Citation
Development of a novel prosthetic wrist device incorporating the dart thrower's motion

Material Information

Title:
Development of a novel prosthetic wrist device incorporating the dart thrower's motion
Creator:
Davidson, Matthew Lee ( author )
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
1 electronic file (148 pages) : ;

Subjects

Subjects / Keywords:
Wrist ( lcsh )
Prosthesis ( lcsh )
Prosthesis ( fast )
Wrist ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
The purpose of this research was to identify limitations people with arm amputations face completing daily living tasks and to design a new prosthesis which alleviates these deficiencies. State of the art prosthetic devices can mimic many of the motions of an intact limb but are controlled by a limited number of signals from the muscles in the residual limb. The majority of current research is focused on improving the control of these devices by increasing the number of inputs or using software to interpret the limited inputs in a more meaningful way. This research instead determined that the mechanics of the prosthesis could be simplified while maintaining functionality and a simple control system. Specifically, this research tested the hypothesis that the three degrees of freedom in the wrist (flexion-extension, radial-ulnar deviation, and rotation), could be combined into a single degree of freedom, known as the Dart Thrower's Motion, in a way that preserves most of the wrist's motion and functionality and could be controlled with a simple input method. There are currently no commercially available wrist flexion devices which utilize this motion. The studies presented in this dissertation surveyed people with upper limb amputations and found that they are less satisfied performing tasks that utilize the Dart Thrower's Motion with their prosthesis. The major angle of the Dart Thrower's Motion was identified in able-bodied individuals to be 22 degrees offset from the anatomical flexion-extension plane. Finally, a new prosthetic wrist device was developed based on this angle. This new prosthesis improved functionality over a traditional flexion wrist and was no more difficult to use than a device without a wrist. This research helps to alleviate many of the barriers to inclusion which people living with upper limb deficiency regularly face.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: Adobe Reader.
Statement of Responsibility:
by Matthew Lee Davidon.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
on10030 ( NOTIS )
1003045385 ( OCLC )
on1003045385
Classification:
LD1193.E56 2017d D38 ( lcc )

Downloads

This item has the following downloads:


Full Text
Development of a novel prosthetic wrist device
INCORPORATING THE DART THROWER'S MOTION
by
Matthew Lee Davidson B.S., Reed College, 2006 M.S., University of Colorado at Denver, 2011
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfilment of the requirements for the degree of Doctor of Philosophy Bioengineering Program
2017


2017
MATTHEW LEE DAVIDSON ALL RIGHTS RESERVED
ii


This thesis for the Doctor of Philosophy degree by
Matthew Lee Davidson has been approved for the Bioengineering Program by
Cathy Bodine, Chair Richard F. ff. Weir, Advisor James J. Carollo Kendall S. Hunter William Sullivan
Date: July 29, 2017
mi


Davidson, Matthew Lee (Ph.D., Bioengineering Program)
Development of a Novel Prosthetic Wrist Device Incorporating the Dart Thrower's Motion Thesis directed by Research Associate Professor Richard F. ff. Weir
Abstract
The purpose of this research was to identify limitations people with arm amputations face completing daily living tasks and to design a new prosthesis which alleviates these deficiencies. State of the art prosthetic devices can mimic many of the motions of an intact limb but are controlled by a limited number of signals from the muscles in the residual limb. The majority of current research is focused on improving the control of these devices by increasing the number of inputs or using software to interpret the limited inputs in a more meaningful way. This research instead determined that the mechanics of the prosthesis could be simplified while maintaining functionality and a simple control system. Specifically, this research tested the hypothesis that the three degrees of freedom in the wrist (flexion-extension, radial-ulnar deviation, and rotation), could be combined into a single degree of freedom, known as the Dart Thrower's Motion, in a way that preserves most of the wrist's motion and functionality and could be controlled with a simple input method. There are currently no commercially available wrist flexion devices which utilize this motion. The studies presented in this dissertation surveyed people with upper limb amputations and found that they are less satisfied performing tasks that utilize the Dart Thrower's Motion with their prosthesis. The major angle of the Dart Thrower's Motion was identified in able-bodied individuals to be 22 degrees offset from the anatomical flexion-extension plane. Finally, a new prosthetic wrist device was developed based on this angle. This new
IV


prosthesis improved functionality over a traditional flexion wrist and was no more difficult
to use than a device without a wrist. This research helps to alleviate many of the barriers to inclusion which people living with upper limb deficiency regularly face.
The form and content of this abstract are approved. I recommend its publication.
Approved: Richard F.ff. Weir
v


