measurement and (tc) represents the measurement of the neutral posture at calbration time, by the same calculation method. 4.2.6.5 Neutral posture angles Since marker arrays are attached to the upper limbs without strict concern for orientation relative to the limb or other arrays, it is important to perform a static posture measurement in order to determine the relative position of certain arrays in an anatomically neutral posture. From this posture calibration, we obtain neutral angles for forearm pronation-supination, and wrist flexion-extension and radial-ulnar deviation. There is no need to calibrate against a neutral position for the elbow joint, since the joint angle is calculated directly at each point of interest. Similarly for the shoulder joint, the neutral posture is not required as the joint angle is calculated relative to the fixed shoulder joint frame, the orientation of which depends on the placement of frame 1 on the torso. Following Anglin (1993), the neutral posture for forearm rotation is conventionally defined with the thumb facing upwards while the elbow is flexed at 90° and the upper arm vertical, as illustrated in Figure 26. Supinotwn Neutral Pronation Figure 26: Forearm neutral position and rotation definitions (Source: Anglin 1993) Neutral wrist posture is measured with the hand straight relative to the forearm and fingers and thumb extended, again illustrated in Figure 26. This definition is also used by Schmidt et al. (1999). 4.2.7 Validating posture angle measurements 56 Due to time constraints, no controlled studies of the posture angle measurements could be performed. The validity of joint angle measurements was evaluated subjectively based joint angles observed in the calibration motions from the OR test and two additional laboratory tests. Figure 27 below illustrates the spread of joint angle data from a laboratory subject (AH) during calibration motions of the wrist joint in flexion-extension. The top graph represents the physical marker locations on the hand relative to the wrist frame, and the bottom graph shows the joint angles for each point. Y-axis WHstfe^u 40 60 Samples (n=100) Figure 27: Example of visual joint angle validation for the wrist. (Top) Plot of points in wrist frame generated by wrist flexion-extension calibration test for laboratory subject AH. (Bot.) Wrist flexion-extension (solid) and radial-ulnar deviation (dashed) joint angles for the wrist f-e calibration plot. 57 Note in Figure 27 that the deviation angle of the wrist is relatively stable throughout the range of wrist flexion. Similar plots for the shoulder, elbow and forearm were used to assess the validity of joint angle measurements subjectively. It is recommend for future work to prepare studies comparing angles measured with Polaris against a controlled joint angle output in order to assess the accuracy of these measurements. For the purposes of this ergonomic study, which is based on visual assessments of joint angles, it is assumed that the angles calculated reasonably represent the physical joint angles. 4.2.8 Study protocol The following protocol was adopted in this experiment. The surgeon is asked to scrub and enter the OR while the patient is being anesthetized. Before the operation begins one researcher also scrubs in, in order to attach the sterilized marker arrays to the surgeon's hands, lower arms and torso (Figure 28, left panel). The researcher remains scrubbed in to assist the surgeon in attaching the markers to the laparoscopic instruments during a tool change, and to be available to make any adjustments to the marker arrays or support cuffs (Figure 28, right panel). If the researcher is unable to scrub in, they instruct the OR nurse on the correct procedure for attaching the marker arrays from outside the sterile field. Figure 28: Study protocol: (L.) Researcher fitting surgeon with passive and active marker arrays, (R.) Surgeon performing the operation while researcher stands ready to assist with marker arrays problems and tool changes. A second researcher (not shown) remains outside the sterile field to operate the motion analysis system. Once the marker arrays are secured in place, another researcher outside the sterile field performs a test to ensure that the motion analysis equipment is operational. A subsequent test is preformed on the Polaris optical tracking system to ensure that there are no external sources of infra-red light in the OR which would interfere with marker tracking. Once the equipment is tested and confirmed operational, the researcher informs the surgeon that they are ready to begin recording. When the surgeon confirms they are ready to begin surgery, the researcher begins tracking the marker arrays with the Polaris system, as well as begins recording the operation from both the external video camera and laparoscopic camera. The equipment records for the duration of the initial open and subsequent laparoscopic portion of the surgery. When the laparoscopic portion of the surgery is completed, the surgeon informs the researcher and the equipment is shut off. The surgeon continues to wear the marker arrays once the surgery is completed in order to perform the joint centre calibration. The researcher instructs the surgeon to perform a five minute sequence of shoulder, elbow and wrist rotations which are used postoperatively to approximate the surgeon's joint centre locations and neutral posture orientation of the arrays. Once the joint centre calibration procedure is complete, the surgeon is free to remove the marker arrays. If the surgeon feels uncomfortable wearing the markers at any time during the surgery, they are free to stop the experiment and remove the markers. Approval for this experiment was granted through the University of British Columbia Clinical Research Ethics Board and the Vancouver Hospital. 4.3 Results Posture data was collected with the motion analysis equipment during a laparoscopic cholecystectomy at the University of British Columbia site of the Vancouver Hospital. Joint postures are calculated for intervals of the standardized laparoscopic cholecystectomy operation based on the surgical task segments defined by Traverso et al. (1997). Though Traverso and colleagues segmented each stage of the laparoscopic cholecystectomy from placing the initial trocar to close, we examine only the main surgical tasks directly related to the minimally invasive portion of the procedure: cystic duct dissection, gallbladder dissection, and gallbladder removal. Table 5 below lists the surgical steps, endpoints, and mean percentage of time for operative tasks as described by Traverso et al. (1997). Note that, though included in the study by Traverso et al. (1997), the intraoperative cholangiogram (IOC) was not performed in the operation examined in the current study and is thus excluded from the examination. From the combined OR/laparoscope video of the studied surgery, the start and end times for each of the tasks noted in Table 5 were collected and presented in Table 6 below. Table 5: Surgical steps, endpoints, and percentage of overall surgery time for operative tasks examined in posture sampling study. Categories and task times from Traverso et al. (1997). Task Start point End point Mean operative time mins. std. dev. ('% total time) Total operation time Initial incision Close incision 72±28 mins. Cystic duct dissection begin cystic duct dissection end cystic duct dissection 15±11 mins. (20.5%) Gallbladder dissection begin gallbladder dissection end gallbladder dissection 14±8 mins. (19.2%) Remove gallbladder end gallbladder dissection gallbladder removed from abdomen 12±6 mins. (6.8%) 59 Table 6: Breakdown of surgical steps, clock times, and percentage of overall surgery time for operative tasks from the laparoscopic cholecystectomy studied during the motion analysis. Task Start time (hh:mm:ss) End time (hh:mm:ss) Mean operative time mins. (%total time) Total operation time 08:15:27 09:38:12 82.2 mins. Cystic duct dissection 08:30:40 08:43:45 13.1 mins. (15.3%) Gallbladder dissection 08:43:22 09:17:21 34.0 mins. (41.1%) Remove gallbladder 09:17:21 09:28:22 11.0 mins. (13.3%) Note that all tasks, with the exception of Gallbladder dissection, observed in this surgery are within the normal limits stated in Table 5. In the operation observed for this experiment, the gallbladder dissection was complicated by an aberrant right artery following the normal line of dissection between the gallbladder bed and the liver. The increased operative time was due to careful dissection of the artery until it could be properly identified and clipped. Having identified the occurrence time for the three operative tasks of interest, the posture analysis outlined in Table 4 is applied over the duration of each task. Figure 29, Figure 30 and Figure 31 below illustrate the percentage contribution of ergonomic stress level for each joint of interest according to the prescribed posture definitions. A normalized, weighted postural stress score (in %) is presented for each joint angle in the following figures as determined by the following equation. n = 2,3, £ (Posture_scoreJ° in l * %Contributionfnt) Postural_stressH,,„g,=^  7 — jjn (12) (max_Posture_scoreJ J The normalized, weighted postural stress score is determined in order to provide an easily comparable measure of overall postural stress in each joint. Note that the normalized stress score excludes the contribution of normal posture (z-1) in order to examine only how ergonomically stressful a given task is. If the joint angle remains totally within the normal posture limits during a task, the contribution to actual postural stress is 0%. Further, the maximum posture score is used as the baseline to normalize stress contribution across each joint. Using the maximum score as a baseline represents a maximally stressful posture adopted for the duration of the task as more stressful than a moderately stressful posture adopted for an equal amount of time. The overall normalized, weighted postural stress scores for all joints and tasks are illustrated in Figure 35. 60 Stress contribution levels: Cystic duct dissection (CDD) Norm Wgt Stress yy0 61% 74% 52% 6% 100 Shoulder Elbow Wrist f-e Wrist r-u Forearm p-s Normal posture CM Q2 B 3 U4 Highly stressing posture Figure 29: Cystic duct dissection (CDD) stress levels for each joint angle, including the normalized, weighted ergonomic stress for each joint angle (top). Posture scores are based on definitions presented in Table 4. Stress contribution levels: Gallbladder dissection (GBD) Norm Wgt Stress «| o/Q 54% 94% 65% 87% 100 T S 8 0 £ w o I 60 c o a l_ O B 40 c o S n | 20 o o o.o. / 1.0 99.0 2.7 4.9 81.1 65.4 34.6 Shoulder Elbow Wrist f-e 87.4 12.6 Wrist r-u Forearm p-s Normal posture • "] Q2 • 3 M4 Highly stressing posture Figure 30: Gallbladder dissection (GBD) stress levels for each joint angle, including the normalized, weighted ergonomic stress for each joint angle (top). Posture scores are based on definitions presented in Table 4. 61 Stress contribution levels: Gallbladder removal (GBR) Norm Wgt Stress <| Q/Q 100 80 o I 60 o c o D) s c o '*-> £ 20 o o 1.4 98.6 Shoulder 38% 81% 59% 43% 59.1 43.5 56.5 40.9 Elbow Wrist f-e Wrist r-u Forearm p-s Normal posture D1 02 14 Highly stressing posture Figure 31: Gallbladder removal (GBR) stress levels for each joint angle, including the normalized, weighted ergonomic stress for each joint angle (top). Posture scores are based on definitions presented in Table 4. 4.3.1 Dealing with corrupted data It is important to note that the percentage contribution to ergonomic stress illustrated in Figure 29 to Figure 31 is not determined as a factor of total task time, but rather a factor of total available angles for a given joint. Joint angles which were corrupted due to marker occlusion or high RMS error in marker localization (> 0.5 mm) were excluded in determining the total percentage spent in a given posture. RMS error (in mm) is calculated internally by the Polaris system at each sample as the best least-squares fit between pre-calibrated and measured marker positions. The RMS error rejection threshold value of 0.5 mm was determined based on the 99 t h percentile (by standard deviation) over average values collected in laboratory trials, and also represents - 1 % of the minimum physical distance between markers on the arrays. 62 Table 7: Sample block of joint angle from experimental data illustrating the definitions of missing data, partial data, and complete data. Note the colour code for data status (Missing -black, Partial - gray, Complete - white). Joint angles are presented in radians Corruption Status HHHl Sample clock time (HMMssmsms) Shoulder elev. Elbow Wrist wKKk Wrist Forearm Hands out to sides Missing u 0 0 0 0 0 Missing 0 0 0 0 0 0 Partial 8393477 0.55 • 1.72 -0.19 -0.15 0 -1 Partial 8393494 0.53 1.66 -0.15 -0.07, 0 .1 Complete 8393510 0.52 1.60 -0.14 -0.10 0.07 1 As illustrated in Table 7, joint angle data for a given sample may be either missing, partial, or complete. A potential problem with excluding corrupted (either missing or partial) joint angle data in calculating the percentage time spent in a given posture is that large gaps in joint angle data may occur during surgical task performance which would skew the results. For example, if the markers happened to be occluded by the surgeon contorting their shoulder into a highly stressful posture while performing a surgical task, this data would be lost from the posture analysis even though it represents an important contribution to ergonomic stress during task performance. Such a loss of data could potentially represent a serious problem as an average of 33% of all joint angle data were found to be corrupted over the duration of the surgery. If, however, the lost data is associated with tasks such as tool changes or dealing with experiment specific problems (i.e. marker array adjustments), we could safely ignore these periods of corrupted data as they are not associated with normal instrument manipulations and therefore not useful in targeting task specific postures. Consequently, it is important to confirm that large blocks of missing data occur most frequently in "non-surgical tasks", that is, tasks which are not associated with instrument manipulations within the abdomen. In examining the cystic duct dissection (CDD) task as representative of the whole surgery, it was found that a greater percentage of totally corrupted data could be attributed to marker occlusion during instrument changes or in dealing with problems not related directly to performing the surgical task (e.g. de-fogging laparoscope lens, dealing with Polaris marker array problems) as illustrated in Table 8 below. Here, the percentage of corrupted data is taken as a function of the number of samples noted during a prescribed task interval in which all joint angle data were missing (see Table 7). Such an occurrence is assumed to correspond to periods of marker occlusion rather than due to Polaris marker localization errors. Table 8: Percent missing data during tool change/error tasks vs. normal surgical tasks for cystic duct dissection (CDD) task. Time: sec. (%) Percent missim ] data Cystic duct dissection task 785 i100%) mean std dev median Non-surgical tasks ni 13) 298 (38%) 63%u 30% Normal surgical tasks (n=11) 487 (62%) 31% 29% 19% (*) Nonparametric Mann-Whitney test: greater amount of corrupted data (<x=0.05, 0.01<P<0.025) 63 A nonparametric Mann-Whitney test was performed testing the null hypothesis (Ho) that the percentage of totally corrupted data is not greater in non-surgical tasks compared to normal surgical tasks (analysis expanded in Appendix E. The Ho was rejected (a=0.05, 0.01 < P < 0.025). This result is interpreted to indicate that a significant majority of totally corrupted or occluded data within a data set is attributed to non-surgical tasks, and can therefore be excluded from the posture analysis. Knowing this, we can ignore 19% of data corruption as it is totally corrupted. Figure 32 below illustrates breakdown in percentages of good or corrupted data for each joint over the duration of the surgery (from the start of CCD to the end of GBR). re *J re •o re *-> o +•> 4— o co c a> u i_ a. 100% 80% 60% 40% 20% 0% 19% 19% 19% 19% S%t 7% 23% 21% 76% 74% 59% 60% Shoulder Elbow Wrist Forearm • Complete • Partial • Missing Figure 32: Breakdown of data status showing complete, partial and missing data in each joint over the duration of the surgery from start CCD to end GBR. Table 9 below illustrates the distribution of missing, partial and complete data over each sampled minute of surgery (at -300 data samples per minute). Annotations of the major surgical events occurring within each minute are presented. In this table, only the data status which was most frequent within that minute is illustrated by colour code. The percent frequency for majority data status is given in each case. Table 9: Annotated frequency of data status within each minute of recorded surgery. Surgery time Task Status (%) Status colour code legem Partial "• Missing 8:29 % 8:30 Start cystic duct dissection (CDD) 1 8:31 Localize duct 8:32 53 8:33 63 8:34 Polaris marker array falls off tool 59 8:35 62 8:36 91 8:37 Clip and cut cystic artery 59 8:38 (several tool changes) . 