To my parents, Niza and Richard Davidson, for making this possible.
VI


Acknowledgements
Thank you to my committee for their support, encouragement, and patience as I pursued this research. Their contributions helped steer this work to a successful completion. I am especially grateful to Richard Weir for his invaluable insights into the world of prosthetics and to Cathy Bodine for making me feel like I could accomplish this work. Thanks also to the other members of the biomechatronics lab for their help, both technical and theoretical, and to Jac Corless and the Mechanical Engineering Department's machine shop for building parts for me. A very big thanks goes to all my friends who supported me and volunteered their time to participate in my experiments: your volunteered time was invaluable. Finally, I want to thank Jyll Tuggle at Under the Umbrella Cafe for feeding me and providing a safe, comfortable space for me to work countless thousands of hours over the last eight years, and Peter Chan of Peter's Chinese Cafe for bringing me delicious food so many times when I was too busy or tired to feed myself.
VII


Table of Contents
I. MOTIVATION AND RESEARCH GOALS...........................................................1
1.1 Summary of Goals.....................................................................3
1.1.1 Specif ic Aim 1: Identify what ADL tasks people with amputations find challenging via online
survey and determine what improvements they are interested in seeing in their devices..3
1.1.2 Specif ic Aim 2: Measurement of the Dart Thrower's Motion in the Context of a Prosthesis-
Specific Functional Test................................................................4
1.1.3 Specif ic Aim 3: Testing a novel prosthetic wrist that incorporates the Dart Thrower's Motion.. 5
II. BACKGROUND..............................................................................7
2.1 Physiology of the Upper Limb.........................................................7
2.1.1 The Wrist........................................................................9
2.1.2 The Dart Thrower's Motion.......................................................11
2.2 The Amputated Upper Limb............................................................14
2.2.1 Etiology........................................................................14
2.2.2 Anatomy and Surgery of the Upper Limb...........................................15
2.2.3 Considerations for People with Upper Limb Amputations...........................16
2.3 Current Prosthetics.................................................................17
2.3.1 Types of Prosthetics............................................................18
2.3.2 Prosthetic Wrists...............................................................19
2.3.3 Limitations of Current Prosthetic Devices.......................................22
2.4 Myoelectric Control.................................................................23
III. USER SURVEYS SUGGEST INCORPORATING THE DTM INTO A PROSTHETIC WRIST WOULD BE USEFUL 26
3.1 Methods: Development................................................................27
viii
3.1.1 Human Factors Testing
27


3.1.2 Survey and Interview Questions
28
3.1.3 Survey Validation....................................................................31
3.1.4 Interview............................................................................32
3.1.5 Power Analysis.......................................................................35
3.2 Methods: Experiment......................................................................36
3.2.1 Inclusion/Exclusion Criteria.........................................................36
3.2.2 Recruitment..........................................................................38
3.2.3 Analysis.............................................................................39
3.3 Results..................................................................................40
3.4 Discussion...............................................................................44
3.5 Conclusion...............................................................................47
IV. MEASUREMENT OF THE DART THROWER'S MOTION IN THE CONTEXT OF A PROSTHETICS-SPECIFIC FUNCTIONAL TEST..................................................................................48
4.1 Methods: Development of a Wrist Measurement System.......................................49
4.1.1 Motion Capture.......................................................................49
4.1.2 Flex Sensitive Resistors.............................................................52
4.1.3 Commercial Electrogoniometers........................................................55
4.2 Methods: Standardized Motions............................................................56
4.3 Methods: Principal Component Analysis....................................................58
4.3.1 Spatial vs. Temporal PCA.............................................................61
4.3.2 A Statement About Error Reporting in PCA.............................................62
4.4 Methods: Experiment......................................................................63
4.4.1 Inclusion/Exclusion Criteria.........................................................63
4.4.2 Recruitment..........................................................................65
4.4.3 Electrogoniometers...................................................................65
4.4.4 Assessment...........................................................................66
IX