57 8:39 48 8:40 Lens fogged 8:41 Clip and cut cystic duct 87 8:42 End CCD 54 8:43 Start gallbladder dissection (GBD) 'JS 8:44 60 8:45 65 8:46 79 8:47 73 8:48 63 8:49 81 8:50 •71 8:51 98 8:52 T C - graspers to cauterizing spatula 71 8:53 Localizing aberrant right artery 72 8:54 8(> 8:55 100 8:56 79 8:57 79 8:58 62 8:59 69 9:00 80 9:01 71 9:02 *a 9:03 '• 79 9:04 72 9:05 81 9:06 T C - spatula to clip applier 55 9:07 Clip and cut aberrant r. artery 45 9:08 Resume G B D 66 9:09 97 9:10 G B slips out of graspers, regrasp G B 65 9:11 Camera problems 52 9:12 Resume G B D 58 9:13 1111111111I11I111I6JI: 9:14 75 9:15 "4 9:16 End GBD 93 9:17 Start gallbladder removal (GBR) 71 9:18 Surgeon leaves O R table briefly 53 9:19 Prepare specimen bag for G B 95 9:20 49 9:21 G B into bag 59 9:22 Waiting for irrigation tool, resting 33 9:23 52 9:24 T C - graspers to irrigation tool 52 9:25 Move camera to new port 54 9:26 Insert graspers to retrieve G B bag 56 9:27 Remove specimen bag w/ open tools 43 9:28 End GBR 36 65 1 150 300 450 Frame no. Figure 33: Frame-by-frame data status distribution over each minute of recorded surgery time. Occurrence periods are illustrated for the three major surgical events: cystic duct dissection (CDD), gallbladder dissection (GBD), gallbladder removal (GBR). Note again that the majority of missing or partial data, with the exception of cystic duct and artery clipping and cutting, occurs during non-surgical tasks such as tool changes. Figure 33 complements Table 9 with an illustration of the frame-by-frame data status distribution within each minute of recorded surgery time. In this figure, each row represents one minute of surgery time, corresponding to the rows of Table 9, where the status of each sample frame within that minute represent the cells along the row. Note that the number of frames per minute varies throughout the surgery, indicating variable sampling rates per minute (zero padding compensates for shorter segments in the image above). This problem stems from the current motion analysis software provided by Northern Digital Inc. which allows the frame update rate to increase when marker arrays are missing from view. Future improvements to the software will maintain a constant sampling rate. A further analysis was performed to examine the duration of gaps of missing data in the individual joint data. Here, the occurrence of partially complete data in a given sample frame was recorded only if there were other joint angle data available in the sample (see 'partial', Table 7). This type of partial corruption can not be attributed to non-surgical tasks or tool changes, but does not occur for substantially great lengths. The frequency distribution for missing data gap duration in a given joint angle within a partially complete data sample is presented in Table 10 below. 66 Table 10: Percent frequency of duration for partial samples in each joint angle from data collected during motion analysis. Duration of missing data by joim t (%) Duration: no. samples (sec.) Shoulder Elbow Wrist Forearm <=1 (0.067) 40 6 40 2 39.6 32 8 >1; <= 5 (0.33) 35.4 38.9 34.9 35.5 > 5; <= 10 (0.67) 12.6 10.8 10.6 12.4 > 10; <= 15 (1.00) 3.4 3.0 3.3 5.5 > 15; <=20(1.33) 1.7 2.0 2.0 2.5 > 20 (1.33) 6.3 5.1 9.5 11.4 The information gleaned from Table 10 is that the length of gaps in the data during periods of partial corruption is less than 0.33 seconds approximately 70% of the time in each joint. Thus, the effective sampling frequency becomes ~3Hz from 15Hz, which is still quite generous for a posture sampling analysis. Bhattacharya et al. (1999), for example, use a 1Hz sampling frequency on their ergonomic posture sampling dosimeter for assessing carpentry work. The implications of small gap duration is that the loss of data likely does not effect the essence of the surgical task postures captured by the analysis, and may therefore be ignored in calculating the overall percentage of time spent in various posture zones. Figure 34 illustrates the breakdown of data status for each marker array over the course of the surgery. These plots illustrate the problems with the choices of marker location. For example, the torso array is visible for nearly the entire surgery, while the proximal forearm is occluded for a good portion of the surgery. This type of visual analysis may be used in the future to optimize the marker location in order to maximum marker visibility and, therefore, data collection. Torso (F1) ' i .>" -"wff*""' ..'iiiwii^ iiiiiiiiiiiiiiii^ w S^wBf Distal forearm (F3) Proximal forearm (F2) D Complete | Missing | Zero padding Figure 34: Status of each joint angle sample over the surgery duration. Each row represents one minute of surgical time containing N samples which are either complete (white) or missing (black). Zero padding compensates for variable sampling frequency per minute. 67 4.4 Discussion The normalized ergonomic stress for each joint is presented in Figure 35; Here, the amount of time spent in ergonomically stressful postures outside of normal posture limits is counted as a contribution to overall ergonomic stress of a task, and weighted according to the posture scores presented in Table 4. Further, the maximum posture score for each joint is used as the baseline to normalize stress contribution. Thus, a maximally stressful posture adopted for the duration of a task receives a normalized score of 100% ergonomic stress, while adopting a neutral posture . for the same duration and task registers as 0%. 100% 80% in | 60% in c o I 40% c o o 20% Normalized Ergonomic stress Mean • CCD 39% I GBD 60% • GBR 45% Shoulder* Elbow Wrist f-e* Wrist r-u Forearm p-s Figure 35: Normalized, weighted postural stress contribution of each joint outside of normal, ergonomically safe posture range. Mean posture scores for each task over all joints are also illustrated (legend), though no significant difference was found between stress levels1. A significant difference was found between ergonomic stress in wrist flexion-extension (f-e) postures compared to shoulder postures, indicated by (*) (') Kruskal-Wallis single-factor nonparametric A N O V A by ranks, cc=0.05 As illustrated in Figure 35, the shoulder joint angle is subjected to the lowest postural stress throughout the surgery while the relative wrist stress is highest in each of the three surgical stages examined. A nonparametric analysis using a Kruskal-Wallis single-factor A N O V A by ranks is performed to test the null hypothesis that the ergonomic stress level is the same across all five joints (Zar 1999) (see Appendix E). In this analysis, each data value of percent normalized stress is assigned an integer rank and the analysis is performed much like a single-factor block A N O V A with one measurement per cell. Here, the null hypothesis of equal stress is rejected (ct-0.05, 0.02 < P < 0.01). Based on this, a nonparametric multiple comparisons test was performed between each group of joint stress (see Appendix E) and a significant difference between shoulder and wrist flexion-extension stress was found (a=0.05, 0.02 < P < 0.01). No significant differences were found between any other joint angles. The mean stress values over each surgical task suggest that GBD is the most stressful stage of the surgery, though another 68 Kruskal-Wallis test showed that there is no significant difference between the ergonomic stress levels in the three surgical tasks (a=0.05 ). The finding that posture of the wrist in flexion-extension is more stressing across all stages of this particular surgery than shoulder posture agrees with the common concern that laparoscopic instruments cause excessive wrist flexion (Berguer 1998a). Another note of interest is that our surgeon (AN) uses a palm gasp technique to reduce stress on the wrist during laparoscopic procedures. This is likely accounts for the lower stresses noted in Figure 35 due to radial-ulnar deviation relative to flexion-extension in each surgical task. Figure 36 below is presented to visualize the time evolution of the average normalized stress measure over the course of the surgery. Here, the percentage normalized stress is the combined mean of normalized stress for each joint. In Figure 36, every second minute is annotated for clarity. Full surgical task annotation is illustrated in Table 9. Start cystic duct dissection (CDD) Localize duct Clip and cut cystic artery Clip and cut cystic duct Start gallbladder dissection (GBD) GBD Localizing aberrant right artery GBD Clip and cut aberrant r. artery GBD Camera and instrument problems GBD Start gallbladder removal (GBR) Prepare specimen bag for GB GB into bag Waiting for irrigation tool, resting Insert graspers Remove specimen bag w/ open tools 1 0 0 % 5 0 % 0 % Average normalized stress level (%) Figure 36: Time evolution of average normalized stress over the duration of the surgery 69 Averaging the normalized stress, as illustrated in Figure 36, has the advantage Of showing the overall pattern of stress over the course of the surgery. For example, the relatively higher stress levels noted during the period of gallbladder dissection (GBD) may indicate the need for detailed analysis of this task and the instruments used in order to reduce ergonomic stress. The generality of the integrated measure, however, is also a disadvantage since we lose information regarding the stress contribution from particular joints. Thus an analysis which considers both individual and integrated measures may be most effective. Some problems we discovered in using this motion analysis system are worth noting. One problem was that the marker arrays needed to track the instruments, as originally proposed in the study protocol, need to be placed on the tools with each use and were thus cumbersome to deal with. Further, because of the surgeon's unconventional grip style, the tool marker arrays were often turned away.from the Polaris position sensor, and were therefore not tracked. Due to inconsistencies with tool marker placement and occlusion, it was decided to exclude tool motion from the present data analysis and concentrate on posture. Another problem was that the original intent of the study was to track the motion of both of the surgeon's arms and instruments. In two initially unsuccessful attempts at running the study, however, it was discovered that the Polaris position sensor could not be placed in such a way as to allow for consistent tracking of markers on both arms. In the successful attempt reported herein, it was decided to track the motions of only the surgeon's dominant arm and orient the position sensor in an optimal position in order to capture the maximum amount of arm motion data. Suggestions for future considerations to deal with these problems include developing a more robust method for instrument tracking, with the Polaris system or otherwise, and exploring options for successfully tracking the non-dominant arm and instrument. Options include more testing with single position sensor to determine a position for the unit which captures the maximum amount of data from both arms, adding another Polaris system, or investigating alternative systems which require on-site calibration but allow for greater freedom in positioning and. increasing the number of sensing cameras. Though no general inferences can be made from this pilot study, the potential for further ergonomic study with such a motion analysis system is illustrated. For example, an interesting study could compare the use of the unconventional palm grip against using the handle rings provided on standard pistol grip instruments in order to determine which technique has a lower impact on ergonomic stress, which may suggest new handle designs. Another study may compare the trade-off between increasing shoulder elevation angles in an effort to decrease wrist strain. The demonstrated advantage of using the modified RULA posture sampling scores presented in this study allows one to determine an overall normalized, weighted score for all joint angles, which provides for easy comparison between joints. The main potential of the motion analysis system is to develop simulations for training and assessment based on expert surgeon experience in the OR. Simulation has been defined as 'a technique of substituting a synthetic environment for a real one, so that it is possible to work under laboratory conditions of control' (Harman as quoted by Sanders 1991). Simulators are designed with different objectives in mind: training to develop skills, system design to reveal design deficiencies, and personnel assessment for predicting ultimate task performance of individuals (Sanders 1991). In the area of laparoscopic or minimally invasive surgery (MIS), most of the current simulator development is aimed at training and assessing new surgeons (Way 1995). Our primary interest as instrument designers, however, is using a simulation to assess system design; that is, to accurately assess and evaluate new instruments and combined toolsets for MIS in a controlled environment. We reason that improved tool designs which perform well 70 on accurate simulations of tasks representative of MIS will show a performance benefit during actual surgery. An important step in developing a accurate simulation is validation. The usual simulation validation method is to assess the transfer-of-training from simulator to reality based on some measure of subject performance. Mudd (1968) notes that the disadvantage of using transfer-of-training performance measures for simulator validation is that they do not indicate which functional elements of the simulation may be missing, distorted, or misleading to the trainee. From the system design viewpoint, the value of a MIS simulation depends on how closely it captures the elements of perceptual-motor skills used in actual surgery, not how well it develops those skills in trainees. Sanders (1991) notes that "direct correspondence is usually assessed when the major aim of the simulation concerns equipment design." On this basis, we propose to use a motion analysis system in studies of live minimally invasive surgeries, in order to develop surgical task simulations based on a measure of correspondence between subject responses in both situations. The subject responses measured by the motion analysis system are the coupled factors of surgical task, ergonomic stress level, and tool motion. 4.5 Conclusions The combined optical IR/video motion analysis system presented in this pilot study demonstrates the potential for continuous posture sampling of the surgeon's dominant arm in an ergonomic analysis of a laparoscopic cholecystectomy. On the whole, the system was found to be successful in accomplishing the goal of free motion tracking of the surgeon's dominant arm over the duration of a laparoscopic procedure. A method was presented and demonstrated for continuous posture sampling and ergonomic assessment during live surgeries in the OR. Though no inferences can be made from this one surgery, the advantages of using the motion analysis system for ergonomic studies and simulation validation are illustrated and discussed. 71 Chapter 5 Conclusions and Future Work 5.1 Introduction The goals of this project were to establish a foundation for research into the design and assessment of improved instruments for minimally invasive surgery. Minimally invasive surgery has exploded into general surgical practice in the last decade with the introduction of the laparoscopic cholecystectomy. Though beneficial to patients, the new technique is complicated by physical and mental challenges which make it difficult for surgeons to master. Improved instrumentation and teleoperators have the potential to ease the strain of performing the technique and improve surgeon performance. Initial experimentation on the effects of physical constraints on surgeon performance suggested that an economical and effective solution is to develop a teleoperator device constrained to 4 degrees of freedom which restores the natural motion mapping between the hand and instrument tip. A prototype teleoperator was constructed to test this theory. Fundamental problems with the novel drive mechanism, however, prevent the prototype from functioning satisfactorily, leaving the project open to future improvements. In order to effectively test for real improvements in surgeon performance, we also proposed a motion analysis system for measuring surgeon performance during actual surgeries. This system will be employed in future studies to validate surgical simulations in the laboratory which may be used to test and develop new instruments. A pilot study was performed to assess the potential of using the system to track the unconstrained motion of the surgeon's dominant arm over the course of a laparoscopic cholecystectomy. An ergonomic posture sampling study was performed to show the usefulness of the system. 5.2 Review of present research 5.2.1 Effects of physical constraints In this experiment, we set out to determine how constrained motion affects the surgeon performance in order to establish guidelines for the development of advanced MIS instrumentation and teleoperators. Using a simulation of a totally transparent teleoperator system (direct vision, direct force and motion scaling), we assessed task completion times for suturing and pick-and-place tasks under three open tool and two laparoscopic tool configurations with various degrees of freedom: 6, 5 and 4 DOF open and 6 and 4 DOF laparoscopic. These configurations represent the most likely candidates for teleoperated devices. Results of tests on 15 surgeons of various expertise show that avoiding reversed hand-tool motion mapping and increasing the number of DOF significantly increases suturing performance. Avoid reversed motion also significantly improves pick-and-place task performance, though adding DOF to the open tool configuration showed no improvement. In correlating performance improvements on simulated tasks to the relative time spent performing similar tasks in surgery, we concluded that the most useful and economic configuration for a teleoperator is one which emulates the feeling of using an open tool and has 4 DOF motion at the instrument tip. 5.2.2 Mechanical teleoperator Based on the results of the physical constraints study, we proposed to develop a four (4) degree of freedom mechanical teleoperator for MIS. The device was designed to mechanically couple tool motion from the master handle controlled by the surgeon to the slave tip within the abdomen. The decision to design a totally mechanical teleoperator was twofold. Firstly, to create a simple device which would theoretically show the same performance benefit as expensive electro-mechanical robot systems yet be economical to manufacture an affordable for health institutions. Secondly, eliminating electrical parts from the teleoperator improves the opportunity for the device to forego intense approval testing and be rapidly adopted into mainstream surgical practice. The prototype device developed for this thesis features a parallel linkage design to produce tool pitch and yaw relative to a fixed fulcrum. The fixed fulcrum is positioned at the entry point into the abdomen and acts as a passive safety mechanism, preventing lateral strain on the abdominal wall. Another unique design feature of the teleoperator is the friction drive mechanism for tool plunge and roll, developed as a solution for reducing the moving mass of the teleoperator about the pivoting axis. In practice, it was found that the friction drive mechanism was inadequate for effective and reliable motion transmission from master to slave. Recommended testing of various friction coatings for the drive ball may yet prove the theory of the friction drive to be applicable for this device. The base of the teleoperator may serve for future prototypes. 