4.4.5 Processing.
66
4.4.6 Principal Component Analysis.......................................................68
4.4.7 Coupling Counts and Position Analysis...............................................70
4.5 Results..................................................................................70
4.5.1 PCA on Reach and Grasp..............................................................71
4.5.2 2-Dimensional PCA on Averaged Data..................................................73
4.5.3 2-Dimensional PCA on Full Data Set..................................................74
4.5.4 Coupling............................................................................76
4.6 Discussion...............................................................................77
4.7 Conclusion and Next Steps................................................................79
V. TESTING A NOVEL PROSTHETIC WRIST THAT INCORPORATES THE DART THROWER'S MOTION...............81
5.1 Methods: Development.....................................................................82
5.1.1 Development of the DTM Wrist........................................................82
5.1.2 Hand Modifications..................................................................87
5.1.3 Functional Characteristics of the Prosthetic Wrist and Hand........................88
5.2 Methods: Experiment......................................................................89
5.2.1 Population..........................................................................90
5.2.2 Inclusion Criteria..................................................................91
5.2.3 EMG and Prosthesis Fitting..........................................................92
5.2.4 Myoelectric Control.................................................................95
5.2.5 Limited SHAP........................................................................99
5.2.6 Compensatory Motions...............................................................100
5.2.7 NASA-TLX...........................................................................102
5.3 Results.................................................................................102
5.3.1 SHAP Results.......................................................................103
5.3.2 Shoulder Results...................................................................104
X


5.3.3 NASA-TLX Results.................................................................106
5.4 Discussion..........................................................................107
5.4.1 SNAP...........................................................................108
5.4.2 Shoulder.......................................................................109
5.4.3 NASA-TLX.......................................................................Ill
5.4.4 Corrected Angle................................................................Ill
5.5 Conclusion..........................................................................112
VI. CONCLUSIONS AND FUTURE WORK...........................................................114
6.1 Conclusions.........................................................................114
6.1.1 Survey.........................................................................114
6.1.2 DTM angle......................................................................115
6.1.3 DTM Prosthetic Wrist...........................................................115
6.2 Future work.........................................................................116
6.2.1 Rotator vs DTM.................................................................116
6.2.2 Correction of the DTM Angles...................................................117
6.2.3 Assessment of Compensatory Motions.............................................117
6.2.4 Combination of Wrist and Hand Position in Postural Controller..................119
6.3 FinalThoughts.......................................................................120
REFERENCES 122
XI


List of Figures
Figure 2.1: Homunculus drawing by Penfield showing that neurons directing the control of
the hand dominate the motor cortex (34).......................................9
Figure 2.2: Bones of the hand and wrist...........................................10
Figure 2.3: Range of motion of the anatomical wrist ((14), modified)..............11
Figure 2.4: An example of The Dart Thrower's Motion...............................12
Figure 2.5: Radial plot of the average maximum passive range of motion of 6 cadaver wrists as described by Crisco. The DTM plane is visible as a plane rotated 20-40 degrees from the plane of pure flexion-extension (20)...............................................13
Figure 2.6: Passive wrists. Terminal devices may be attached to these devices and passively positioned to achieve flexion. From left to right: Hosmer constant friction wrist and Hosmer flexion wrist (Hosmer, Campbell, CA), Texas Assistive Devices N-Abler wrist, Medical Bionics Robo-Wrist, and TRS passive flexible wrist.............................21
Figure 2.7: Examples of powered wrists. The left two devices (Otto Bock wrist rotator, VASI children's electric wrist rotator) only provide rotation. The right two devices (Centri Hand, Shanghai hand) combine rotation and flexion-extension............................21
Figure 2.8: The Defense Advanced Research Projects Agency (DARPA) RP2009 APL Modular Prosthetic Limb (MPL) Arm developed as part of the DARPA RP2009 initiative includes a multi-DOF wrist........................................................................22
Figure 2.9: Diagram of myoelectric control for a powered prosthesis. The neural signal from the brain innervates the muscle in the residual limb, changing the surrounding electric
XII