5.2.3 Motion analysis study In order to initiate the development of motion analysis for surgical task simulation validation, a pilot study was performed sampling surgeon posture during a laparoscopic cholecystectomy in the OR. Here, the Polaris optical tracking system, combined with video analysis, was used to produce a continuous ergonomic posture sample of the surgeon's dominant arm throughout the course of the surgery. Joint angle measurements for the shoulder, elbow, forearm and wrist were sampled and categorized according to a modified RULA ergonomic posture sampling grade scale based on the severity of stress caused by the posture. Results showed a significantly higher contribution of stress due to wrist flexion-extension compared to shoulder elevation in the 73 surgery studied. The success of implementing a continuous motion analysis during live surgery shows promise for future simulation validation and ergonomic studies. 5.3 Future research recommendations 5.3.1 Mechanical teleoperator The failure of the teleoperator prototype to produce suitable plunge and roll motion coupling between the master and slave tools lends itself to several recommendations for future work. • Continued evaluation of a variety of friction surfaces is recommended. A more suitable coating may be available which would act as a more effective friction drive, in which case the device may be shown to work effectively. • Examine alternative designs for plunge-roll motion transmission which do not rely on friction drive. The current teleoperator is effective for transmitting pitch-yaw, so the base of this prototype may be salvaged to construct a second generation prototype teleoperator with an improved plunge-roll motion transmission mechanism. • Test for improved performance with teleoperator over conventional MIS instruments. 5.3.2 Motion analysis system , Several recommendations were derived from the motion analysis study in the OR. The suggestions for future work in this area are: • Develop a constraining jig for the wrist to use during the wrist calibration process. This will help eliminate some of the problems encountered with 'free motion' calibration where we rely on the subject to constrain their posture and motions during the F-E and R-U tasks. • Examine the potential to improve shoulder joint centre tracking by including a marker to track motion of the acromion (following Schmidt 1999). • Determine a definition for shoulder joint angle which is both robust for free motion analysis and may be related to ergonomically stressful or important postures. Standard Eulerian methods, floating axis methods (Grood & Suntay 1983) or latitude-longitude measures (An 1991) may prove to be appropriate. Quaternion (screw axis, angle-axis) methods are also available, thought they are not readily interpreted physically (Grood & Suntay 1983). • Examine performance measures such as measures of joint angular velocity and acceleration, following Emam (1999). • Improve instrument motion tracking methods by improving on the Polaris marker arrays, or by a recommending a new system-for tool motion tracking. • Automate the identification of instrument movement in order to improve interoperative task tracking. • Study alternative positions for the Polaris position sensor within the OR in order to maximize the amount of useful data captured by the system during surgery. 74 5.4 Future studies: Performance measures and simulation validation As mentioned previously, our long-term research goal is to develop validated simulations of surgical tasks based on motion analysis studies in the OR, and use these controlled simulations as test-beds for improved MIS instrumentation and toolsets. It is our belief that designing new instruments to improve surgeon performance on well validated simulations of actual surgeries will allow us to infer real performance gains in the OR before testing the instruments there. The advantages of iterative design improvements and rapid prototyping are also attractive reasons for developing such simulations. The question which arises, however, is what measure of 'performance' should be used in the initial study of the OR which would allow us not only to track important aspects of surgeon behaviour in the OR, but provide a good measure of comparison in validating laboratory simulations of real surgical tasks. The concept of performance or skill measurement has become a relevant topic in the area of minimally invasive surgery. Currently, most researchers who are involved in developing training programs or simulations propose some measure of performance. Generally, task completion time is employed as the performance measure in evaluating skill improvement (Chung & Sackier 1998) as it is straightforward to measure and is directly effected by the task being performed (Sanders 1991). Though simple to measure and compare, task completion time may not sufficiently describe the complete skill set required in performing a task. Derossis et al. (1998) address this concern with a performance score that rewards both precision and speed. Hanna and Cuschieri have employed various measures of skill based on knot quality (Hanna 1997), economy of arm motion (Emam 1999), instrument trajectory, errors and task times (Hanna 1998), and error propagation and human reliability in actual surgeries (Joice 1998). Any one of these measures may show differences due to various treatments in a skills testing experiment, but the question is which measure, or combination of measures, captures the relevant aspects of real surgery. If a measure does not correlate to actual surgery, then conclusions drawn in simulations may be irrelevant in practice (Mudd 1968, Sanders 1991). To answer this question', we propose to first examine the skills surgeons call upon during actual operations in the OR and discritize them into similar tasks in the lab. If one accepts the learning theory that skills transfer to a real situation through practice of similar discrete training tasks, then it is reasonable to assume that by decomposing the complex movements of the surgeon into discrete skill sets, we will be better able to evaluate which skill set is most influential in performing.surgery. Knowing this, we may address what mechanical solutions may lend themselves to complementing or enhancing these skills. 5.4.1 Simulation validation There are currently two approaches to MIS simulations seen in the literature. In the first case, researchers are developing complex simulations which aim to immerse the subject into a highly realistic 'virtual world' (Satava 1993, Kuhnaphel 1997, Playter 1997). In the other case, researchers are developing partial simulations which address basic perceptual-motor tasks using either computer assisted assessment (Hanna 1998, Wilson 1997) or direct assessment of simplified MIS tasks on desk-top trainers (Sackier 1991, Derossis 1998). Sanders (1991) notes that human performance measurement in partial simulations is easier to analyze but the 75 simulations have low fidelity and realism, while highly complex and realistic simulations are less open to analysis. Given our ultimate interest in evaluating instrument and system design, it appears more desirable to assess task performance in a setting where the outcome measures can be objectively controlled, as in the partial simulation case. The question raised is whether partial simulations adequately capture or reflect the essential motor-skills necessary to perform a given task, or if there is some advantage to the highly complex system which may not be immediately apparent. Sanders (1991) states, however, that "it is perfectly feasible that... a [highly realistic] simulation still misses some essential characteristic, rendering it useless as training device." Indeed, Allen (1986) showed that a highly physically realistic simulation does not necessarily imply a successful simulation. What is important is that the simulator captures the behaviour of the subject in the actual situation (Mudd 1968). Thenmportant issue in simulator design is then one of functional realism, more than physical realism (Allen 1986). Though current MIS simulations have face, validity to actual surgical tasks, how can one assess if the surgeon actions on a simulator are functionally similar to reality? This then becomes an issue of simulation validation based on reality. The usual validation method is to assess the transfer-of-training from simulator to reality based on some measure of subject performance. Mudd (1968) notes that the disadvantage of using transfer-of-training performance measures for simulator validation is they do not indicate which functional elements of the simulation may be missing, distorted, or misleading to the trainee. From the system evaluation viewpoint, the value of a MIS simulation depends on how closely it captures the elements of perceptual-motor skills used in actual surgery^ not how well it develops those skills in trainees. For example, trainees may be able to tie faster sutures after days of practice on a simulator, but the time measure alone does not tell what aspect of trainee behaviour improved. Sanders (1991) notes that when the major aim of the simulation concerns equipment design rather than training effectiveness, direct correspondence is usually assessed. Thus, we propose to study actual minimally invasive surgeries, and build discrete task simulations based on a measure of correspondence between subject responses in both situations. The question raised is, what should be considered good measures of correspondence? 5.4.2 Performance and correspondence measures In order to capture the essence of the surgical skills in a simulation, we need to develop and define performance measures that can be measured in both the OR and laboratory as a gauge for correspondence between the two environments. The problems of performance measurement, however, are dependant on the skills one is interested in capturing. Sanders (1991) defines three issues which are important to performance measurement: 1. Nature of the skill: Is the skill more perceptual-motor like driving a car in heavy traffic, or cognitive, like playing chess? The nature needs to be defined before an appropriate overt measure can be established. 2. Extent of timelock: In composite task analysis, it can be difficult to distinguish what stimulus triggered a given response if too much is occurring simultaneously. 76 3. Optimal performance definition: In every simulation, there is a need to define performance criterion with respect to expertise. It is a problem that expertise in many complex skills cannot be objectively evaluated (Sanders 1991). Commonly called "hand-eye" coordination, perceptual-motor skills are easiest to measure since the timelock is small between outside events and subject activity. This is one reason why task completion time and accuracy are standard output performance measures for simulations aimed at developing motor skills. More complex aspects of perceptual-motor skills can be measured, as long as the correct performance aspects are taken into consideration during simulation design (Sanders 1991). It is instructive at this point to focus on the distinction between work method and performance (Kjellberg 1998), or motion skill and motion pattern (Harrow 1972). Consider the example of lifting a box to waist level presented by Kjellberg (1998). Two different methods for lifting the box are to bend at the knee while lifting (back straight), or bend at the waist. Measures of performance in executing this task could be speed, accuracy, minimizing muscle effort (economy of motion), deviation from the prescribed method (error), or any number of metrics that have been suggested in literature. Harrow defines 'skill' as development of a "certain degree of proficiency or mastery", while 'pattern' is "the acceptable or recognizable performance of a movement for which the outcome alone is important" (Harrow 1972, p.76). Thus, what we wish to accomplish with simulator validation through correspondence is to transfer the methods of performing tasks from OR to simulation. Blaauw (1982) validated a driving simulation by assessing the correspondence between subject behavior responses to real and simulated driving. Outcome measures included subjective subject questionnaire responses and variable physical measurements (e.g. velocity, lateral control, steering-wheel angle) on both an instrumented car and a driving simulation. High correlation between individual behaviour on the simulator and instrumented car, especially with experienced drivers, indicates good validity of the simulation. In our examination of surgical tasks, we require a similar measure of subject behavioral response to both live surgery and controlled simulations which can be readily measured and interpreted. Given that we have established the feasibility of using a motion analysis system during live operations, measures of tool motion, arm motion and posture coupled with specific surgical tasks are all potential metrics for examining work method. Future experimentation will determine which of these measures, or combination thereof, best captures the work method. Once the method is transferred from OR to simulation, we will revisit the data and develop a new and more comprehensive measure for performance, in which case, more complex measures such as error rate and motion quality may also be assessed in defining surgical skill. Salvendy and Pilitsis (1980) performed a validation of three new medical suturing training models by assessing transfer-of-training based on 'first shot' performance measures. The interesting point of this study was that they measured not only performance time and quality as outcomes measures, but also autonomic responses to stress (heart rate, E M G levels, etc.) in an integrated performance index. They argue, as we do, that similar quality of performance does not necessarily indicate the advantages of adopting one method or another method. In some cases, the detrimental long term effects of one task execution method over another may be as, or more important than, short term similarities in quality of performance. In future analyses, we propose to use posture as an outcome measure for indicating the advantage of a given method. Bibliography Aaras, A., Westgaard, R.H. & Stranden, E. (1988). Postural angles as an indicator of postural load and muscular injury in occupational work situations. Ergonomics 31, 915-933. Alarcon, A. & Berguer, R. (1996). A comparison of operating room crowding between open and laparoscopic operations. Surgical Endoscopy 10, 916-919. Allen, J.A., Hays, R.T. & Buffardt, L. (1986). Maintenance training simulator fidelity and individual differences in transfer of training. Human Factors 28, 497-509. Amaral, J. and Levine, A. (1994). Ergonomic evaluation of a new rotating, dual position laparoscopic instrument handle. In Proc 16th IEEE Engineering in Medicine and Biology Ann. Conf. (Sheppard, N. F., Jr., Eden, M. , and Kantor, G., eds.), pp. 315-316. An, K.-N., Browne, A.O., Korinek, S., Tanaka, S. & Morrey, B.F. (1991). Three-dimensional kinematics of glenohumeral elevation. Journal of Orthopaedic Research 9, 143-149. Anglin, C. (1993) A functional task analysis and motion simulation for the development of a powered upper-limb orthosis. M.A.Sc..Thesis, University of British Columbia, Vancouver, BC. . Balazs, M . , Feussner, H., Hirzinger, G., Omote, K. & Ungeheuer, A. (1998). A new tool for minor-access surgery. IEEE Eng.Med.Biol.Mag. 17,45-48. Berguer, R. (1997). The application of ergonomics in the work environment of general surgeons. Reviews on Environmental Health 12, 99-106. Berguer, R. (1998). Surgical technology and the ergonomics of laparoscopic instruments. Surgical Endoscopy 12,458-462. Berguer, R., Forkey, D.L. & Smith, W.D. (1999). Ergonomic problems associated with laparoscopic surgery. Surgical Endoscopy 13,466-468. Berguer, R., Gerber, S., Kilpatrick, G. & Beckley, D. (1998). An ergonomic comparison of in-line vs pistol-grip handle configuration in a laparoscopic grasper. Surgical Endoscopy 12,805-808. Berguer, R., Rab, G.T., Abu-Ghaida, H., Alarcon, A. & Chung, J. (1997). A comparison of surgeons' posture during laparoscopic and open surgical procedures. Surgical Endoscopy 11, 139-142. Berguer, R., Remler, M . & Beckley, D. (1997). Laparoscopic instruments cause increased forearm fatigue: a subjective and objective comparison of open and laparoscopic surgery. Minimally Invasive Therapy and Allied Technologies 6, 36-40. 