potential and causing it to contract. This change is detected by electromyogram
electrodes on the surface of the arm, electronically amplified and processed, and sent as a control signal to a motor which moves the prosthesis..............................24
Figure 3.1: Power analysis indicating that 21 participants were sufficient to detect a 10% difference in responses with 97% confidence..................................................36
Figure 3.2: Pie Chart showing the how often participants made each of the keyed comments during their interview. Overall, nearly one third of comments were about difficulty using tools or completing other ADL tasks. Another third addressed common complaints with prosthetics (e.g. durability, socket, cost, weight, and electronics).........44
Figure 4.1: Attempts at overcoming the Optotrak limitations. Rigid plates (above) were
attached to body segments which correctly located the segment but could only be seen from a fixed vantage point. Multiple markers attached to each segment and defined as a "rigid body" (below). Lighter lines and circles indicate connections and markers on the opposite side of the limb. This method was also ineffective because the Optotrak software was unable to account for soft tissue artefacts (i.e. minor motion between markers from finger abduction or forearm skin stretch) in determining the location of the rigid body.........................................................................51
Figure 4.2: Circuit diagram for the flex sensitive resistor setup. Vm and Voutwere generated
XIII
and recorded in Labview,
53


Figure 4.3: Angle vs Voltage in flex sensitive resistor (FSR) testing. FSR was attached to a plastic goniometer and the voltage was recorded via Labview at various angles and
moving in the forward (black) and backward (white) directions......................54
Figure 4.4: Mockup of the FSRs attached to the dorsal and ulnar sides of the wrist......54
Figure 4.5: Plot of the wrist angles flexion-extension (FE) and radial-ulnar deviation (RU)
recorded with the FSR system during simulated dart throwing......................55
Figure 4.6: Results of calibration test of the goniometers and torsion meter over 180 degree operating range. Flysteresis was found to be <2.5 degrees over 180 degrees. Relationship between output voltage and angle is highly linear........................56
Figure 4.7: SHAP design showing the abstract objects (ball, cylinder, etc.) and simulated ADL objects (cup pouring, key turning, etc.). Participants perform timed movements of the objects from which the assessor can calculate the SHAP score (89).....................57
Figure 4.8: Frequency spectrum of sample wrist motion showing that the majority of information is below 10 Hz. This justifies the use of a filter with a 10 Hz cutoff frequency.............................................................................67
Figure 4.9: Demonstration of the phases of a SHAP task during a single trial. The initial rest phase results from the need to start recording before the participant begins the task. The reach phase typically involves the greatest motion as the hand is positioned at the object. The Grasp phase is typically characterized by the least motion and determined by recordings of the participant only holding the object. The final rest phase begins
XIV


once the object is released and continues beyond the end of the task until the
recording can be manually stopped..................................................68
Figure 4.10: Motion of the wrist during the reach phase (top) and grasp phase (bottom) of each task in the SHAP averaged across participants. Error bars indicate standard error. For each task, a distinctive motion from the rest position to the beginning of the grasp position is visible. During the reach phase, there is a visible coupling between FE and PS. During the grasp phase, many of the tasks have similar motions while the more complex tasks (coins, buttons, pouring) have distinctive profiles. During grips, FE and RU are commonly correlated. Simulated dart throwing (DTM) was included as a comparison..............................................................................72
Figure 22: 2-dimensional radial plot in the FERU plane on the averaged wrist data. Red and green lines represent the first and second principal component. Their lengths represent the percent contribution of each PC...........................................74
Figure 4.12: Two-dimensional radial plots of wrist angle in the three planes during the reach phase (left) and grasp phase (right). Red and green lines represent the first and second principle components, respectively. Lengths indicate the percent contribution to the variance of each PC. Boxed plots show the largest deviation in each of the phases......75
Figure 4.13: Coupling in the three degrees of freedom of the wrist. Error bars indicate one standard deviation from the mean. We found significant coupling in the reach phases of FERU- (DTM), RUPS-, PSFE+/- and in all three degrees of freedom. Greater coupling in the grasp phase was found in all but the FERU+ (Anti-DTM) mode......................76
xv


Figure 4.14: Percentage of SHAP task grasp phases in given wrist orientation (F-flexion, E-
extension, R-radial deviation, U-ulnar deviation, P-pronation, S-supination). During the reach phase, wrist position is present in all combinations with FUP the most common position. A plurality of grasps takes place with the wrist in flexion, ulnar deviation, and pronation or in extension, radial deviation, and supination. This presentation is consistent with the DTM........................................................77
Figure 5.1: The transmetacarpal hand used as the starting point for the DTM wrist flexion device used in this study (99). The fingers were removed and replaced with an adjustable flexion part and the gears replaced with a commercial gear box...........84
Figure 5.2: Design of the DTM wrist from CAD (left), to plastic 3D printed (middle), to a
machined device (right)........................................................84
Figure 5.3: CAD view of the final DTM wrist (left) and exploded view (right) created for this study. Custom parts for this study are colored: Flexion Part (red), Adjustable Adapter Plate (green), Securing Annulus (yellow), and Wedge (teal). With the wedge removed and the adapter plate rotated to 0, this wrist functions as a standard flexion wrist... 85
Figure 5.4: Demonstration of the DTM wrist's ability to mimic the anatomical DTM....86
Figure 5.5: Modified transmetacarpal hand used in this study. Cotton padding and rubber gloves were added to the fingers to improve grip. The wedge and adapter plate of the
DTM wrist can be seen at the bottom
88