78 Bhattachaya, A., Warren, J., Teuschler, J., Dimov, M . , Medvedovic, M . & Lemasters, G. (1999). Development and evaluation of a microporcessor-based ergonomic dosimeter for evaluationg carpentry tasks. Applied Ergonomics 30, 543-553. Blaauw, G.J. (1982). Driving experience and task demands in simulator and instrumented car: A validation study. Human Factors 24, 473-486. Boehm, D.H., Reichenspurner, H., Gulbins, H., Detter, C., Meiser, B., Brenner, P., Habazettl, H. & Reichart, B. (1999). Early experience with robotic technology for coronary artery surgery. Ann.Thorac.Surg. 68, 1542-1546. Brown, C , Reeves, C , Loge, A (1998) "Design of a tool tip and handle for the mechanical teleoperator for laparscopic surgery" M E C H 455 Design report. The University of British Columbia, Vancouver, British Columbia, Canada. Buess, G. (1997). SMITROB and other tasks of SMIT. Minimally Invasive Therapy and Allied Technologies 6, 1 Buess, G.F., Mentges, B., Weiss, O. & Mailander, L. (1997). Training courses in Tuebingen. Minimally Invasive Therapy and Allied Technologies 6, 6-13. Cao, C. G. L. (1996) A Task Analysis of Laparoscopic Surgery: Requirements for Remote Manipulation and Endoscopic Tool Design. M.A.Sc. Thesis, Simon Fraser University, Burnaby, BC. Cao, C. G. L., MacKenzie, C. L., Ibbotson, J. A., Turner, L. J., Blair, N. P., and Nagy, A. G. (1999) Hierarchical decomposition of laparoscopic procedures. In Proc. Medicine Meets Virtual Reality: 7. IOS Press, Amsterdam. Chung, J.Y. & Sackier, J.M. (1998). A method of objectively evaluating improvements in laparoscopic skills. Surgical Endoscopy 12,1111-1116. Claus, G.P., Sjoerdsma, W., Jansen, A. & Grimbergen, C A . (1995). Quantitative standardised analysis of advanced laparoscopic surgical procedures. Endosc.Surg.Allied.Technol. 3, 210-213. Coope, I.D. (1993). Circle fitting by linear and nonlinear least squares. J. Optimization Theory & Application 76, 381-388. Cox, M.R., McCall, J.L., Toouli, J., Padbury, R.A., Wilson, T.G., Wattchow, D.A. & Langcake, M . (1996). Prospective randomized comparison of open versus laparoscopic appendectomy in men. World J.Surg. 20, 263-266. Dautzenberg, P., Neisius, B., Trapp, R., and Buess, G. (1995) A Powered Dexterous Instrument with Surgical Effectors for Telemanipulator Assisted Laparoscopy. In Proc. 17th IEEE Engineering in Medicine and Biology Int. Conf. Davis, J. and Darryn, G. (1998) Proposal for the Design of a Teleoperator for Laparoscopic Surgery. Internal document. Department of Mechanical Engineering, University of British Columbia, Vancouver, BC. 79 Derossis, A . M . , Bothwell, J., Sigman, H.H. & Fried, G.M. (1998). The effect of practice on performance in a laparoscopic simulator. Surgical Endoscopy 12, 1117-1120. Derossis, A . M . , Fried, G.M., Abrahamowicz, M . , Sigman, H.H., Barkun, J.S. & Meakins, J.L. (1998). Development of a model for training and evaluation of laparoscopic skills. Am.J.Surg. 175, 482-487. Drown, M . , Oshika, J., Schab, L, Schmidt, G., and Smith, D. (1998) Design of a 4-Degree-of-Freedom Mechanical Teleoperator for Laparoscopic Surgery. Internal document. Department of Mechanical Engineering, University of British Columbia, Vancouver, BC. Dunlap, K.D. & Wanzer, L. (1998). Is the robotic arm a cost-effective surgical tool? AORN Journal 68,265-272. r Dussaults, C. (1998) A 4-DOF Teleoperator for Laparoscopic Surgery. Internal Report. 1998. Internal document, Department of Mechanical Engineering, University of British Columbia, Vancouver, BC. Eldridge, B., Gruben, K., LaRose, D., Funda, J., Gomory, S., Karidis, J., McVicker, G., Taylor, R.H. & Anderson, J. (1996). A remote center of motion robotic arm for computer assisted surgery. Robotica 14, 103-109. Emam, T.A., Frank, T.G., Hanna, G.B., Stockham, G. & Cuschieri, A. (1999). Rocker handle for endoscopic needle drivers. Technical and ergonomic evaluation by infrared motion analysis system. Surgical Endoscopy 13, 658-661. Faraz, A. & Payendeh, S. (1997). Synthesis and Wordspace Study of Endoscopic Extenders With Flexible Stem. Journal of Mechanical Design 119, 412-414. Finlay, P. A. & Ornstein, M.H. (1995). Controlling the Movement of a Surgical Laparoscope. IEEE Engineering in Medicine and Biology 289-291. s Fisher, K.S., Reddick, E.J. & Olsen, D.O. (1991). Laparoscopic cholecystectomy: cost analysis. Surgical Laparoscopy & Endoscopy. 1,77-81. Frank, T.G. & Cuschieri, A. (1997). Prehensile atraumatic grasper with intuitive ergonomics. Surgical Endoscopy 11, 1036-1039. Fried, G.M., Derossis, A . M . , Bothwell, J. & Sigman, H.H. (1999). Comparison of laparoscopic performance in vivo with performance measured in a laparoscopic simulator. Surgical Endoscopy 13, 1077-1081. Funda, J., Eldridge, B., Gruben, K., and Taylor, R. H. (1994) Comparison of two manipulator designs for laparoscopic surgery. In SPIE 2351- The international society for optical engineering. Telemanipulator and Teleprescence Technologies (Das, H., ed.), pp. 172-183. Funda, J., Taylor, R.H., Eldridge, B., Gomory, S. & Gruben, K. (1996). Constrained cartesian motion control for teleoperated surgical robots. IEEE Transactions on Robotics and Automation 12, 453-465. 80 Garcia-Ruiz, A., Gagner, M. , Miller, J.H., Steiner, C P . & Hahn, J.F. (1998). Manual vs robotically assisted laparoscopic surgery in the performance of basic manipulation and suturing tasks. Archives of Surgery. 133,957-961. Genaidy, A . M . , Al-Shedi, A.A. & Karwowski, W. (1994). Postural stress analysis in industry. Applied Ergonomics 25, 77-87. Green, P.S., Hill, J.W., Jensen, J.F. & Shah, A. (1995). Telepresence Surgery. In Proc. 17th IEEE Engineering in Medicine and Biology Int. Confi, pp. 324-329. Gupta, V., Reddy, N. P., and Batur, P. (1996). Forces in surgical tools: Comparison between laparoscopic and surgical forceps. In Proc. 18th IEEE Engineering in Medicine and Biology Int. Conf. (no pg.) Halvorsen, K., Lesser, M . & Lundberg, A. (1999). A new method for estimating the axis of rotation and the center of rotation. J. Biomechanics 32, 1221-1227. Hanna, G.B., Drew, T., Clinch, P., Hunter, B. & Cuschieri, A. (1998). Computer-controlled endoscopic performance assessment system. Surgical Endoscopy 12, 997-1000. Hanna, G.B., Frank, T.G. & Cuschieri, A. (1997). Objective assessment of endoscopic knot quality. American J. Surgery. 174,410-413. Harrow, A.J. (1972). A taxonomy of the psychomotor domain. Longman, New York. Hasson, H.M. (1993). Rotational handle for laparoscopic instrumentation. J. Reproductive Medicine. 38, 494-496. Herder, J.L., Horward, M.J. & Sjoerdsma, W. (1997). A laparoscopic grasper with force perception. Minimally Invasive Therapy and Allied Technologies 6, 279-286. Hill, J. W., Green, P. S., Jensen, J. F., Gorfu, Y., and Shah, A. S. (1994). Telepresence surgery demonstration system. In Proc. IEEE Int. Conf. on Robotics and Automation, pp. 2302-2307. Hodgson, A. J., Pantazopol, R. A., Visser, M . D., Salcudean, S. E. , and Nagy, A. G. (1997). Assessing potential benefits of enhanced dexterity in laparoscopic surgery. In Proc. 19th Ann. IEEE Engineering in Medicine and Biology Society Conference, Chicago. Hoening, D.M., Shalhav, A.L. , rcangeli, C.G., trander, D.D., lbahnasy, A . M . & layman, R.V. (1997). Under-table mounting for the AESOP robot for laparoscopic flank surgery. Minimally Invasive Therapy and Allied Technologies 5/6, 460-462. Horgan, L.F., O'Riordan, D.C. & Doctor, N. (1997). Neuropraxia following laparoscopic procedures: an occupational injury. Minimally Invasive Therapy and Allied Technologies 6,33-35. Howe, R.D., Peine, W.J., Kontarinis, D.A. & Son, J.S. (1995). Remote palpitation technology. IEEE Engineering in Medicine and Biology 318-323. 81 Hu, J., Ren, J., and Sheridan, T. (1996). Telerobotic surgery: Stable force feeback with time delay. In SPIE 2901 - The international society for optical engineering. Telemanipulator and Telepresence Technologies 3. (Stein, M . R., ed.), 22-33. Hunter, J.G. (1997). The Learnining Curve in Laparoscopic Cholecystectomy. Minimally Invasive Therapy and Allied Technologies 6, 24-25. Joice, P., Hanna, G.B. & Cuschieri, A. (1998). Errors enacted during endoscopic surgery - a human reliability analysis. Applied Ergonomics 29, 409-414. Kant, I.J., de, J.L., van Rijssen-Moll, M. & Borm, P.J. (1992). A survey of static and dynamic work postures of operating room staff. Int. Arch. Occup.Environ. Health 63, 423-428. Karhu, O., Kansi, P. & Kuorinka, I. (1977). Correcting working postures in industry: A practical method for analysis. Applied Ergonomics 8, 199-201. Keyserling, W.M. (1986). A computer-aided system to evaluate postural stress in the workplace. • Am.Ind.Hyg.Assoc. J. 47, 641-649. Kim, C.A. & Smith, C D . (1999). Manual vs robotically assisted laparoscopic surgery in the performance of basic manipulation and suturing tasks. Surgical Endoscopy 13, 723 Kjellberg, K., Lindbeck, L. & Hagberg, M . (1998). Method and performance: two elements of work technique. Ergonomics 41, 798-816. Kroemer, K.H.E. (1989). Cumulative trauma disorders: Their recognition and ergonomics measures to avoid them. Applied'Ergonomics 20,274-280. Kiihnaphel, U.G., Kuhn, Ch., Hubner, M . , Krumm, H.-G., Maass, H. & Neisius, B. (1997). The Karlsruhe Endoscopic Surgery Trainer as an example for virtual reality in medical education. Minimally Invasive Therapy and Allied Technologies 6, 122-125. MacKenzie, C. L. , Graham E.D., Cao C.G.L, and Lomax A.J. (1999). Virtual Hand Laboratory Meets Endoscopic Surgery. In Proc. Medicine Meets Virtual Reality: 7. IOS Press, Amsterdam. Matern, U. & Waller, P. (1999). Instruments for minimally invasive surgery: principles of ergonomic handles. Surgical Endoscopy 13, 174-182. Matern, U., Eichenlaub, M. , Waller, P. & Ruckauer, K. (1999). MIS instruments. An experimental comparison of various ergonomic handles and their design. Surgical Endoscopy 13, 756-762. McAtamney, L. & Corlett, E.N. (1993). RULA: A survey method for the investigation of work-related upper limb disorders. Applied Ergonomics 24, 91-99. Melzer, A., Schurr, M.O., Kunert, W., Buess, G:, Voges, U. & Meyer, J.U. (1993). Intelligent surgical instrument system ISIS. Concept and preliminary experimental application of components and prototypes. Endosc.Surg.Allied.Technol. 1, 165-170. 82 Mudd, S. (1968). Assessment of the fidelity of dynamic flight simulators. Human Factors 10, 351-358. Mueglitz, J., Kunad, G., Dautzenberg, P., Neisius, B. & Trapp, R. (1993). Kinematic problems of manipulators for minimal invasive surgery. Endosc.Surg.Allied.Technol. 1, 160-164. Mueller, L.P. (1993). Laparoscopic instrument grips. An ergonomic approach [letter]. Surgical Endoscopy 7, 465-466. Mukherjee, R., Song, G., and Satava, R. M . (1996). An Articulating Manipulator for Enhanced Dexterity in Minimally Invasive Surgery. In Proc. 18th IEEE Engineering in Medicine and Biology Int. Conf. Nagai, H. & Araki, S. (1999). Semiautomatic suturing device (Maniceps) for laparoscopic surgery. Surgical Endoscopy 13, 191-193. Neisius, B., Dautzenberg, P., Trapp, R., and Buess, G. (1995) Robotic Telemanipulator for Laparoscopy. In Proc. 17th IEEE Engineering in Medicine and Biology Int. Conf. Neuhaus, S.J. & Watson, D.I. (1997). Laparoscopic surgeons' thumb - is it a training phenomenon? Minimally Invasive Therapy and Allied Technologies 6, 31-32. Ohgami, M . (1998). Robotics in Endoscopic Surgery. Unpublished presentation at Society of American Gastrointestinal Endoscopic Surgeons (SAGES) Annual Meeting, Seattle, WA, April 1-4,1998. Ortega, A.E. , Hunter, J.G., Peters, J.H., Swanstrom, L.L. & Schirmer, B. (1995). A prospective, randomized comparison of laparoscopic appendectomy with open appendectomy. Laparoscopic Appendectomy Study Group. Am.J.Surg. 169, 208-212. Ottensmeyer, M . P., Thompson, J. M . , and Sheridan, T. B. (1996). Telerobotic Surgery: Experiments and Demonstration of Tele-Surgeon/Assistant Cooperation Under Different Time Delays and Tool Designs. In SPIE 2901 - The international society for optical engineering. Telemanipulator and Telepresence Technologies 3. (Stein, M. R., ed.), pp. 156-166. Patkin, M . & Isabel, L. (1995). Ergonomics, engineering and surgery of endosurgical dissection. JR. Coll.Surg.Edinb. 40, 120-132. Pellegrini, C A . & Sinanan, M.N. (1997). Training, proctoring, credentialing in endoscopic surgery. Minimally Invasive Therapy and Allied Technologies 6, 26-30. Peshkin M . and Colgate J.E. (1996) Cobots work with people. IEEE Robotics & Automation Magazine 3(4), (no pp.) Peterson, B. & Palmerud, G. (1996). Measurement of upper extremity orientation by video stereometry system. Med.Biol.Eng Comput. 34, 149-154. Playter, R. & Raibert, M. (1997). A virtual reality surgery simulator using advanced haptic feedback. Minimally Invasive Therapy and Allied Technologies 6,117-121. 83 Radermacher, K., von Pichler, K. C , Erbse, St, Boeckmann, W., Rau, G., Jaske, G., and Staudte, H.-W. (1996). Using human factor analysis and VR simulation techniques for the optimization of the surgical worksystem. In Medicine Meet Virtual Reality: 4: Health Care in the Information Age (Sieburg, H., Weghorst, S., and Morgan, K., eds.), IOS Press and Ohmsha, Amsterdam, NL. Rahman T., Ramanathan R., Seliktar R. and Harwin W. (1995) A simple technique to passively gravity-balance articulated mechanisms. Journal of Mechanical Design 117, 655-658. Rau, G., Radermacher, K., Thull, B. & von Pichler, C. (1996). Aspects of ergonomic system . design applied to medical work systems. In Computer Integrated Surgery: Technology and clinical applications (Taylor, R.H., LavallUe, S., Burdea, G.C. and Mosges, R., eds.), The MIT Press, Cambridge, M A , pp. 203-221. Reinhardt-Rutland, A. H. and Gallagher, A. G. (1995) Visual depth perception in minimally invasive surgery. In Contemporary Ergonomics (Robertson, S. A., ed.), Taylor & Francis Rovetta, A., Sala, R., Cosmi, F., Wen, X., Milanesi, S., Sabbadini, D., Togno, A., Angelini, L.*& Bejczy, A.K. (1996). A New Telerobotic Application: Remote Laparoscopic Surgery Using Satellites and Optical Fiber Networks for Data Exchange. International Journal of Robotics Research 15, 267-279. Rovetta, A., Sala, R., Wen, X., Cosmi, F., Togno, A. & Milanesi, S. (1995). Telerobotic surgery project for laparoscopy. Robotica 13, 397-400. Sackier, J.M. & Wang, Y. (1994). Robotically assisted laparoscopic surgery. From concept to development. Surgical Endoscopy 8,63-66. Sackier, J.M., Berci, G. & Paz-Partlow, M. (1991). A new training device for laparoscopic cholecystectomy. Surgical Endoscopy 5, 158-159. Salvendy, G. & Pilitsis, J. (1980). The development and validation of an analytical training program for medical suturing. Human Factors 22, 153-170. Sanders, A.F. (1991). Simulation as a tool in the measurement of human performance. Ergonomics 34 , 995-1025. Santos-Munne, J. J. (1996) Stereotactic system for linear-trajectory medical interventions. M.Sc. Thesis, North Western University, Chicago, IL, USA. Satava, R. M . (1994). Telemanipulation, telepresence and virtual reality for surgery in the year 2000. In SPIE 2351- The international society for optical engineering. Telemaipulator and telepresence Technologies (Das, H., ed.), pp. 162-169. Satava, R.M. (1993). Virtual reality surgical simulator: The first steps. Surgical Endoscopy 7, 203-205.' Satava, R.M. (1994). The modern medical battlefield: Sequitur on advanced medical technology. IEEE Robotics and Automation Magazine Sept., 21-25. 84 Schmidt, R., Disselhorst-Klug, C , Silny, J. & Rau, G. (1999). A marker-based measurement procedure for unconstrained wrist and elbow motions. J. Biomechanics 32, 615-621. Schurr, M.O. & Buess, G. (1995). OREST II~ergonomic workplace and systems platform for endoscopic technologies. Endosc.Surg.Allied.Technol. 3, 193-198. Schurr, M.O., Buess, G.F. & Heyn, S.P. (1997). Task-force robotics and microsystems in minimally invasive therapy - SMITROB - or the Society for Minimally Invasive Therapy. Minimally Invasive Therapy and Allied Technologies 6, 2-5. Sciavicco, L. & Siciliano, B. (1996). Modeling and control of robot manipulators. McGraw-Hill, New York. Shell, B. and Payandeh, S. (1996) Endoscope. Integrated System based project (IS-9) of the Institute of Robotics and Intelligent Systems, Simon Fraser University, Burnaby, BC. Sjoerdsma, W., Herder, J.L., Horward, M.J., Jansen, A., Bannenberg, J.J.G. & Grimbergen, C A . (1997). Force transmission of laparoscopic grasping instruments. Minimally Invasive Therapy'and Allied Technologies 6,21'4-278. Smith, J.F., Boysen, D., Tschirhart, J., Williams, T. & Vasilenko, P. (1992). Comparison of laparoscopic cholecystectomy versus elective open cholecystectomy. J.Laparoendosc.Surg. 2, 311-317. Stringer, N.H., Walker, J . C & Meyer, P.M. (1997). Comparison of 49 laparoscopic myomectomies with 49 open myomectomies. J.Am.Assoc. Gynecol.Laparosc. 4, 457-464. Swain, C P . (1997). Endoscopic sewing and stapling machines. Endoscopy 29, 205-210. Taylor, R. H., Funda, J., LaRose, D., and Treat, M . (1992) A telerobotic system for augmentation of endoscopic surgery. In Proc 14th IEEE Engineering in Medicine and Biology Ann. Conf, pp. 1054-1056. Taylor, R.H., Funda, J., Eldridge, B., Gomory, S., Gruben, K., LaRose, D., Talamini, M . , Kavoussi, L. & Anderson, J. (1995). A Telerobotic Assistant for Laparoscopic Surgery. IEEE Eng.Med.Biol.Mag. 279-288. Taylor, R.H., Funda, J., Joskowicz, L., Kalvin, A.D., Gomory, S., Gueziec, A.P. & Brown, L.M.G. (1996). An overview of computer-integrated surgery at the IBM Thomas J. WatsonResearch Center. IEEEEng.Med.Biol.Mag. 40, 163-183. Tendick, F. and Cavusoglu, M . C. (1997). Human-machine interfaces for minimally invasive surgery. In Proc. 19th IEEE Engineering in Medicine and Biology 19th Int. Conf. 2771-2776. Tendick, F., Bhoyrul, S., and Stark, L. (1995) Comparison of laparoscopic imaging systems and conditions using a knot tying task. In 2nd Ann. Int. Symp. on Medical Robotics and Computer Assisted Surgery, pp. 238-245 85 Tendick, F., Jennings, R.W., Tharp, G. & Stark, L. (1996). Perception and manipulation problems in endoscopic surgery. In Computer-Integrated Surgery: Technology and Clinical Applications (Taylor, R.H., Lavallee, S., Burdea, G.C. and sges, R., eds.), The MIT Press, Cambridge, Massachusetts, pp. 567-575. Tendick, F., Mori, T. & Way, L.W. (1995). Future of Laparoscopic Surgery. In Fundamentals of Laparoscopic Surgery (L.W.Way, S.Bhoyrul and T.Mori, eds.), Churchill Livingstone Inc., New York, pp. 235-252. Traverso, L.W. (2000). Surgical outcomes: What are they and why should they be measured? Surgical Endoscopy 14,103-105. Traverso, L.W., Koo, K.P., Hargrave, K., Unger, S.W., Roush, T.S., Swanstrom, L.L. , Woods, M.S., Donohue, J.H., Deziel, D.J., Simon, I.B., Froines, E. , Hunter, J. & Soper, N.J. (1997). Standardizing laparoscopic procedure time and determining the effect of patient age/gender and presence or absence of surgical residents during operation. A prospective multicenter trial. Surgical Endoscopy 11,226-229. Treat, M . (1996). A surgeon's perspective on the difficulties of laparoscopic surgery. In Computer-Integrated Surgery (Taylor, R.H., Lavallee, S., Burdea, G.C. and Mosges, R., eds.), MIT Press, Cambridge, MA. , pp. 559-560. van Bergen, P., Kunert, W., Bessell, J. & Buess, G.F. (1998). Comparative study of two-dimensional and three-dimensional vision systems for minimally invasive surgery. Surgical Endoscopy 12, 948-954. van Hemert tot Dingshof, V., Lazeroms, M. , van der Ham, A., Jongkind, W., and Honderd, G. (1996). Force Reflection for a Laparoscopic Forceps. In Proc. 18th IEEE Engineering in Medicine and Biology Int. Conf. Wallwiener, D., Stumpf, B., Bastert, G. & Mueller, W. (1995). Multifunctional instrument for operative laparoscopy: technical," experimental and clinical results in gynaecology. Endosc.Surg.A11 ied. Technol. 3, 119-124. Way, L.W., Bhoyrul, S. & Mori, T. (1995). Learning Laparscopic Surgery. In Fundamentals of Laparoscopic Surgery (L.W.Way, S.Bhoyrul and T.Mori, eds.), Churchill Livingstone Inc., New York, pp. 225-233. Wilson, M.S., Middlebrook, Al, Sutton, C , Stone, R. & McCloy, R.F. (1997). MIST VR: a virtual reality trainer for laparoscopic surgery assesses performance. Ann.R.Coll.Surg.Engl 79,403-404. Zar, J.H. (1984). Biostatistical Analysis, 3rd ed. Prentice-Hall, Inc. Englewood Cliffs, NJ., Zar, J.H. (1999). Biostatistical Analysis, 4th ed.. Prentice Hall Upper Saddle River, NJ, Zill, D.G. & Cullen, M.R. (1992). Advanced engineering mathematics. PWS-Kent Boston, pp. 465 Zucker, K.A. , Bailey, R.W., Graham, S.M., Scovil, W. & Imbembo, A .L . (1993). Training for laparoscopic surgery. World J Surg. 17,3-7. 8 6 Appendix A Analysis of physical constraints task time data A . l Suturing task times and Analysis of Variance (ANOVA) Table 11 shows the suturing task completion times for 15 surgeons on the five instrument conditions. Table 11: Suturing task completion times (seconds) for surgeons (n=15) on 5 tool arrangements Surgeon Task 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 6 open 56.91 28.13 31.03 19.80 13.17 11.04 35.21 16.21 14.00 11.19 32.13 13.22 20 22 18 46.15 24.89 16.97 10.04 9.19 13.27 26.16 21.12 10.17 15.01 21.15 12.18 17 22 19 32.06 13.94 17.00 13.82 9.09 12.03 21.08 12.21 9.05 12.13 19,11 11.21 15 20 20 28.87 16.02 17.18 8.83 13.12 9.09 17.29 9.17 11.20 9.29 15.27 12.25 15 20 18 5 open 65.06 35.14 40.04 23.23 25.21 19.25 51.18 23.16 25.08 23.12 45.21 32.14 35 37 55 43.15 21.91 32.91 20.23 50.28 18.15 28.1 40.04 49.24 20.11 21.14 16.13 35 40 55 86.86 17.92 48.06 32.13 21.15 20.03 43.15 25.12 21.16 22.15 23.12 20.10 30 35 33 44.99 16.94 22.98 18.16 18.26 23.26 25.18 13.13 20.08 26.14 36.29 22.14 30 30 60 4 open 69.83 37.06 36.95 34.24 30.06 17.06 56 57.04 31.27 20.03 58.18 25.26 40 3. 47.74 39.93 45.91 38.93 17,09 21.02 20.16 51,04 32.11 25.15 22.29 51.23 44.27 37 4- ' 43J4 58.07 36.07 48.86 27.25 25.17 14.24 38.12 18.03 31.03 17.20 52.23 20.17 37 3. v - '-41.38 56.06 51.95 31.22 36.08 17.18 34.13 50.2 20.17 19.26 37.05 62.04 32.20 30 25 i . ' ; 44.89 6 lap 55.05 34.81 38.83 23.22 21.05 26.09 39.21 31.12 21.18 34.05 42.07 22.19 33 55 40 41.87 39.21 27.93 25.13 22.19 17.2 47.18 42.29 23.12 23.19 46.26 51.02 40 67 25 59.84 28.74 22.08 21.27 26.29 13.25 44.01 15 27.03 16.17 29.24 49.00 40 45 45 102.75 22.95 21.01 33.19 21.15 28.22 39.1 42.01 22.20 31.16 38.15 29.05 30 37 37 4 lap 179.05 55.02 54.01 29.17 34.09 26.03 77.01 56.1 38.21 29.07 169.25 74.01 80 115 127 122.93 92.05 38.94 25.16 35.21 24 47.24 52.06 37.05 23.25 75.12 41.04 100 55 150 85.98 58.10 52.13 29.19 55.12 42.15 53.21 47.09 55.25 29.27 118.20 67.27 100 75 72 71.04 63.17 87.99 53.16 23.02 46.03 78.19 72.1 25.06 .49.21 52.23 48.02 78 120 83 INB: Shaded area indicates estimation for missing data according to Zar (1984). The data for surgeons 13 through 15 represent the original pilot study data from Hodgson et al. (1997). The pilot study timed five sutures, so I removed the final time in all cases to match the four timed sutures studied in the present data analysis. Normally, one would randomize data before removing one point, but since the fifth observation may be lower than the previous four due to learning effects, it was removed to remain consistent with the present analysis. Note also the shaded area for Surgeon 15 under the 4 open condition. In this instance the surgeon complained of an aggravated thumb injury and subsequently opted not to performing the suturing task under the 40 condition. In order to compensate for the missing data, values were estimated according to statistically valid methods presented by Zar (1984, p.216). Each point of missing data point, Eh is estimates as follows. aA+bB.-X £ , = — J-N + (l-a-b) a = row b - column N = total an * bm 4 = 2>, (13) less missing values 87 Where £, is the estimated value, Aj is the sum of the row in which the missing data point lies, Bt is the column, and T is the total for all values in the table, less the missing values. The randomized block A N O V A table generated from Microsoft Excel 97 is illustrated in Table 12. This A N O V A analysis is available in MS Excel 97 as 'two-factor with replication' in the data analysis package. Table 12: Suturing task randomized block ANOVA table generated in MS Excel 97 (a=0.001). SUMMARY 6 open 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Total Count Sum Average Variance 5 open 4 163.99 41.00 168.90 4 82.98 20.75 46.78 4 82.18 20.55 48.87 4 52.49 13.12 24.33 4 44.57 11.14 5.35 4 45.43 11.36 3.12 4 99.74 24.94 60.13 4 58.71 14.68 26.76 4 44.42 11.11 4.50 4 47.62 11.91 5.68 4 87.66 21.92 52.32 4 48.86 12.22 0.67 4 67.00 16.75 5.58 4 84.00 21.00 1.33 4 75.00 18.75 0.92 60 1084.65 18.08 81.55 .Count Sum Average Variance 4 open 4 240.06 60.02 418.76 4 91.91 22.98 70.37 4 143.99 36,00 113.61 4 93.75 23.44 37.91 4 114.90 28.73 214,62 4 80.69 20.17 4.83 4 147.61 36.90 152.59 4 101.47 25.37 123.30 4 115.56 28.89 188.67 4 91.52 22.88 6.30 4 125.76 31.44 129.48 4 90.51 22.63 46,44 4 130.00 32.50 8.33 4 142.00 35.50 17.67 4 203.00 50.75 145.58 60 1912.73 31.88 201.89 Count Sum Average Variance 4 223.89 55.97 151.27 4 170,99 42.75 57.21 4 155.96 38.99 53.99 4 114.66 28.67 74.02 4 93.43 23.36 30.61 4 85.59 21.40 77.90 4 195.36 48.84 57.62 4 127.35 31.84 320.67 4 106.71 26.68 32.46 4 96.57 24.14 78.38 4 223.68 55.92 26.06 4 121.90 30.48 108.89 4 144.00 36.00 18.00 4 135.00 33.75 72.92 4 177.64 44.41 6.68 60 2172.73 36.21 180.85 Count Sum Average Variance 4 lap 4 259.51 64.88 695.21 4 125.71 31.43 50.37 4 109.85 27.46 66.68 4 102.81 25.70 27.40 4 90.68 22.67 6.09 4 64.76 21.19 50.80 4 169.50 42.38 15.50 4 130.42 32.61 164.80 4 93.53 23.38 6.54 4 104.57 26.14 65.29 4 155.72 38.93 52.70 4 151.26 37.82 206.81 4 143.00 35.75 25.58 4 204.00 51.00 168.00 4 147.00 36.75 72.25 60 2072.32 34.54 216.97 Count Sum Average Variance Tola! 4 459.00 114.75 2313.23 4 268.34 67.09. 288.29 4 233.07 58.27 437,59 4 136.66 34.17 163.87 4 147.44 36.86 178.46 4 138.26 34.57 124.20 4 255.65 63.91 255.97 4 227.35 56.84 117.11 4 155.57 38.89 154.26 4 130.80 32.70 128.94 4 414.80 103.70 2657.68 4 230.34 57.59 242.94 4 358.00 89.50 147.67 4 365.00 91.25 989.58 4 432.00 108.00 1348,67 60 3952.30 65.87 1276.41 Count Sum Average Variance 20 1346.45 67.32 1250.76 20 739.93 37.00 382.07 20 725.05 36.25 285.63 20 500.39 25.02 102.56 20 491.02 24.55 143.16 20 434.73 21.74 99.11 20 867.86 43.39 261.25 20 645.30 32.27 321.15 20 515.79 25.79 145.97 20 471.08 23.55 92.69 ' 20 1007.62 50.38 1339.47 20 642-87 32.14 341.94 20 842.00 42.10 676.41 20 930.00 46.50 819.74 20 1034.64 51.73 1202.87 ANOVA Source of Variation SS df MS F P-valua For it Tools Surgeon Interaction Within 73442.4 47114.0 24819.5 43569.1 4 14 56 221 18360.6 3365.3 443.2 197.1 93.13 17.07 2.25 3.66E-47 1.07E-28 1.02E-05 4.789 2.722 1.539 F>Fcrit F>Fcrit F>Fcrit Reject Ho Reject Ho Reject Ho Total 188945 295 Note that the 4 missing data points are not included in calculating the within cells and total DOF (Zar 1984, p.217), so the DOF values shown in the A N O V A table are less 4. As the A N O V A analysis indicates significant differences, a Tukey multiple comparison test between means is performed (Zar 1984, p. 186, p.213). The Tukey test tests the null hypothesis that means are equal between various treatments (instrument configurations). A confidence interval (CI) analysis is also performed to set the 95% confidence bounds on the mean difference. Results of the Tukey comparison and CI analysis are presented in Table 13. Table 13: Tukey comparison of means for suturing data (a=0.001) Tukey multiple comparison test (Zar, p.186, 213) Ho: Means are equal Tukey's CI Analysis alpha: 0,001 Diff. B/w q(0.001, Diff. B/w "+/- Lower Upper Comparisons means SE q 221,5) Conclusions means q_crit*SE" Bound Bound 6 0 vs. 5 0 13.80 1.81 7.61 5.484 Reject Ho 60 vs. 50 13.80 9.940658 3.86 23.74 6 0 vs. 4 0 . 18.13 1.81 10.00 5.484 Reject Ho 60 vs. 40 18.13 9.940658 8.19 28.08 6 0 vs. 6L 16.46 1.81 9.08 5.484 Reject Ho 60 vs . 6L 16.46 9.940658 6.52 26.40 6 0 vs. 4L 47.79 1.81 26.37 5.484 Reject Ho 60 vs. 4L 47.79 9.940658 37.85 57.73 5 0 vs. 4 0 4.33 1.81 2.39 5.484 Accept Ho ' 5 0 vs. 6L 2.66 1.81 1.47 5.484 Accept Ho 5 0 vs. 4L 33.99 1.81 18.75 5.484 Reject Ho 50 vs. 4L 33.99 9.940658 24.05 43.93 4 0 vs. 6L 1.67 1.81 0.92 5.484 Accept Ho 4 0 vs. 4I_ 29.66 1.81 16.36 5.484 Reject Ho 40 vs. 4L 29.66 9.940658 19.72 39.60 6L vs. 4L 31.33 1.81 17.29 5.484 Reject Ho 6L vs. 4L 31.33 9.940658 21.39 41.27 The power of a performed test is evaluated following Zar (1984, p.228). <t> = . \{nl0l>l-l)*(MSt00l-MSwilhin) n,ooi * MSwilhin / (5 - 1 ) * (18360.6 -197.1) 5*197.1 = 8.59 (14) Consulting Zar (1984, Table p.660) with v/=4, v2=221, aj=0.01, the power is found to be > 0.99 The power of detecting an interaction was also calculated (Zar 1984, p.227) with v7=4, v2=221, c^0.05. 0 _ I u l J i n t e r a t c t i o n ( ^ ^ n t e r a t c h o n - l ) * ^ , w „ 24819.5 (15) \(56 + l)(197.1) 1.49 Here the power is determined from tables (Zar 1984, p.660) to be 0.76. 89 A.2 Pick-and-place task times and Analysis of Variance (ANOVA) Table 14 shows the suturing task completion times for 14 surgeons on the five instrument conditions. Note that one surgeon in the pilot study did not perform the pick-and-place task. 90 Table 14: Pick-and-place task completion times (seconds, n=14). Tool 6 Open Surgeon 6 Lap 1.26 1.20 1.16 1.29 2.09 3.00 2.01 1.88 1.95 1.17 1.93 1.83 1.20 1.97 1.90 2.85 2.12 1.91 1.98 1.86 2.06 1.87 1.91 1.90 1.98 1.90 0.99 1.06 1.05 1.05 1.22 1.17 1.12 1.22 1.13 2.86 1.94 1.16 2.06 1.12 1.95 1.11 1.89 . 1.91 1.78 1.26 2.06 1.88 1.21 1.08 1.93 1.82 2.11 1.94 1.18 1.04 0.97 1.04 1.03 1.73 2.13 1.13 2.06 1.25 1.22 2.86 1.93 2.05 1.92 1.82 1.24 1.20 2.02 1.91 1.14 1.95 1.85 2.02 1.21 1.87 1.94 1.94 2.12 1.94 1.18 1 1 1 | • B 1.82 1.03 1.04 1.25 1.16 2.02 1.19 1.11 2.01 1.08 1.26 1.86 1.16 1.95 2.85 1.97 2.01 1.80 1.92 1.12 2.00 3.01 1.09 1.98 1.84 2.03 1.93 1.86 1.96 1.15 1.81 2.12 0.98 2.13 1.16 1.27 1.27 1.23 2.14 1.96 1.99 1.91 2.00 2.01 1.17 1.97 1.93 2.80 1.94 1.84 1.96 _™1.S1^  4.89 ' 1.10 2.01 1.90 2.07* 2.13 1.98 1.81 1.07 2.77 3.08 1.18 2.11 1.24 2.01 2.14 1.11 1.99 1.88 2.03 2.05 1.86 1.97 1.97 2.08 2.00 1.86 2.06 1 . 9 4 Q ? ; i l l 1.96 1.13 1.98 1.26 2.02 2.03 0.99 1.09 176F::-' mm 3.09 2.21 3.00 2.23 3.24 3.07 2.95 3.02 2.87 3.82 3.88 2.99 3.11 2.93 1.98 3.00 3.11 4.98 2.04 4.10 2.23 3.79 3.10 2.06 2 . 0 7 p 1.89 1.21 2.79 1.15 1 0 8 i L 2.15 2.13 2.22 2.08 2.17 2.87 2.16 5.78 4.14 1.92 3.25 2.78 2.29 2.03 4.18 2.82 2.04 2.81 2.02 2.79 2.91 3.01 2.95 2.05 2.07 1.21 1.10 2.24 1.79 1.91 4.00 2.22 3.04 2.24 3.10 3.10 5.95 3.01 2.99 4.18 3.03 2.10 3.24 4.05 2 . 9 6 " 3.87 6.80 2.96 2.96 2.86 3.12 2.18 4.04 2.95 2.92 3.95 1.92 2.89 1.94 2.11 3.14 2.23 3.27 2.17 2.23 4.01 2.78 3.79 2.92 3.00 2.98 4.19 3.02 2.06 2.06 2.90 2.08 3.01 2.95 2.79 2.97 2.96 2.98 2.96 2.13 2.12 1.96 2.15 2.01 5.02 2.13 2.95 2.14 2.89 2.08 1.09 2.06 2.06 2.00 1.98 2.00 1.10 2.06 2.07 2.04 1.97 2 . 0 7 | 1.02 2.13 4.26 2.77 2.16 2.82 1.15 2.05 2.00 2.02 1.99 1.98 1.07 1.29 2.72 1.27 1.901 2 .03* 1.05 3.08 2.98 2.21 2.89 2 F 7 I m ^ .13 2.10 2.05 2.88 2.06 1.79 2.25 2.85 "2.86' 4.14 2.90 1.22 4.06 2.12 •2.94 2.14 3.02 1.92 7 8 9 10 11 12 13 2.07 1.13 2.06 • 2.10 2.26 2.28 1.77 2.03 1.11 1.21 2.90 2.80 1.98 2.20 2.02 1.78 1.89 2.16 2.18 1.91 1.60 2.10 1.06 1.88 2.89 2.82 1.91 1.77 2.02 1.08 1.98 2.08 2!08 1.97 1.33 1.87 0.96 1.05 1.06 2.08 1.17 2.23 .2.01 1.11 1.15 2.05 2.01 1.27 1.67 1.22 1.07 1.'14 2.03 3.03 2.80 1.20 1.96 1.09 1.80 2.01 2.00 1.94 1.40 1.90 1.76 1.12 1.98 2.07 2.02 1.60 1.96 .1 .05 1.83 2.03 2.15 1.21 2.00 1.99 0.21 1.01 1.06 1.00 1.79 1.33 2.00 1.08 1.18 2.07 2.16 1.18 1.90: 1.24 1.08 1.85 2.08 2.88 1.92 1.33 1.92 1.07 1.14 2.01 2.15 1.94 1.67 3.08 1.06 1.89 2.00 2.83 1.99 2.33 Hat.oij 1.78 1.15 2.04 2.04 1.23 1.83 1.09 1.16 1.03 1.00 1.08 1.76 1.47 1.28 1.09 1.08 2.15 2.08 1.17 1.60 2.82 2.02 1.20 2.12 2.00 1.86 1.33 1.91 1.09 1.79 2.77 1.94 1.15 2.13 1.21 1.80 1.10 2.15 . 1.23 1.89 3.00 1.96 1.13 1.85 2.81 1.97 1.13 1.57 1.08 0.99 0.98 1.11 1.02 1.85 1.67 2.01 1.21 1.28 2.06 3.03 1.15 1.67 1.21 1.84 2.00 2.02 2.25 1.12 2.27 1.99 1.06 2.81 2.02 2.88 1.85 2.00 1.97 1.11 1.62, 1.98 2.90 1.16 1.00 2.05 1.81 2.76 1.97 3.07 1.87 2.00 1.97 0.99 2.07 1.03 1.87 1.03 1.33 2.00 1.07 1.29 1.25 2.11 1.12 1.90 2.01 1.08 2.73 2.75 2.91 1.12 1.47 1.22 1.08 2.07 1.24 2.04 1.80 1.67 1.71, 1.05-/" 1.93 2.09 1.13 1.70 2.87 1.78 "* 2 do 2.03 2.99 1.12 1.93 1.20 0.97 5.12 1.06 1.88 1.72 1.67 5.17 1.29 . 3.26 3.09 3.00 1.27 3.00 3.02 2.79 3.84 3.02 3.05 2.73 4.67 3.83 1.97 1.95 2.90 3.01 2.14 2.33 4.07 1.22 4.13 2.16 4.05 2.10 2.90 3.17 4.98 2.03 2.88 3.12 2.03 2.60 3.02 1.81 1.81 2.06 2.85 1.77 2.53 3.12 2.07 2.16 2.12 3.09 1.21 2.83 2.94L-, •US] 1.97 2.14 3 - 0 2 R M S 3.00 6.11 1.98 4.87 2.78 2.92 2.02 3.67 3.08 1.95 2.09 2.14 1°2 2.98 2.07 2.05 2 .07 ' 3.06 2.13J3 2.83 2.84 1.03 1.91 1.77 1.96 1.96 3.14 2.21 3.09 2.28 4.20 3.04 4.43 3.04 2.82 4.11 2.87 4.00 3.02 5.40 3.02 2.14 3 0 2 O lilts 3.96 2.10 4.00 3.09 2.08 2.86 4,14 3.97 3*24 5.00 2.95 4.80 2.92 2.90 4.95 3.03 4.00 2.86 1.21 2.14 1.19 2.21 1.87 3.50 2.26 3.04 2.23 3.29 4.05 2.23 3.77 3.83 2.01 2.97 2.84 3.99 2.81 5.23 2.96 3.21 2.09 2.93 3.97 2.98 3.33 3.03 3.86 2.77 2.15 3.21 2.15 4.40 3.01 3.97 2.15 2.99 3.97 2.12 . 3.30 2.18 3.13 4.98 1.95 2.81 1.83 4.40 14 3.83 5.20 2.00 2.33 1.70 1.67 2.00 2.00 2.00 1.60 1.83 3.00 2.37 2.33 2.13 2.20 1.87 1.67 2.00 2.47 1.87 1.80 1.87 1.50 2.53 3.00 3.00 2.00 2.00 1.67 , 1.67. 2.00 1.80 1.63 1.57 3.00 3.00 1.43 2.00 1.83 1.80 1.87 3.00 2.00 1.83 1.83 1.67 1.83 3.37 3.00 4.00 3.00 3.63 3.00 5.17 4.00 3.00 5.33 3.53 3.67 Shaded cells represent missing or anomalous data (due to task errors) which were corrected using the missing data technique illustrated in equation (1) above. The same randomized block A N O V A technique employed in examining the suturing results was performed on the pick-and-place tasks. The results and A N O V A table are presented in Table 15: Pick-and-place task randomized block ANOVA table generated in MS Excel 97 (a=0.001). S U M M A R Y 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Total 6 Open Count 12 12 12 12 12 12 12 12 12 12 12 12 12 12 168 S u m 2 3 . 2 6 19 .09 18 .32 19.51 18 .07 2 4 . 4 8 2 3 . 1 5 13.41 18 .12 2 4 . 3 5 2 6 . 4 8 2 2 . 2 5 2 0 . 1 0 2 9 . 1 7 2 9 9 . 7 6 A v e r a g e 1.94 1.59 1.53 1.63 1.51 2.04 1.93 1.12 1.51 2 . 0 3 2.21 1.85 1.68 2 . 4 3 1.78 V a r i a n c e 0 .46 0.21 0 .18 0 .18 0 .17 0.31 0 . 0 5 0 .16 0 .18 0.31 0 .27 0 . 2 2 0.11 1.17 0 .37 5 Open Count 12 12 12 12 12 12 12 12 12 12 12 - 1 2 12 12 168 S u m 2 2 . 3 2 1 8 . 4 9 2 3 . 1 8 2 1 . 3 8 16 .27 2 6 . 5 2 2 1 . 6 0 1 5 . 3 5 16.24 24 .31 2 3 . 3 8 19.07 2 1 . 8 3 2 4 . 0 7 2 9 4 . 0 1 A v e r a g e 1.86 1.54 1.93 1.78 1.36 2.21 1.80 1.28 1.35 2 . 0 3 1.95 1.59 1.82 2.01 1.75 V a r i a n c e 0 23 0 . 3 0 0 . 0 5 0 . 3 0 0 . 1 3 0 . 6 9 0 . 4 3 0 . 1 3 0 .14 0 . 2 8 0 .35 0 .14 0 . 2 3 0 . 0 9 0.31 4 Open Count 12 12 12 12 12 12 12 12 12 12 12 12 12 12 168 S u m 2 8 . 5 0 17 .34 22.21 2 0 . 1 6 2 3 . 4 8 2 1 . 3 7 22 .21 1 5 . 0 5 2 7 . 1 7 21 .34 3 0 . 0 2 16 .19 2 0 . 6 0 2 5 . 8 7 311 .51 A v e r a g e 2 . 3 8 1.45 1.85 1.68 1.96 1.78 1.85 1.25 2 .26 1.78 2 .50 1.35 1.72 2 .16 1.85 V a r i a n c e 0 . 7 2 0 .17 0 . 1 0 0 . 1 3 0 .24 0 . 2 5 0 . 2 2 0 . 1 2 1.11 0 .27 0.24 0 . 1 2 0 . 1 2 0 . 3 3 0 . 4 0 Count 12 12 12 12 12 12 12 12 12 12 12 12 12 12 168 S u m 3 2 . 3 7 2 9 . 4 8 3 8 . 2 9 2 7 . 3 9 3 1 . 3 3 2 8 . 8 2 4 3 . 3 5 2 5 . 2 9 32 .07 2 9 . 1 3 3 6 . 1 2 2 3 . 6 6 3 7 . 1 0 2 4 . 1 0 4 3 8 . 5 0 A v e r a g e 2 .70 2 .46 3 .19 2 .28 2.61 2 .40 3.61 2.11 2.67 2 .43 3.01 1.97 3 .09 2.01 2.61 V a r i a n c e 0 .51 0 . 6 3 1.18 0 . 5 5 1.01 0 .22 1.07 1.04 1.13 0 . 2 2 0 . 2 0 0 .17 0 .46 0 .24 0 . 7 8 4 Lap Count 12 12 12 12 12 12 12 12 12 12 12 12 12 12 168 S u m 3 9 . 1 9 3 7 . 3 7 3 7 . 4 0 . 3 2 . 2 0 3 5 . 3 6 3 2 . 4 2 3 5 . 3 7 3 4 . 4 8 3 5 . 