Figure 5.6: Process for simulating an amputation in an intact participant. Attach electrodes
and EMG sensors (Top left), apply Coban (top right), apply padding (bottom left), splint with prosthesis (bottom right).......................................................94
Figure 5.7: Process for building a custom myoelectric socket for a participant with an
amputation. Bare residual limb (top left), apply electrodes and EMG sensors (top right), wrap in cotton and fiberglass casting material (bottom Left), apply second layer of fiberglass cast to secure the prosthesis to the inner cast (bottom right)............94
Figure 5.8: Illustration of the hybrid velocity-position control system. Top left: raw EMG, Top right: thresholded EMG, Bottom: hand and wrist positions. Participant activates flexors (red) to flex the wrist, co-contracts for the required time to switch to hand control, activates flexors which now open the hand, moves to the target (gap in top right), and then activates extensors (white) to open the hand. The resultant movements can be seen in the bottom plots..................................................................96
Figure 5.9: Co-contraction display showing the controlled variables (threshold, refractory time, and co-contract time). Participant has held a co-contraction for 6 out of 10 required bins but missed one in the process, probably by dropping below threshold. Each green bin represents 10ms of co-contraction time. Participants could see this display during the training session but not while performing the SHAP.....................96
Figure 5.10: Index of function scores showing significant differences between the No Wrist condition and all other conditions and between the DTM wrist and the FE wrist (black). There was significant difference between the FE and DTM wrists (red). The score for
xvii


the DTM wrist used by people with transradial amputations was not significantly
different than the DTM wrist used by people with simulated amputations (blue). Error bars represent standard error............................................104
Figure 5.11: Comparison of shoulder ab/adduction ROM during the SHAP tasks. No
significant differences were found between the different prosthetic wrist conditions. Error bars represent standard error......................................106
Figure 5.12: Compensatory motions employed by participants during the SFIAP. A wide
variety of compensatory motions were used, including: large abduction (left), shoulder lift (middle), and torso bending (right).................................105
Figure 5.13: Comparison of NASA TLX scores. Only the No Wrist and AMP (DTM) conditions were significantly different. Higher score indicates greater difficulty.......107
XVIII


List of Abbreviations
ADL
BP
COMIRB
DAQ
DASH
DOF
DTM
FE
FSR
IFS
IR
LED
ME
OPUS
PC
PCA
PS
RU
SHAP
SL
Activity of Daily Living Body Powered
Colorado Multiple Institution Review Board Data Acquisition Board
Disabilities of The Arm, Shoulder and Hand Outcome Measure
Degree Of Freedom
Dart Thrower's Motion
Wrist Flexion-Extension
Flexion Sensitive Resistor
Index Of Function Score
Infrared
Light Emitting Diode Myoelectric
Orthotics And Prosthetics User Survey Principal Component Principal Component Analysis Wrist Pronation-Supination Wrist Radial-Ulnar Deviation Southampton Hand Assessment Procedure Scaffoid-Lunate Joint
xi x


Chapter I
Motivation and Research Goals
There are an estimated 41,000 Americans currently living with major upper limb loss (proximal to the hand) and this number is expected to more than double in the next 30 years (1). Over three-fourths of upper limb amputations are the result of traumatic injury and the majority of people with upper limb amputations are under the age of 65 (1,2). This would suggest an active and adaptable population, but according to the World Health Organization's International Classification of Functioning and Health (WHO-ICF), this population often feels severely limited in their daily lives and report difficulty participating in social activities, finding fulfilling employment, and living independently (3). This is consistent with older studies such as the 1998 Harris survey which found that people with disabilities have a difficult time reintegrating into the workforce and with social interactions after an injury (4). It is also reflected in US employment rates: recent census data show that a smaller percentage of people with disabilities are in the work force than people without disabilities (5-7). Furthermore, as many as 69% of people with upper limb amputations become unemployed or have to change their career after an injury (5-8).
A prosthetic arm enables people with upper limb amputations to be better able to perform activities of daily living (ADL), such as personal hygiene and food preparation, and gain employment in positions that require specific motions (2). Body powered prosthetics that use a cable to actuate a hook or simple hand are fairly limited, but modern myoelectric powered prosthetics can pick up muscle signals from the residual limb and translate that into multiple motions (9-11). Many of these new devices feature multiple degrees of
1


freedom (DOF) in the hand and wrist in an attempt to mimic the anatomical arm (12). However, the control systems for these devices lack the ability to translate the myoelectric signals into complex movements. This lack of control means that prosthetic devices often do not meet their user's needs and is part of the reason that about a third of people with upper limb amputations reject their prosthesis (13).
A great deal of prosthetics research is focused on solving this limitation by increasing the level of control by adding more surface electrodes, implanting electrodes into individual muscles, or designing control schemes to translate patterns of signals into specific motions (14-18). These solutions are promising, but have not yet had great commercial success. The research presented in this dissertation instead focused on simplifying the mechanical properties of the prosthetic wrist with the goal of improving functionality while avoiding complicating the control system. This was done by determining which DOF are most important for performing ADLs and including those while removing the others.
This research was inspired by the work done by Palmer who, in 1985, showed that the motion of the wrist is offset from the anatomical flexion-extension plane in what is known as the Dart Thrower's Motion (DTM) (19). More recently, Wolfe and Crisco found that the DTM is an important DOF to preserve wrist function when surgically fusing bones of the wrist (20,21). The specific goal of this research was to determine whether a simple prosthetic wrist using the DTM could accomplish most of the function of an anatomical wrist.
2


1.1 Summary of Goals
The purpose of this research was to test the hypothesis that people with upper limb amputations face particular difficulty with ADL tasks that require wrist mobility due to the limited degrees of freedom found in most existing prosthetic wrists; and that limitation could be alleviated by a prosthetic wrist that incorporates the DTM controlled with a simple myoelectric system. To this end, this dissertation tested three specific aims:
1.1.1 Specific Aim 1: Identify what ADL tasks people with amputations find challenging via online survey and determine what improvements they are interested in seeing in their devices.
It is most important to identify the needs of a population before designing any new device for them. Performing human factors testing with people with upper limb amputations and integrating their input into the design from the beginning improves the chances of making a successful device. Several studies have surveyed people about their relationship with their prosthesis and found that they often have difficulty performing ADL, especially those that involve tool use (8,22,23). However, none of these surveys specifically inquired about tasks that utilize the DTM. Twenty-three people completed a novel, online survey developed for this study to determine the levels of difficulty, satisfaction, and importance of performing DTM and non-DTM tasks with their prosthesis. The survey was based on existing surveys (24,25) and supported by a team of experts in the field of human factors testing and rehabilitation. Afive-person subset of the people surveyed were also interviewed to clarify their responses and add additional feedback. The survey and
3


interviews indicated that people were less satisfied performing DTM tasks with their prosthesis than non-DTM tasks and that people were interested in having a more functional wrist in their device. This motivated the precise measurement of the DTM and development of a prosthetic wrist that incorporated the DTM.
1.1.2 Specific Aim 2: Measurement of the Dart Thrower's Motion in the Context of a Prosthesis-Specific Functional Test
The goal of creating a simple prosthetic wrist was informed by the results of specific aim 1. In order to design a wrist that incorporates the DTM it was necessary to first precisely identify the angle of the DTM. Previous studies on the DTM have described a range of angles rotated from pure flexion-extension (20,26,27). Many of these studies have limited application to prosthetic design because they used cadaver arms, or examined full range of motion, or measured non-standardized motions. To create a clinically relevant measure of the DTM, fifteen intact participants performed the Southampton Hand Assessment Protocol (SHAP) while wearing goniometers to record the angle of the wrist (28). The SHAP is the current standard of care measurement of prosthetic function and involves moving several abstract objects (e.g. ball, cylinder, tab) and simulated ADL objects (e.g. turning a key, picking up coins, pouring water from a cup). The results indicated that a prosthetic wrist could mimic the DTM by combining flexion-extension and radial-ulnar deviation at an angle of 22 degrees during the grasp phase.
4