3 3 3 2 . 4 3 4 5 . 2 9 3 0 . 4 2 5 0 . 7 7 44 .71 5 2 2 . 7 3 A v e r a g e 3 .27 3.11 3 .12 2 .68 2 . 9 5 2 .70 2 .95 2 .87 2 .94 2 .70 3 .77 2 .54 4 . 2 3 3 . 7 3 3.11 V a r i a n c e 0 . 3 3 2 . 7 5 0 .22 0 . 3 8 0 . 7 8 0.71 0 .18 1.02 0 .74 0 . 5 5 0.51 0 28 0 .51 0 .64 0 . 8 5 Total Count 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 6 0 S u m 1 4 5 . 6 4 121 .77 1 3 9 . 4 0 120.64 124.51 133.61 1 4 5 . 6 8 1 0 3 . 5 8 1 2 8 . 9 3 131 .56 161 .29 111 .59 1 5 0 . 4 0 147.91 A v e r a g e 2 . 4 3 2 . 0 3 2 .32 2.01 2 .08 2 . 2 3 2 . 4 3 1.73 2 . 1 5 2.19 2.69 1.86 2.51 2.47 V a r i a n c e 0 . 6 9 1.19 0.81 0 .46 0 . 8 2 0 .51 0 . 9 0 0 . 9 2 1.01 0 .41 0 . 7 2 0 . 3 3 1.30 0 . 8 9 A N O V A Source of Variation S S df MS F P-vatue Fait S a m p l e 2 5 0 . 5 7 4 6 2 . 6 4 144 .10 1 . 9 E - 9 3 4.67 F>Fcrit Reject H o C o l u m n s 5 6 . 3 3 13 4 .33 9.97 5 . 2 1 E - 2 0 2 .70 F>Fcr i t Reject H o Interaction 6 8 . 7 1 52 1.32 3.04 1 . 2 9 E - 1 1 1.76 F>Fcr i t Reject H o Wi th in 3 2 7 . 7 7 754 0 .43 Total 7 0 3 . 3 7 823 Significant effects warrant further testing with the Tukey multiple comparison test (Zar 1984, p. 186, p.213). The Tukey test tests the null hypothesis that means are equal between various treatments (instrument configurations). A confidence interval (CI) analysis is also performed to set the 95% confidence bounds on the mean difference. Results of the Tukey comparison and CI analysis are presented in Table 16. Table 16: Tukey comparison of means for Pick-and-place data (a=0.001) Tukey multiple comparison Ho: Means are equal Tukey's CI Analysis alpha: 0.001 Diff. B/w Diff. B/w Lower Upper Comparisons means S E q q_crit Concls. means q . _crit*SE" Bound Bound 6 0 vs. 5 0 0.03 0.050868 0.672 5.484 Accept Ho 6 0 vs. 4 0 0.07 0.050868 1.375 5.484 Accept Ho 6 0 vs. 6L 0.83 0.050868 16.235 5.484 Reject Ho 60 vs. 6L 0.826 0.279 0.547 1.105 6 0 vs. 4L 1.33 0.050868 26.092 5.484 Reject Ho 60 vs. 4L 1.327 0.279 1.048 1.606 5 0 vs. 4 0 0.10 0.050868 2.047 5.484 Accept Ho 5 0 vs. 6L 0.86 0.050868 16.907 5.484 Reject Ho 50 vs. 6L 0.860 0.279 0.581 1.139 5 0 vs. 4L 1.36 0.050868 26.764 5.484 Reject Ho 50 vs. 4L 1.361 0.279 1.082 1.640 4 0 vs. 6L 0.76 0.050868 14.860 5.484 Reject Ho 40 vs. 6L 0.756 0.279 0.477 1.035 4 0 vs. 4L 1.26 0.050868 24.717 5.484 Reject Ho 40 vs. 4L 1.257 0.279 0.978 1.536 6L vs. 4 1 0.50 0.050868 9.857 5.484 Reject Ho 6L vs. 4L ' 0.501 0.279 0.222 0.780 Value for qcrit at (ot=0.001,754,5) found in Table B.5, Zar (1984, p.522). Power for a performed test (((>= 10.7, 1-/J> 0.99) and power to detect interaction (</)= 1.73,1-/? = 0.88) calculated as illustrated for suturing test in equations (2) and (3) with v/=4, V2=754, with c^O.Ol and a=0.05, respectively. 92 A.3 Analysis of open vs. laparoscopic surgery operation times We examined 25 references (totaling 29 comparisons) from literature comparing various equivalent open and laparoscopic procedures on the basis of operation time and found that open procedures take approximately 74% ± 17% (mean ± s.d.) of the operation, time required for equivalent laparoscopic procedures. Table 17: Ratio of equivalent open to laparoscopic mean operative time in 29 references. References shown correspond to those listed in the subsequent reference list Reference •BflHf Mean operation time (mins.) Lap Open Stat. Diff ? y/n/ (-) not available Ratio (open/lap) 1 52.6 28.4 - "~54% 2 260 251 - 97% 304 212 y 70% 139 128 - 92% 3 202 144 y 71% 4 221 59 y 27% 5 58.7 50.6 - 86% 6 282 277 - 98% 7 115 110 - 96% 8 103.78 87.79 y 85% 9 2000 1000 - 50% 10 164 131 - 80% 11 164 124 y 76% 12 289 201 y 70% 13 35 30.5 - 87% 14 102.2 81.7 y 80% 15 107 72 - 67% 16 230 165 - 72% 17 66 58 - 88% 68 58 - 85% 18 218 168 y 77% 19 212 174 - 82% 212 139 y 66% 20 47.5 41 n 86% 21 78.8 62.7 y 80% 22 122 95 - 78% 23 264 133 y 50% 24 70.3 46.5 y 66% 25 226 101 y 45% Average: 74% St. Dev.: 17% Count: 29 93 A.3.1 Reference list Winfield H.N., Hamilton B.D., Bravo E.L. and Novick A.C. (1998) Laparoscopic adrenalectomy: the preferred choice? A comparison to open adrenalectomy. J. Urol. 160, 325-329. Khalili T .M. , Fleshner P.R., Hiatt J.R., Sokol T.P., Manookian C , Tsushima G. and Phillips E.H. (1998) Colorectal cancer: comparison of laparoscopic with open approaches. Dis. Colon Rectum 41, 832-838. Holub Z., Voracek J. and Shomani A. (1998) A comparison of laparoscopic surgery with open procedure in endometrial cancer. Eur.J.Gynaecol.Oncol. 19, 294-296. Rosser J.C.J., Rosser L.E. and Savalgi R.S. (1998) Objective evaluation of a laparoscopic surgical skill program for residents and senior surgeons. Arch.Surg. 133, 657-661. Rogers D.A., Hatley R.M. and Howell C.G.J. (1998) A prospective, randomized comparison of traditional and laparoscopic inguinal exploration in children. Am.Surg. 64, 119-121. Doehn C , Fornara P., Fricke L. and Jocham D. (1998) Comparison of laparoscopic and open nephroureterectomy for benign disease. J.Urol. 159,732-734. ^ Fornara P., Doehn C. and Jocham D. (1997) Laparoscopic nephropexy: 3-year experience. J.Urol. 158 , 1679-1683. Ramos J.R., Petrosemolo R.H., Valory E.A., Polania F.C. and Pecanha R. (1997) Abdominoperineal resection: laparoscopic versus conventional. Surg.Laparosc.Endosc. 7, 148-152. So J.B., Kum C.K., Fernandes M.L. and Goh P. (1996) Comparison between laparoscopic and conventional omental patch repair for perforated duodenal ulcer. Surg.Endosc. 10, 1060-1063. Elashry O.M., Nakada S.Y., Wolf J.S.J., Figenshau R.S., McDougall E .M. and Clayman R.V. (1996) Ureterolysis for extrinsic ureteral obstruction: a comparison of laparoscopic and open surgical techniques. J.Urol. 156, 1403-1410. Cox M.R., McCall J.L., Toouli J., Padbury R.A., Wilson T.G., Wattchow D.A. and Langcake M . (1996) Prospective randomized comparison of open versus laparoscopic appendectomy in men. World J.Surg. 20, 263-266. Naito S., Uozumi J., Shimura H., Ichimiya H., Tanaka M . and Kumazawa J. (1995) Laparoscopic adrenalectomy: review of 14 cases and comparison with open adrenalectomy. JEndourol. 9,491-495. 94 Yoshida K., Yamazaki Y., Mizuno R., Yamadera H., Hara A., Yoshizawa J. and Kanai M . (1995) Laparoscopic splenectomy in children. Preliminary results and comparison with the open technique. Surg.Endosc. 9, 1279-1282. Dubois B., Nagy A.G. , Anderson D., Simpson W.T. and Appleby J.P. (1995) Comparison of initial laparoscopic cholecystectomy at a community hospital versus a teaching hospital. Can.J.Surg. 38, 439-444. Martin L.C. , Puente I., Sosa J.L., Bassin A., Breslaw R., McKenney M.G. , Ginzburg E. and • Sleeman D. (1995) Open versus laparoscopic appendectomy. A prospective randomized comparison. Ann.Surg. 222, 256-261. Prinz R.A. (1995) A comparison of laparoscopic and open adrenalectomies. Arch.Surg. 130, 489-492. Scorpio R.J., Tan H.L. and Hutson J.M. (1995) Pyloromyotomy: comparison between laparoscopic and open surgical techniques. J.Laparoendosc.Surg. 5, 81-84. Maddern G.J., Rudkin G., Bessell J.R., Devitt P. and Ponte L. (1994) A comparison of laparoscopic and open hernia repair as a day surgical procedure. Surg.Endosc. 8, 1404-1408. Payne J.H.J., Grininger L . M . , Izawa M.T., Podoll E.F., Lindahl P.J. and Balfour J. (1994) Laparoscopic or open inguinal herniorrhaphy? A randomized prospective trial. Arch.Surg. 129,973-979. Andrew D.R., Middleton S.B. and Richardson D.R. (1994) A comparison of laparoscopic and open inguinal hernia repair in servicemen. J.R.Army.Med.Corps. 140, 76-78. Hardy K.J., Miller H., Fletcher D.R., Jones R.M., Shulkes A. and McNeil J.J. (1994) An evaluation of laparoscopic versus open cholecystectomy. Med.J.Aust. 160, 58-62. Tate J.J., Dawson J.W., Chung S.C., Lau W.Y. and Li A.K. (1993) Laparoscopic versus open appendicectomy: prospective randomised trial. Lancet 342, 633-637. Williams L.F.J., Chapman W.C., Bonau R.A., McGee EC Jr, Boyd R.W. and Jacobs J.K. (1993) Comparison of laparoscopic cholecystectomy with open cholecystectomy in a single center. Am.J.Surg. 165,459-465. Herbst C.A.J., Elliott L., Koruda M . and Maxwell J.G. (1993) Laparoscopic cholecystectomy: comparison of university and community experience. Surg. Laparosc. Endosc. 3, 95-99. Fisher K.S., Reddick E.J. and Olsen D.O. (1991) Laparoscopic cholecystectomy: cost analysis. Surg. Laparosc. Endosc. 1, 77-81. .95 Appendix B Shop drawings of the Mechanical Teleoperator for Minimally Invasive Surgery Appendix B: Table of Contents Teleoperator isometric view Teleoperator rear isometric view Teleoperator side view Teleoperator front view Teleoperator top view Teleoperator end plate detail (slave end) Assembly - end plate components Part 1 - Middle arm Part 2 - Vertical drive shaft Part 2a - Vertical drive shaft metric sleeve Part 3 - Horizontal drive shaft Part 4 - Tool Part 5 - Ground shaft Part 6 - Ground (Photo stand mount) Part 7 - Main horizontal member Part 8 - Offset spacer \ Part 9 - Mounting end plate (slave end) Part 9a - Mounting end plate (master end) Part 10 - Mounting end plate arm Part 12 - Ball transfer mounting arm Part 17 - Tool holder arm Part 23 - Counter weight Part 24 - Bearing pin Teleoperator Parts List 96 , 2 ! | | D S - a; z IB-Is - u c 15| U 0 ^ O c > E | S O o <D ^ a > OJ 0 C ZOr> o 3! 55 £ « E ^ ~= 3 -Si m uJ <=> 1 * * S> 5 O " „-z 5 oo I j ^ j u-i Is I 2 I* "-o-w E2 0 SE 97 98 99 1 is . u-y, a-c. £ U O S i l l !8J Si U 0) o a iS ^ dj < ° 1 * ^ 5 8 i S5 « 2 < O O £ 5 Q 5 D o ^ Q ? l i s 5 r ^ Z O 5 | 103 /In <2 & L 3 CO 0 CM uS CO CM i , *v_ CO O o o 0 o d co ty. 5± GO 0 O CO co - o ; g ^_ • & GO ress fi - 1 ress fi o ress fi t o _J CD Q_ z 5S i i - o f ^1 J r o U 51°, l i t 1 2 o a) 3 Q-> 0) 0) c o •o 5 1 < < | 0 51 z 2 S o ^ 5 ^ y< I 5 S I ^ o 1 6 5 - s , E s o l ) Z o 2 Qo 5oS SS | | € o O 5 | o •Q. CN CN CN o o CO CO 00 o - o - o ~o CO - o o o 00 CM CO :n=r S3 •o co Q. O D_ 1/1 O co i / i = E o 5 £ 104 13 8 : 5 - or Z -V O O E ° E ^ O CD O (D o • o </>"o; E O J - ^ USE U OS 2 o « => o_> 0J 0J c za=> •c > . . • UJ TJ O 56 105 o o CN o o +i o-o CN u o u o ©5 -° s= i o > u o.Q ° o*-~ O 75 0 O OcNi/5 a .0-u O b 2 UJ Z 12 u u i 5 ° § 0 < CO d So si IF . u-S J3E £ 0 0 = o < J O £ I f 2 o o 3 a > D i e ZQ=> •a o c o o I i ! . o u o ^z9 J o s Q < 3 l s ° - g l l 0 5 £ 1 E2 IS §1 o E D E ^ O <D O (D o •*-on oo _ , O : a), ! T J , . Q CJ <D E l - ^ O 107 o o CN c UJ OS u4?E o oS - ' . C O . u o, 51°, l i f i O a CD D a> o t i c ZOzj O O i as 2 : S5 3 i | £ >U |o < l o l l z US Q'bSi < B ° -115 O 5 E 01 5ZS 0 0 o o .9 o •-CN o -©. CN o / 108 NOIl !y y ui Z LIST z g J2 Z O < i n d So PART IDENTIFYI g o t Z o o ^ 151 S t t2 2ocs D Q. > (!) OJ'C O O I u z i z < < |> z o .NCE z o z < Oo 2 Q TOLE < * S o s •§is O 5 6 1 D 109 CN 0 CO D CO Sc d H> o — o o o o +' l " CO r-iil o. ^ CO co — Z J. •E o D 0 co u < ' o n 5 CL> 1) 0) c ! D 3 D O E c o So . o S So 15 31 - Q:2 |o < = So H o l g Q 5 £ O 5 | 1 1 1 to o O O o o +' l" 0 u a a. CN o o +i o 0 c I-119 Q 0 t o CO •Q-q o CM 01 o o -2 -• .c 0 2 ° " Si 's , HI 2 o a> 5 a> 0) 0) c n E o 2 'OS O n J J cc a. OS § 5 ij b P I g a. 5 z>-u 1 6 1 8 1 1 1 O 5 E ON CO < CN U . o-y o o =: - ' c o S5| <J 0 £ 2 o a) 3 0-> H O C zo=> 0 O O a O t°- c < 'o CES Q CO z z o < OO < TOL < ^ e < 5 j y i u W 115 J o t u f j y 4 m L o r u W CN C >~<D E <D Q . *• > o So n c E o o = S5| u o s |!° O c > E l i , 2 o a <D CD c Z Q j a =1 «; J O LO O O ' o o < CN I u 1 5 ^ S u 5 S 3 ? a 0 § 1 ^ u Q < =5 S-9n CO "Q-' o m Z U 2 Q 5 E iii O 5 S I I I uUm Teleoperator parts list From file: Teleoperator parts list.xls 121 Bill of Materials For two teleoperators Prepared by: John G. Person (M.A.Sc. Candidate) Neuromotor Control Lab, UBC Phone: 822-8785 Date: ' 25.03.99 Number Part name Qty per teleoperator Total Qty Material Stock material dims, (in.) 1 Middle arm 2 4 Al .375 0 D X 7 L 2 Vert drive shaft (rotation) 2 4 SS .197 O D x 1.732 L (*) 2a Vert roller metric sleeve 2 4 Al .25 ODx .1875 IDx 375L 3 Horz drive shaft (plunge) 1 2 SS .197 O D x 25.906 L (*) 4 Tool 2 4 SS .25 ODx .1875 IDx 12.0L 5 Ground shaft 1 2 Al .375 OD x 6.693 L 6 Photo stand mount 1 2 Al 1.750 L x 3.0 H x 1.5 H 7 Main horz. member 1 2 Al .375 O D x .25 ID x 23.622 L 8 Linear bearing set (LMW-250) 4 8 . . . — 9 Mounting end plate (slave end) 1 2 Al 2.244Lx 1.744WX 1.535H 10 Mounting end plate arm 2 4 Al 2.243 L x 1.901 W x .866 H 11 Ball transfer 2 4 . . . . . . 12 Ball transfer mounting arm 2 4 Al 2.25L x 0.906W x 1.0H (**) 13 Mouse ball 2 4 . . . . . . 14 Knurled thumb screw (PQH-26) 2 4 . . . . . . 15 Belleville washer (STM-6) 2 4. . . . . . . 16 Shoulder screw (PL-13) 2 4 . . . . . . 17 Tool holder arm 2 4 Al 0.906Lx 1.423WX.4.056H 18 . Offset spacer 2 4 SS 0.25D x 0.394L 19 Lower arm 2 4 Al 10.296LX 0.125WX 0.375H 20 Timing belt pully 2 4 . . . . . . 21 Timing belt 1 2 . . . . . . 22 Mounting end plate (master end) 1 2 Al 2.244LX 1.744WX 1.535H 23 Counterweight 2 4 Steel (see NOTE) 24 Bearing pin 2 4 SS 5/32 OD x .75 L Material Notes Al Aluminum — Stock Part SS Stainless Steel NOTE: Counter weight drawings to follow (*) 5mm OD x 900mm SS bar stock provided (on order from BERG) (**) stock for ball transfer mounting arm determined as follows: Ordered parts for teleoperator 122 Prepared by: John G. Person (M.A.Sc. Candidate) Neuromotor Control Lab, UBC Phone: 822-8785 Date: 25.03.99 Teleoperator part number Teleoperator part name Ordered part name Quantity (for one teleoperator) Part number ( B E R G Cat. #) 1 Middle arm Ball bearinqs (to main horz memb) 4 , n Ball bearings (to ground shaft) 4 B1-25-Q3 Shoulder screws 2 PL-13-3 2 Vertical drive shaft (rotation) Timing belt pulley 2 8TP4-20 Timing belt (factory splice: 1 8TB-674 673-674 pitches for L = 55.25") Ground shaft 1 S1-105 (**) 3 Horizontal Drive shaft Ground shaft 1 S1-105 (**) 6 Photo stand mount Thumb Screw 1 PQ-40 Ball Bearings 2 B1-13-Q3 9 & 2 2 Mounting end plate Ball Bearings (metric) 6 B1-53 Small mounting screws 4 Y6-S4-A4 Large mounting screw 2 Y6-S4-A10 12 Ball transfer mounting arm Shoulder screw 2 PL-13-3 (Seating force screw) Belleville washer 2 STM - 6 Knurled thumb screw 2 PQH-26 17 Tool holding arm (master/slave) Rotating linear bearing set (2 re ta ine rs , 2 s p a c e r s s , 1 rad ia l b e a r i n g ) 4 LMW-250 Linear ball bearing 4 LMB-1-B Angular ball bearing 2 B14-4 Ball bearing 2 B1-25-Q3 Inner race shaft spacer 2 SS1-23 Shoulder screw 2 PL-10-3 Inner race shaft spacer (small) 4 SS1-37 18 Offset spacer Machine screw 2 Y9-S4-A10 19 Lower arm Shoulder screws 4 PZ-1-3 Ball Bearings 4 B1-35-Q3 (*) Salvage bearings from old teleoperator prototype (**) 5mm OD x 900 mm L stainless steel shafting -- cut two vertical and one horizontal drive shaft from this stock 1-23 Appendix C Validation of Circle and Sphere fitting methods C . l Introduction The basis of kinematic limb calibration is to accurately fit a circle (or sphere) to a cloud of points collected from limb markers in order to determine the location of the centre of rotation, and therefore the approximate joint centre. The circle and sphere fitting methods used in the biomechanical analysis of the upper limb were derived from methods proposed by Santo-Munne (1996), Halvorsen et al. (1999), and Zill & Cullen (1992). The three methods proposed are : True Least Squares (TLS) - Zill & Cullen (1992) Linearized Least Squares (LLS) - Santo-Munne (1996) Linearized Perpendicular Bisector (LPB) - Halvorsen et al. (1999) This section describes the how each of the methods are applied, and compares accuracy and precision of each method against the other in order to objectively select the best method for our biomechanical analysis. It is shown that TLS is the most accurate method, and is therefore preferred. TLS does, however, seriously break down at high noise levels and small arc size. The LBP method is applied to circle and sphere fitting problems when TLS fails. Failing that, LLS may be applied as it is more robust, but less accurate than either LPB or TLS. C.2 Problem description In general, each of the three methods is used to determine the approximate rotation centre of a cloud of data points describing the surface of a circle or a sphere. Consider the example of a circular arc formed by a cloud of points as illustrated in Figure 23 below. If we consider each point as various displacement 'snapshots' of a single point fixed to a rigid body in pure rotation, then one should be able to describe the distance from the centre of rotation to each 'snapshot' point as the radius of a circle. The basis of the LLS and TLS methods is that it is possible, given such a cloud of points, to find the optimal base circle which, fits the all data. The LPB method, on the other hand, localizes the centre by solving the intersection of the set of perpendicular bisectors generated between two points, also illustrated in Figure 23. Figure 37: Illustration of a cloud of data points describing a 90 °arc of a circle (R = 100, noise = ±5%). The LLS and TLS methods fit an optimal circle to the data, while the LPB method solves the intersection of all perpendicular bisectors of sets of two points in the data cloud. C.2.1 TLS theory Both the TLS and LLS methods find the optimal circle or sphere to fit a cloud of data points. In order to explain thev procedure, we examine the steps involved in describing the optimization procedure on a circle. A circle is defined as: f(xi,yi) = ^ xi-x0)2+(yl-y0)2 (16) Where x, indicates a measured point on the surface of the circle and x0 the sphere centre (similarly for yt andy0). In a TLS fit of the data, one measures the error between the expected radius of the circle, R,, and measured radius by the following equation. e,.