1.1.3 Specific Aim 3: Testing a novel prosthetic wrist that incorporates the Dart Thrower's
Motion
A novel prosthetic wrist incorporating the results from specific aim 2 was designed and tested. The wrist was based on an existing wrist flexion unit rotated to the DTM angle. The prosthesis was controlled by a simple, modified position-control myoelectric system. The goal of this study was to determine whether the DTM wrist would perform better than a flexion wrist and would still be as easy or easier, to use than a prosthesis without a wrist. To test this hypothesis, a group of 10 able-bodied participants performed the SHAP using this device with a splint to simulate an amputation. Each person in this group completed the SHAP using the DTM wrist, a flexion wrist, and a fixed wrist (no-wrist). The results from this group were compared to three able-bodied people performing the SHAP with their intact limbs and to five people with upper limb amputations using a prosthesis with the DTM wrist. Improvement in function was measured by the SHAP Index of Function Score (related to time-to-completion) and shoulder compensatory motion (shoulder angle measured with goniometers as in aim 2). Ease of use was measured by the NASA Task Load Index (NASA-TLX), a subjective questionnaire filled out at the end of each session.
The results of this study indicated that the DTM wrist performed better than the flexion wrist on the SHAP although both active wrist conditions were slower than the no-wrist condition. The shoulder data measuring compensatory motions did not indicate a reduction in ab/adduction compensatory motion but a range of other unmeasured compensatory motions were observed during the test. The NASA-TLX indicated that neither the DTM wrist nor the flexion wrist were more difficult to control than a device without a wrist. This is a
5


strong indication that the DTM wrist improves prosthetic wrist function without being too
difficult to control.
6


Chapter II
Background
This chapter contains the background necessary to understand the challenges facing anyone who wants to design a better prosthetic device for people with upper limb amputations (ULA). This includes: the physiology of the intact and amputated upper limb, specifically focusing on the function of the anatomical wrist; a discussion of electromyogram signals from the muscles and how they can be used to control powered prosthetic devices; and an overview of existing upper limb prosthetic devices, particularly commercial and research grade powered wrist devices. This research was focused on the mechanics of the anatomical wrist and design of a new prosthetic wrist. Therefore, a majority of this space is devoted to describing those systems.
2.1 Physiology of the Upper Limb
The human arm evolved over millions of years to perform both dexterous and gross movements using 36 muscles in the hand and forearm, 27 bones, and 18 joints, resulting in 27 possible motions, or degrees of freedom (DOF), in the hand and wrist and an additional 7 DOF in the arm (11). The hand alone contains 22 DOF while the wrist has three functional DOF arising from movement of the eight metacarpal bones and the relative motion of the radius and ulna. These motions are primarily actuated by muscles in the proximal forearm. They allow the arm to perform a wide array of motions from large, course movements, such as swinging and grasping, to precise positioning and fine motor control (29).
7


The upper limb consists of the shoulder joint, upper arm, elbow, forearm (consisting of
the radius and ulna bones), wrist (made up of the metacarpal bones), and the hand. The role of the proximal arm is primarily to produce course movements for positioning and power, while the distal components of the arm are typically responsible for fine control and precision.
It has been understood for over a century that a great amount of brain power is devoted to controlling the arm and hand (30). The function of the upper limb is the result of mental planning and neural signaling, proprioceptive and sensory feedback loops, and the biomechanics of the muscles, tendons, and bones (31). These actions must be timed and coordinated precisely to interact with the world. Penfield famously illustrated the outsized proportion of the motor cortex devoted to the hand in his homunculus drawing (Figure 2.1) (32). The evolution of the hand and motor cortex in early hominids likely developed simultaneously and set humans apart from other primates (33).
8