=|*,-/(W,)| (") In the TLS optimization, the object is to minimize the sum of squares of the error shown above, expressed as: 125 mmE = fj[Rl-f(xl,yi)]2 (18) Solving this minimization for n points measured points yields optimal values for R, x0 and y0. Calculation of the sphere is similar to the circle, with z, added to equations (1) - (3). In order to begin the TLS optimization, however, one must have an initial approximation of the centre. This initial estimate may be produced using the LLS method. C.2.2 LLS theory Santos-Munne (1996) describes a method of linear least squares (LLS) in which the circle or sphere fitting problem is posed as a system of linear equations and solved using the pseudoinverse technique. An example of linearizing a circle is used to illustrate this point. In the equation illustrated below, the original equation for the circle is expanded and reordered: R2=(x-xQf+(yi-y«y (19) 2 2 n , 0 2 2 D 2 *i +y, = ^ xix0+2yiy0-x0 -y0 +R Given this form, and a set of n points (x„ yj, a system of linear equations may be represented as: (20) 2 , 2 " x, + ;/, 2x z.x0 iy0 2 , 2 _X" y" . „ x n yn .1 D 2 2 2 _R -x0 -y0 •_ Which is solved using the pseudo-inverse: A = Bp p = [BTB B T A (21) The linearization of the sphere is similar, but the additional z variable leads to the following matrix representation: 2 , 2 , 2 x, +yl + z. 2 , 2 , 2 *n +yn + Z n x, yx z, 1 X„ yn Z n i. 2x 2>^o . 2z0 • n 2 2 2 2 ^ - ^ o - ^ o - ^ o (22) C.2.3 LPB theory The Linear Perpendicular Bisector (LPB) method described by Halvorsen et ah (1999) localizes the centre of rotation by solving the intersection of the set of perpendicular bisectors generated between any two points, as illustrated in Figure 23. The points may be selected using the following algorithm proposed by Halvorsen et al. (1999): {(1, n 12 +1), (2, n 12 + 2),..., (/, n 12 + /),:.., (" / 2 ,«)} (23) Where n is the number of sample points. These data points may be used to determine the displacement between points such that: 126 &Pi=Pi-Pn/2 + i (24) Taking each displacement, Aph perpendicular to the line connecting the centre of rotation and the midpoint of the displacement vector, the following equation is described: Pi+Pn,2 + i -q = o (25) Where q is the centre of rotation. As with the LLS method, similar equations for all pairs of displacement points may be expressed as a set of linear equations such that: Pl+Pn/2 + l Pn/2+Pr, A A , / 2 (26) Here, q is solved using the pseudo-inverse as described in equation (6). One short coming of this method compared to the others is that the method of selecting the displacements is not optimal. It is possible that the points may be poorly arranged and thus cause the solution to break down. C.2.4 Plane fitting for circles An additional step in the circle fitting algorithm is required to ensure that the cloud of points in nominally in the plane of rotation. Using a TLS method similar to that used on the circle fit, we first define a standard form of plane as follows. f(xl,yi^ = a-xi+b-yl+c-zi+d (27) Where a, b, c and d are constants defining the plane. As with the circle TLS, the error between the expected definition of a plane, Pt, and measured plane by the following equation. e,=|/>-/(x,,X,z,.)| (28) The expected plane, P„ is taken as the initial plane formed by the cross product of the first, middle and last points on the data spread. Then the TLS optimization is expressed as follows. mmE = Y\Pi-f{xi,yl,z,y] ; = 1 Once the optimal plane is found (defined by normalized constants a, b, c and d), calculate the transformation matrix to move all of the circular arc of data (in coordinate frame A) into the plane of the normal (frame B). The calculation for rotation by 6 of about an arbitrary axis, r, is detailed in Appendix F. In this instance, the values of #and r are determined as follows. 127 u = Zp = [a b c] : the plane normal v = Z 0° = [0 0 l] : arbitrary nominal axis 8 = cos r = u x v u-v viiwIIIPV (30) Where ZP° is the axis of the plane normal in the.base frame, and Z0 is an arbitrary axis of the base frame. The calculation moves the circular arc of data from the base frame to a frame with z-axis aligned to the plane normal. The homogeneous transform matrix between frame. A and frame B is then defined as follows. . RX0) = Tt Pi P0 P rr 0 (31) Where is the point in the original frame, and P i is the point in the nominal plane. Now the circle fitting algorithm may be applied, in this case to the x and y points of the transformed circle to find the circle centre in the x-y plane. The z coordinate of the circle centre is the average of the cloud of points along the z-axis. C.2 Error Analysis In order to assess the potential for error in the calculation, a comparison of the calculated circle centers using the three proposed methods is prepared, all subject to combinations of variable random noise levels (5 levels) and circular arc segment size (6 levels). Coope (1993) found that the advantage to using the LLS method over TLS for circle fitting was that the LLS fit was more robust to outliers, or to noisy data. The disadvantage of either method is that for smaller sections of the circular arc, the estimated radius is smaller than the true radius. It is the purpose of this examination to determine which of the three methods is most appropriate for our requirements. The noise levels ranging from ±0% to ±25% of the circular radius, and arc sizes from 30°-180° as illustrated in Table 18 below, are applied to an arc of 50 points at a radius of 100 units. The size of the circular arc segments were chosen to represent the physical arcs of data points one could potentially attain with a moving limb. For example, a subject may, for whatever reason, be constrained to flex their elbow only 30°-60° or may be able to flex in the range of 135°-180°. All circle data is presented in two dimensions only as circle fitting is performed in a plane nominally normal to the data points. In this test, 35 repeated measurements of the 30 noise and arc size combinations are performed with each of the three circle fitting methods. In each measurement combination, a set of random numbers is generated and held constant for each of the noise/arc size combinations. The noise added to the data is then simply the set of random numbers multiplied by the noise factor. In each combination, the calculated centre error is taken as the norm distance of the calculated center from the true centre (0,0) as a percentage the ratio of the norm distance against the predetermined radius. The equation (12) below illustrates the calculation procedure for the calculated centre position error. / 2 2 Error =^ ^ x l 0 0 % (32) R 128 The mean and standard deviation for each cell of noise/arc size combination are calculated for the 35 data points collected. Results for circle fitting using each of the three fitting methods are presented in Table 18 below. Results for sphere fitting are presented in Table 19. Table 18: Circle fitting error analysis tables for LLS, LPB, and TLS methods. Error is calculated as the distance of the calculated centre from the actual centre as a percentage of the prescribed arc radius. LLS error analysis for circle fitting methods - Mean (±SD) Noise level Arc Size (°) +0% ±1% ±5% +10% ±25% 30 2.3E-13 (5.1E-29) 10.4 (4.5) 72.9 (6.8) 90.5 (4.6) 97.2 (3.2) 45 4.6E-14 (0.0E+00) 2.7 (1.9) 35.2 (6.7) 67.2 (7.0) 90.5 (4.8) 60 6.7E-14 (0.0E+00) 1.2 (1.0) 15.0 (5.1) 40.0 (7.0) 77.5 (6.5) 90 1.1E-14(0.0E+00) 0.5 (0.4) 3.9 (2.3) 12:2 (4.6) 43.1 (7.3) 135 6.0E-14 (0.0E+00) 0.2(0.1) 1.3 (0.8) 3.4 (1.8) 14.3 (5.2) 180 1.7E-14 (9.6E-30) 0.1 (0.1) 0.8 (0.5) 1.7 (1.0) 6.1 (3.0) LPB error analysis for circle fitting methods - Mean (±SD) Noise level Arc Size (°) ±0% ±1% ±5% ±10% ±25% 30 6.4E-14 (1.3E-29) 6.4 (4.5) 58.1 (9.0) 83.2 (8.2) 94.8 (5.7) 45 2.5E-15 (2.0E-30) 2.4 (1.8) 22.0 (7.9) 51.7(9.1) 83.8 (8.5) 60 2.4E-14 (2.2E-29) 1.3 (0.9) 9.0 (5.5) 25.9 (8.4) 65.2 (9.6) 90 7.8E-15 (1.6E-30) 0.5 (0.4) 3.1 (2.1) 7.7 (4.7) 29.0 (9.2) 135 5.5E-15 (3.2E-30) 0.2 (0.2) 1.2 (0.8) 2.7 (1.8) 9.2 (5.0) 180 3.2E-15 (2.0E-30) 0.1 (0.1) 0.7 (0.5) 1.6 (0.9) 4.6 (2.6) TLS error analysis for circle fitting methods - Mean (±SD) Noise level Arc Size (°) ±0% ±1% ±5% +10% ±25% 30 2.8E-13(1.5E-28) 4.2 (3.2) 1.3E+09 (2.6E+09) 9.0E+08(1.2E+09) 3.9E+08 (7.5E+08) 45 2.3E-13 (5.1E-29) 1.9 (1.4) 1.7E+08 (7.1E+08) 6.1E+08 (1.2E+09) 1.1E+09 (2.0E+09) 60 2.8E-14 (1.6E-29) 1.0(0.8) '5.0 (3.7) 7.0E+08 (3.0E+09) 1.5E+09 (2.3E+09) 90 2.0E-14 (0.0E+00) 0.4 (0.3) 2.2 (1.6) 4.3 (3.2) 2.9E+08 (1.2E+09) 135 2.7E-14 (2.2E-29) 0.2 (0.1) 1.0 (0.7) . 1.9(1.5) 4.6 (3.9) 180 1.3E-14 (8.0E-30) 0.1 (0.1) 0.7 (0.4) 1.4 (0.8) 3.4 (2.2) Table 19: Sphere fitting error analysis tables for LLS, LPB, and TLS methods. Error is calculated as the distance of the calculated centre from the actual centre as a percentage of the prescribed spherical radius. LLS error analysis for sphere fitting methods - Mean (±SD) Noise level Arc Size (°) ±0% ±1% ±5% ±10% ±25% 30 5.2E-13 (4.1E-28) 8.5 (3.8) 69.0 (6.9) 88.8 (5.1) 96.8 (3.9) 45 2.1E-13 (7.7E-29) 2.2 (1.5) 31.0(6.0) 63.2 (7.2) 89.1 (5.7) 60 1.2E-13 (5.1E-29) 1.0 (0.8) 12.6 (4.4) 35.9 (6.4) 74.9 (7.4) 90 8.7E-14 (3.8E-29) 0.5 (0.3) 3.3 (1.9) 10.3 (4.0) 39.4 (7.2) 135 2.0E-13 (1.3E-28) 0.3(0.1) 1.3(0.6) 3.1 (1.5) 12.2 (4.5) 180 1.4E-14(1.4E-29) 0.2 (0.1) 0.9 (0.3) 1.8 (0.7) 5.4 (2.3) 129 LPB error analysis for sphere fitting methods - Mean (dSD) Noise level Arc Size (°) ±0% ±1% ±5% ±10% ±25% 30 1.2E-13 (0.0E+00) 5.0 (3.3) 53.7 (8.8) 81.7 (8.4) 94.7 (5.8) 45 7.2E-15 (0.0E+00) 1.8(1.4) 18.3 (6.7) 47.0 (9.0) 82.8 (9.1) 60 2.4E-14 (1.6E-29) 1.0 (0.7) 7.4 (4.0) 21.9(7.4) 62.2 (10.3) 90 3.0E-14 (6.4E-30) 0.6 (0.3) 2.9 (1.5) 6.8 (3.4) 25.5 (8.2) 135 2.6E-15 (2.4E-30) 0.3 (0.1) 1.7 (0.7) 3.4 (1.3) 9.5 (3.9) 180 1.0E-15 (6.0E-31) 0.3(0.1) 1.3 (0.6) 2.6 (1.1) 6.7 (2.8) TLS error analysis for sphere fitting methods - Mean (±SD) Noise level Arc Size (°) ±0% +1% ±5% ±10% ±25% 30 0.0E+00 (0.0E+00) 3.3(2.8) 16.2 (13.6) 1.9E+06 (1.1E+07) 8.7E+06 (4.0E+07) 45 0.0E+00 (0.0E+00) 1.5(1.2) 7.6 (6.1) 16.4(14.1) 5.3E+04 (3.1E+05) 60 0.0E+00 (0.0E+00) 0.9 (0.7) 4.5 (3.3) 9.0 (6.6) 27.8 (33.4) 90 O.OE+00 (0.0E+00) 0.5 (0.3) 2.2 (1.4) 4.4 (2.8) 11.1 (7.6) 135 0.0E+00 (0.0E+00) 0.3(0.1) 1.2 (0.6) 2.5(1.2) 6.2 (3.2) 180 0.0E+00 (0.0E+00) 0.2 (0.1) 0.9 (0.3) 1.7 (0.7) 4.4 (1.9) C.2.1 Student's t-test of mean differences In order to determine which method gives a more accurate estimate of the true centre, a Student's /-test is performed on both the circle fitting and sphere fitting data in order to determine any statistically significant differences between mean estimates of the arc centre. Table 20 and Table 21 below illustrate the results of the /-test on all combinations of noise and arc size for the circle and sphere fitting results respectively for LLS and LPB methods. Table 20: LPB vs. LLS for circle fitting. Results of Student's t-test for differences between means where: nd - no difference between means, LLS or LPB - indicates difference in means and more accurate method. Noise level Arc Size (°) ±0% ±1% ±5% ±10% ±25% 30 nd LPB LPB LPB nd 45 nd nd LPB LPB LPB 60 nd nd LPB LPB LPB 90 nd nd nd LPB LPB 135 nd nd nd nd LPB 180 nd nd nd nd nd Table 21: LPB vs. LLS for sphere fitting. Results of Student's t-test for differences between means where: nd - no difference between means, LLS or LPB - indicates difference in means and more accurate method. Noise level Arc Size (°) ±0% ±1% ±5% ±10% ±25% 30 nd LPB LPB LPB nd 45 nd nd LPB LPB LPB 60 nd nd LPB LPB LPB 90 nd nd nd LPB LPB 135 nd nd nd nd LPB 130 180 nd nd LLS LLS nd In the Table 20 and Table 21 above, LPB is shown to have better accuracy than LLS at low arc sizes and higher noise. In order to test if the TLS method is more accurate than the LPB method,, a second set of t-test's is performed as above. The results of these tests are illustrated in Table 22 and Table 23 below. Table 22: LPB vs. TLS for circle fitting. Results of Student's X-test for differences between means where: nd - no difference between means, TLS or LPB - indicates difference in means and more accurate method. Note: ( - ) indicates the TLS method failed to find a reasonable solution, so the comparison between means is unnecessary. Noise level Arc Size (°) ±0% ±1% +5% • ±10% ±25% 30 nd TLS - - -45 nd nd - - -60 nd nd TLS - -90 nd nd nd TLS -135 nd nd nd nd TLS 180 nd nd nd nd nd Table 23: LPB vs. TLS Sphere fitting. Results of Student's X-test for differences between means where: nd - no difference between means, TLS or LPB - indicates difference in means and more accurate method. Note: (-) indicates the TLS method failed to find a reasonable solution, so the comparison between means is unnecessary. Noise level Arc Size (°) ±0% ±1% ±5% ±10% ±25% 30 nd nd TLS - -45 nd nd TLS TLS -60 nd nd TLS TLS TLS 90 nd nd nd TLS TLS 135 nd nd TLS TLS TLS 180 nd nd TLS TLS TLS C.3 Discussion In the analysis presented in Table 18 and Table 19 above, the offset from the true circle centre represents the accuracy of a given circle/sphere fitting method, as smaller offset values represent closer proximity to the true circle centre at coordinates (0,0). The standard deviation represents a measure of the precision (repeatability) of the circle fitting method subject to variable random noise. As expected, these tables show a trend towards decreased accuracy and precision in all three methods as the noise level increases and the arc segment length decreases. Table 20 through Table 23, however, illustrate that the order of preference according to accuracy is to apply TLS first, then LPB and finally LLS. The TLS method, however, breaks down dramatically as the noise level increases. Though the results are comparable in all three methods at lower noise. 131 levels, as illustrated in Table 20 though Table 23, the error at high noise is severe using TLS. Considering the noisy nature of the calibration process, the data suggest not to rely on the TLS method alone for circle or sphere fitting. Indeed, an examination of the data collected from elbow joint centre calibration with the surgeon (AN) during the December 23, 1999 OR experiment, dramatically illustrates this effect. The elbow centre locations relative to the marker frame at the elbow (Ce2) illustrated in Table 24 below were calculated using the three fitting methods. Table 24: Calculated elbow centre (Ce2) from calibration during OR experiment (23.12.99) on surgeon subject (AN) using each circle fitting method. Method Ce" (x, y, z) (mm) LLS (-26.22, -9.49, -27.12) LPB (-26.22, -9.49, -27.12) TLS . (3.74E+10, -0.07E+10, -0.21 E+10) It is clear that the TLS method is ineffective on this calibration. As-a further confirmation of the LLS and LPB measurement calculations taken from the OR data, the length of the subject's (AN) forearm (elbow to wrist joint centre) from both the lab and OR calibration measurements is determined, and found to be 263.9mm and 254.3mm respectively; a 3.8% variation in measured length. The convergence problem with TLS noted in this example confirms the findings of Coope (1993) that LLS is more robust to outliers, or noise, in the data. In cases where the TLS method breaks down, the centre may be recalculated using either LLS or LPB. The comparative results illustrated in Table 20 and Table 21 show that LPB is more accurate than LLS, and is thus the preferred method after TLS. In the event that the LPB calculation breaks down, which may occur due to poorly arranged displacements, the LLS method may be used. C.4 Conclusion Considering the higher degree of accuracy shown with the TLS method over LLS and LPB methods, the TLS method is selected for calculating the circle or sphere centre. In the event that the TLS method fails to converge to a reasonable solution, as illustrated in the example of Table 24, the centre is recalculated using the LPB method. Failing that, the LLS may be used as it is the least accurate of the three, but the most robust. 132. Appendix D Clinical Joint Angles D.l v Calculating calibrated joint angles Traditionally in motion analysis studies, angular limb position is expressed either by sequential Cardan (XYZ, Y Z X , etc.) or Euler (ZYZ, Y X Y , etc.) angles taken directly from the rotation matrix between two frames (Schmidt 1999, Peterson 1996), or by directly solving for the joint angle (Anglin 1993). Generally in the rotation matrix (Eulerian) method, marker locations define joint coordinate frames at each joint centre, and the homogeneous transform between frames are used to determine the Euler angles which represent joint rotation. In the direct method employed by Anglin (Anglin 1993), the vector between joint centers is projected onto a predefined plane, and the joint angle is measured directly on that plane. Let us examine an example to clarify these two methods. Take the example of measuring wrist flexion-extension and radial-ulnar deviation on the right hand. We take the joint coordinate frame at the right hand wrist, W, oriented such that the x-axis points distally along the forearm, the y-axis pointing medially along the mediolateral axis of the wrist, and the z-axis determined by the completing 'right hand' orthogonal representation as illustrated in Figure 38 below. The tracked marker on.the hand, represented in the figure by C h W , carries the hand coordinate frame, H, initially aligned with the wrist coordinate frame. Figure 38: Example of the wrist frame. Ch is the hand marker in the wrist frame, W, in the neutral position. Now rotate the hand in radial-ulnar deviation about the wrist axis, Z w , by an angle of 30°. This new posture is illustrated in Figure 39 below. Figure 39: Hand rotated by 30° in radial deviation about the positive Zw axis. By the Eulerian method, the joint angle is solved as follows. Here, RWH represents the rotation matrix between the wrist and hand coordinate frames, and the Euler angles, <p, 6, represent successive rotations about the z-, y- and x-axis, respectively. Note that these angles represent the Roll-Pitch-Yaw (RPY) orientation representation (Sciavicco p.32) RM) = Rv{9) = 0 s* °* 0 0 0 1 se 0 1 0 -se 0 ce 1 0 o~ 0 c, 0 s„ c,„ V Ctj>C6 C t s e s S<j,C8 S ~se ' c (/) = Atan2(r 2 1 ,r„) = 30 0 S J.S aC. cac • CiS... Atan2(-r 3 1 ,Jr 3 2 2 + r 3 3 2 ) : 31 'A/ 32 1 ' 3 3 Atan2(r3 2,r3 3) = 0° Similarly, one could solve the radial deviation angle illustrated in Figure 39 by directly projecting point d , ^ = [x h W yhW Z h W ] onto the x-y plane of frame W, and taking the angle from the x-axis to represent radial deviation. The direct solution to this problem is equated as follows: 134 « = [1 0 Of r w w t-\ m v = K n (0 °J 6> = + C O S •cos u-v vnwllllviiy u-v viiwllllvny i / ( « x v ) ' Z * ' > 0 (radial deviation) if (uxv)-Zw < 0 (ulnar deviation) Where u is the x-axis of frame W. Similarly, the projection onto the x-z plane of frame W is used to directly solve for wrist flexion / extension as follows: u = [\ 0 Of v = [xh 0 Z„ ] 9 RU + C O S - C O S u-v if(uxv)-Yw>0 (flexion) vnllhiy u-v if (uxv)-Yw < 0 (extension) It is illustrative to compare these methods in determining the combined orientation of a wrist in both radial-ulnar deviation and flexion-extension. Taking the example of rotating the hand 30° in radial deviation, we now rotate the wrist from 0° through to 90° flexion, according to Eulerian rotation definitions. Please note that this extreme flexion angle is for illustrative purposes and does not represent common physical flexion angles. Figure 40 below illustrates the graphical representation of this rotation in both the x-y and x-z planes of the wrist frame (top 1. and r. panels, respectively) determined by the Eulerian method; § = 30°, 0 = 0°—>• 90°. The bottom panel illustrates the correspondence between the joint angles calculated with the Direct method against the plot of marker locations determined by the Eulerian method. 135 Figure 40: Graphical representation of the rotation Rz(<p) = 30° Ry(9) = 0°->90° shown 100 Pitch angle about Y w (deg.) in the x-y plane (top I.) and x-z plane (top. r.) of the wrist frame. The bottom panel illustrates the correspondence between the joint angles calculated with the Direct method from the arc of points determined by the Eulerian method. 136 As the bottom plot in Figure 40 illustrates, it is expected that the correspondence between the joint angles calculated using the Direct method from the arc of points generated by the Eulerian rotations will be equal. The degree of correspondence may be quantified by determining the linear correlation between the Eulerian and Direct joint angles, where r = 1 indicates a direct correlation between the calculated joint angles. Indeed, a direct correlation (r = 1) exists for the radial-ulnar deviation angles (RUdirectX indicating that the 30° R-U angle is calculated accurately throughout the range of flexion from 0° to 90° by the Direct method. For the F-E angle, however, the correlation is not exactly linear (r = 0.9988). Nevertheless, this example illustrates how the Eulerian and Direct methods may be used to calculate the orientation of a point relative to its base frame. D.2 Clinical joint angles As illustrated, both the Eulerian and the Direct methods may be used to calculate the angles representing the orientation of the hand relative to the wrist. The question that arises, however, is whether these representations of the joint angles are meaningful in a clinician or ergonomic evaluation, where the degree of angular displacement is often based on direct visual measures taken in the plane of interest. As an example, let us reexamine the data presented in the top panels of Figure 40 and imagine that the coordinate frame W and the marker Ct, w illustrated in Figure 38 are fixed to your own right hand. Now, try to trace the path of the data with that point on your right hand. As your right hand approaches 90° flexion, would you still say, based on the posture of your hand, that you are maintaining 30° of radial deviation? More likely, your estimate of radial deviation would be closer to 0° based on this visual evaluation . Here lies the problem inherent in the measuring joint angles by Eulerian and Direct calculation methods; the representation of compound postures depends on sequential rotations about various axes while physically one is interested in the rotation away from a neutral plane of interest, independent of the other joint angles. Since, in this study, we are interested in the clinical posture angles based on the visual inspection methods of the R U L A technique, I propose to use a new method for calculating joint posture angles at the wrist and shoulder which I will refer to as the Clinical joint angle calculation. In the present example, let us define the Clinical joint angle for radial-ulnar (RU) deviation as the elevation angle of the hand marker from the x-z plane at the wrist, and flexion-extension (FE) as the elevation angle of the hand marker from the x-y plane at the wrist, as illustrated in Figure 41 below. Now, the measures of RU and FE are independent of each other, and more accurately represent the physical joint angles seen clinically. 137 Figure 41: Clinical joint angles for flexion-extension (Q^B) and radial-ulnar deviation (Q^u) of the hand in the wrist frame. 0FE is the elevation angle ofpoint C^from the X-Y plane, and GRU is the elevation of point C^ from the X-Z plane. Flexion is in the -Z axis direction, and radial deviation is in the + Y axis direction. In order to calculate the Clinical joint angles, I take the projection of the marker point into the neutral plane of interest and determine the angle between the vectors described by these points. The following equations illustrate the calculation of the flexion-extension angle, OFE-u = v = w w w xh yh z h w w " A ( + C O S u-v -cos Vinlniy u-v if zhw < 0 (flexion) if zhw =0 (neutral) if zhw > 0 (extension) Similarly, for radial-ulnar deviation angle off of the X - Z plane, 6\ 'RU-u v = w w w x h yh z h w A w Xu 0 z. 0 RU + C O S •cos f \ u-v u-v vinlriiy if yh > 0 (radial deviation) if Jn - 0 (neutral) if yh <0 (ulnar deviation) Returning to the example illustrated in Figure 40, we recalculate the flexion-extension and radial-ulnar deviation using the Clinical joint angle method. These angles are presented in Figure 42 below. 138 90 i r i i / 80 CD Q) -o 70 -• f * y nt angles 60 - ^direct , ' / nt angles 50 ' / ^^clinical O lated 40 -• j * j * y • J » S Calcu 30 / / R U direct * * 20 ^s ^ c h n i c a l 10 0 - f i i 1 ! 1 — ^ 0 ' 2 0 40 60 80 100 Pitch angle about (deg.) Figure 42: Comparison of Direct joint angles against Clinical joint angles on the data presented in Figure 40. FEdirect and RUdirect are the flexion-extension angles and radial-ulnar deviation angles calculated by the Direct joint angle calculation, while FEcuniCaiand RUciiniCai are the same angles determined by the Clinical joint angle calculation The results of the Clinical joint angle calculations illustrated in Figure 42 show a more meaningful interpretation of the compound joint angles defining hand posture relative to the wrist. Thus, the Clinical joint angle representation is more useful in clinical situations or ergonomic analysis such as the RULA technique where joint angles are determined according to visual interpretations of posture. i39 Appendix E Nonparametric analyses and task time annotations E . l Task time annotation for cystic duct dissection Task Time annotations for laparoscopic cholecystectomy - Dec 23, 1999 Section 1 - Cystic duct dissection (see Matlab program: p e r c _ c o r r u p t e d . m for output details) Start l-'.nil Duration % lime time (sec.) corrupted Initial incision 8:15:27 start PSVSAMPL.EXE 8:28:36 Cystic duct dissection 8:30:40 8:43:45 785 > good data 8:30:40 8:34:00 200 31.3 array falls off tool 8:34:00 8:34:46 46 76.9 good data ,8:34:46 8:37:00 134 0.0 TC - grasper -> clip applier 8:37:00 8:37:17 17 81.5 good data 8:37:17 8:37:52 35 55.0 TC - clip applier -> new clip 8:37:52 8:38:09 17 37.1 good data 8:38:09 8:38:20 11 15.5 TC - clip applier new clip 8:38:20 •8:38:37 17 82.4. good data 8:38:37 8:38:46 9 14.8 TC - clip applier -^scissors 8:38:46 8:39:02 16 75.5 good data 8:39:02 8:39:26 24 7.7 140 TC - scissors -> graspers 8:39:26 8:39:37 11 67.1 array on tool 8:39:37 8:39:48 11 0.0 lens fogged 8:39:48 8:40:44 56 7.9 good data . 8:40:44 8:40:57 13 19.4 TC - grasper -> clip applier 8:40:57 8:41:16 19 91.3 good data 8:41:16 8:41:31 15 82.1 TC - clip applier -> new clip 8:41:31 8:41:48 17 87.6 good data 8:41:48 8:41:59 11 78.7 TC - clip applier -> new clip 8:41:59 8:42:13 14 • 78.4 good data 8:42:13 8:42:23 10 35.2 TC - clip applier -> scissors 8:42:23 8:42:55 32 57.0 good data 8:42:55 8:43:20 25 4.0 TC - scissors -> graspers 8:43:20 8:43:45 25 77.6 Non-surgical n = 13 Total time: sec. (%) 298 (38%) mean (sd) % corrupted 63 % (3 0%) median % 76.9% Normal surgical n = 11 Total time: sec. (%) .487(62%) mean (sd) % corrupted 31% (29%) median % 19.4% E.2 Nonparametric analysis of surgical/non-surgical cystic duct dissection tasks The frequency distributions shown in Figure 43 show that the distributions are heavily skewed in opposite directions. Since normality can not be assumed for either distribution, a nonparametric test is appropriate for comparing the two distributions. 141 F r e q u e n c y of data corruption during normal tasks in C C D F r e q u e n c y o f d a t a corrupt ion dur ing n o n - s u r g i c a l l a s k s i n C C D 3 .5 3 2 O ^ ^ H , 1 4 0 6 0 % corrupted Figure 43: Frequency distribution of percentage of total data corruption during periods of(L.) normal surgical tasks and (R.) non-surgical tasks (e.g. errors, tool changes) observed during cystic duct dissection. Higher frequency of data corruption occurs in non-surgical tasks. Using the One-tailed Mann-Whitney test (Zar, 1999, pg. 146-150) Source: Zar J.H. (1999) Biostatistical Analysis. 4 th ed.. Prentice Hall, Upper Saddle River, NJ Group 1: non-surgical tasks (e.g. errors, tool changes) Group 2: direct, normal surgical tasks Ho: Percentage of totally corrupted data is not greater in non-surgical tasks HA: Percentage of totally corrupted data is greater in non-surgical tasks a = 0.051 Percentage of totally corrupted data (%) Non-surgical tasks Normal surgical tasks Totally Rank corrupted [%1 Totally Rank corrupted [%1 0.0 1.5 0.0 1.5 7.9 5 4.0 3 37.1 11 7.7 4 57.0 13 14.8 6 67.1 14 15.5 7 75.5 15 19.4 8 76.9 16 31.3 9 77.6 17 35.2 10 78.4 18 55.0 12 • 81.5 20 78.7 19 82.4 22 82.1 21 87.6 23 91.3 24 142 n, = 13 R i = 199.5 |n 2=11 R 2 = 100.5 = (11)(13) + ( 1 1 ) ( 1 2 1 + 1 ) -100.5 = 108.5 Find the critical value: ^0.05(1) , ) 1,13 = 101 As U' > U'crn, reject H0 P • ^ ( D . i i . 1 3 = 1 ° 8 - 5 ° - 0 1 < P < ° - 0 2 5 ' Conclusion: The percentage of corrupted data is greater in non-surgical tasks, which are recognized as periods which may be excluded from the posture analysis. E .3 Nonparametric analysis of ergonomic stress levels Kruskal-Wallis single-factor A N O V A by ranks (pl95) Source: Zar J.H. (1999) Biostatistical Analysis, 4 th ed.. Prentice Hall, Upper Saddle River, NJ Ho: The ergonomic stress level is the same in all three surgical tasks Ha: The ergonomic stress level is not the same in all three surgical tasks ct= 0.05 CCD GBD GBR value rank value rank value rank 3% 3 1% 1.5 1% 1.5 6% 4 54% 8 38% 5 52% 7 66% 11 43% 6 61% 10 87% 14 60% 9 74% 12 93% 15 82% 13 n 5 5 5 R 36 49.5 34.5 R2/n 259.2 490.05 238.05 Kruskal-Wallis test statistic: 143 H = 12 R: N{N + \)^n, N = fjni=\5 Z— - 3 ( ^ + 1) ;=1 12 H = -15(16) / / = 1.365 # 0 . 0 5 , 5 , 5 , 5 = 5 - 7 8 0 (36)2 { (49.5)2 | (34.5)2 -3(16) Since H = 1.365 < H c r i t = 5.780 - Accept H 0 Conclusion: The ergonomic stress level is the same in all three surgical tasks E.4 Nonparametric analysis testing across joints Kruskal-Wallis single-factor A N O V A by ranks (pl95) Source: Zar J.H. (1999) Biostatistical Analysis, 4 th ed.. Prentice Hall, Upper Saddle River, NJ1 H 0 : The ergonomic stress level is the same in all five joint angles H a : The ergonomic stress level is not the same in all five joint angles a = 0.05 1 ' Shoulder value rank 2 Elbow value rank 3 Wrist f-e value rank 4 Wrist r-u value rank 5 Forearm value rank 1% 1.5 38% 5 74% 12 52% 7 6% 4 1% 1.5 54% 8 82% 13 60% 9 43% 6 3% 3 61% 10 93% 15 66% 11 87% 14 n 3 3 3 3 3 R 6 23 40 27 24 RA2/n 12.0 176.3 533.3 243 192 N 15 H = -12 N(N+ 1)^*1, k n2 £ _ L . _ 3 ( t f + 1 ) tf = 5>,=15 H = 12 15(16) H = 9.830 (6)2 (23)2 (40)2 (27)2 (24)2 j 1 j |- 3(16) Correction factor for tied ranks (eqn. of ties. 10.40, p. 198) where is the number of ties in the ith group C = l '-1 =\-K\ f ; = 0.998 15J -15 H= — = 9.851 c C -^0.05,3,3,3,3,3 = 8.333 Result: Reject H 0 (0.02 < P < 0.01) Accept H a : The ergonomic stress levels are not the same in all five joint angles Nonparametric Tukey-type multiple comparisons, using the Nemenyi test (Zar, 1999, p. 223) Calculate standard error with tied ranks (eqn. 11.28) SE = f m \ N(N + l) tt 12 \2(N-\) \ n A nBj ^15(16) (23 12 •2 )Yl 12(14) - + -3 3 = 3.648 Samples in ranked order 1 2 5 4 3 Rank sums (R) 6 23 24 27 40 mean Ranks (mR) 2 7.67 8 9 13.3 Comparison (B vs. A) Difference (mR B-mR A) SE q q 0.05,5 (table B.15) Conclusion 3 vs. 1 11:3 3.648 3.107 2.807 Reject Ho (*) 3 vs. 2 5.7 3.648 1.553 2.807 Accept Ho 3 vs. 5 5.3 3.648 1.462 2.807 Accept Ho 3 vs. 4 4.3 3.648 1.188 2.807 Accept Ho 4 vs.^ 1 7.0 3.648 1.919 2.807 Accept Ho 4 vs. 2 1.3 3.648 0.365 2.807 Accept Ho 4 vs. 5 . 1.0 3.648 0.274 2.807 Accept Ho 5 vs. 1 • 6.0 3.648 1.645 2.807 Accept Ho 5 vs. 2 0.3 3.648 0.091 2.807 Accept Ho 2 vs. 1 5.7 3.648 1.553 2.807 Accept Ho Conclusion: Ergonomic stress is not the same in the shoulder compared to the wrist flexion-extension Appendix F Coordinate transforms and nomenclature F . l Nomenclature Coordinate Frames Tracked frames 0 Global coordinate frame (Polaris base frame) 1 Body frame (on torso) 2 Proximal forearm frame at elbow 3. Distal forearm frame at wrist 4 Hand frame Virtual frames 5 Shoulder E Elbow W Wrist Other nomenclature TAB Transform from coordinate frame A with respect to coordinate frame B tc Calibration time t Time (variable) C*A(t) (i=s,e,w,h) Joint centre coordinates for shoulder, elbow, wrist and hand1 [XJ yj zi] in frame A at any time, t (or tc) " • . X A X axis of frame A (similarly for Y , Z ) Y B A Y axis of frame B relative to frame A Z B A Z axis of frame B relative to frame A Xj A (i=s,e,w) x-coordinate of joint i in frame A yA(i=s,e,w) y-coordinate of joint i in frame A z A (i=s,e,w) z-coordinate of joint i in frame A 'The joint centre of the hand is not calculated, it is simply the nomenclature adopted in order to remain consistent with the other anatomical landmarks tracked in this study F.2 Homogeneous transforms When dealing with point measurements in 3D coordinates of Cartesian space (x,y,z), it is often necessary to establish the position of a point in multiple coordinate frames. Fortunately, a straight forward method is available, known as a homogeneous transform, which describes the relative rotation and translation between various coordinate frames of interest, illustrated in Figure 44 will demonstrate this point. 146 The example In Figure 44, position of point C, is established relative to both coordinate frame A (CjA) and coordinate frame B (Q B). The relationship between these two points is given by the following equation. C- = TABC,B Or inversely by: c;=(rAByc; (33) (34) Where TAB is the homogeneous transform between frame A and frame B. The homogeneous transform describes mathematically the rotation and translation required to align coordinate frame A with frame B. The homogeneous transform matrix is given by the following equation. T = 1 AB A X0,B RAB A y0.B A Z0,B o7' 1 (35) Here, XQ,BA is the x-coordinate of the origin of frame B with respect to A, similarly for yo and z«. The 3x3 matrix, RAB, is the rotation matrix which describes the order of rotations required to align frame A with frame B, that is the rotation matrix of B with respect to A. A common sequence of rotations employed in rotation matrix calculation is described by RPY (roll, pitch, yaw) rotations about the Z, Y, and X axes, respectively. Other Eulerian rotations (about X Y X , ZXZ, etc.) may also be used to determine RAB- The equations below illustrate the calculation of the RPY rotation matrix from elementary rotations, where eg represents cos(O) (etc.). 147 RRPY=R2(<f>)R(0)Rx(iy) 0" RM = s* ct 0 0 0 1 " ce 0 Ry{0) = 0 1 0 , o ce_ "l 0 0 0 -•v 0 through matrix multiplication sipce V cas,„ (36) Thus, knowing the rotations and translations required to align frame A with frame B, one can determine the homogeneous transform. Matlab robotics toolbox functions such as TRANSL, ROTX, ROTY, ROTZ and RPY2TR, make the calculation of the homogeneous transform quite straightforward. F.3 Rotation about an arbitrary axis In some instances, we require the rotation matrix be expressed as a rotation of a given angle about an arbitrary axis in space. Figure 45 below illustrates this concept on two vectors u and v. An example of such an instance is in aligning the plane of data with one primary plane of the base frame (XY, Y Z or XZ) during planar circle fitting. Here, the angle between the norm to the plane of data (u) and one of the base axes (v) is determined, then the rotation matrix (and hence homogeneous transform) to bring the data into that base plane may be found. Z X Figure 45: Rotation about an arbitrary axis r by angle 9 between two vectors, u and v. 148 Taking the two vectors, u and v, define the angle and axis of rotation to align u with v as follows. - i 6 = cos r - u x v u • v V l H l l V (37) Then, the rotation matrix of u with respect to v defined by rotation by 0 about the vector r is defined by the following equation (Sciavicco 1996, p.28). rxry(\-ce) + rzse rl{\-ce) + cg ryse y rrse (38) Where r = [rx ryrz] and c#and ^represent cos(O) and sin(O), respectively. The equation above is available as the Matlab robotics toolbox function, R O T V E C . From this, we determine the position of C,-v from the following. 0" T = uv 0 0 0 0 0 1_ then c; = c; = (39) 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0080974/manifest

Comment

Related Items