Figure 2.1: Homunculus drawing by Penfield showing that neurons directing the control of the hand dominate the motor cortex (34).
2.1.1 The Wrist
The wrist is used to alter the angle of attack of the hand to allow efficient grasping (35). Without afunctional wrist, people would need to make large compensatory motions with the upper arm to position the hand and perform ADLs. There are redundancies in the muscles that activate the wrist but the primary activators are described below.
The first two DOF of the wrist, flexion-extension (FE) and radial-ulnar deviation (RU), are functions of the movement of the carpal bones relative to the radius and ulna (Figure 2.2). These are activated by several muscles in the forearm. Specifically: the flexors carpi radial is and ulnaris (flexion), extensors carpi radial is longus and brevis (extension), flexor carpi radialis (radial deviation), and flexor carpi ulnaris (ulnar deviation). The third DOF of the wrist, rotation or pro nation/su pi nation (PS), takes place in the forearm as result of the
9


radius and ulna twisting about each other. This motion is activated by the pronatorsteres
and quadratus (pronation) and supinator (supination) (29).
Figure 2.2: Bones of the hand and wrist1.
The functional motions of the wrist allow up to 135 degrees of FE, 40 degrees of RU, and 130 degrees of PS (Figure 2.3) (14). Most tasks, however, only utilize about 40 degrees FE and lessthan 30 degrees RU (19). .Additional flexibility and compliance in the wrist is provided by movement of the carpal bones against each other.
1 www.anatomy-diagram.info
10


7 Kxtensic
80-90 Flexion
30c- 45 Ulnar Deviation (Adduction)
Flexion Extension
Radial-Ulnar Deviation
Rotation
Figure 2.3: Range of motion of the anatomical wrist ((14), modified).
Wrist motion is important for performing many activities of daily living (ADLs). Palmer measured the range of motion (ROM) of wrist FE and RU of normal participants performing a series of ADLs ranging from personal hygiene and food preparation to tool use and secretarial work (19). Fie found that many ADLs, especially tasks that involve tools (e.g. carpentry, culinary, and some personal hygiene) require a significant amount of wrist FE and RU. The wrist is also valuable for positioning the hand at the midline for ADL tasks such as feeding, cooking, and grooming.
2.1.2 The Dart Thrower's Motion
Palmer's study was the first to show that these ADLs often use a combination of wrist FE and RU in in an oblique plane rotated between 20 and 40 degrees from pure FE (20). This motion has become known as the Dart Thrower's Motion (DTM). As the name suggests, the DTM is the motion of the hand and wrist travel from an extended and radially-deviated position to a flexed and ulnar-deviated position when a dart is thrown (Figure 2.4). It is estimated that more than half of ADLs make use of this motion (36).
11


Figure 2.4: An example of The Dart Thrower's Motion2.
The combination of these two DOF allows for efficient positioning of the hand for prehension. In intact wrists, the DTM is a motion guided by the distal surface of the scaphoid bone and minimizes scaphoid-lunate and spheroid-lunate motion (Figure 2.2) (37). This motion allows for the greatest range of wrist circumduction while minimizing required muscle activity and maximizing muscle force output (27,38). The DTM is particularly important for the power stroke of tool use and throwing. Studies have suggested that the development of the bone and ligament structures involved in this motion were one of the major evolutionary advantages that set humans apart from other primates (33,39,40).
Several studies have attempted to identify the specific angle of rotation for the DTM (19,27,41,42). These methods have measured passive and active range of motion, live participants and cadavers, and used goniometers, MRI, and CT scans to try to pin down the DTM angle. One of the best studies was done by Crisco in 2011 measuring range of motion on cadaver arms mounted to a custom jig (Figure 2.5). This showed the passive range of motion of the wrist to be an envelope rotated about 30 degrees from pure FE.
2 www.nicedarts.com
12


Extension
Figure 2.5: Radial plot of the average maximum passive range of motion of 6 cadaver wrists as described by Crisco. The DTM plane is visible as a plane rotated 20-40 degrees from the plane of pure flexion-extension (20).
The concept of the DTM is currently in use in the medical field among hand surgeons. Some diseases of the wrist, including severe arthritis and tumors, require surgeons to fuse some or all of the carpal bones. When these procedures are performed, surgeons often are forced to choose a single DOF that will minimize motion while preserving function. Scott Wolfe, a physician and one of the leading researchers on the DTM, advocates that when wrist mobility must be limited, the DTM should be preserved to retain the majority of functionality (21,23).
The designer of a prosthetic wrist faces a similar choice as the surgeons: the DOF in a prosthetic wrist are limited by weight, power, cost, and ease of use. Therefore, it is prudent to find a motion that maximizes functionality while limiting complexity. The DTM is an excellent candidate for such a motion.
13


2.2 The Amputated Upper Limb
The loss of a major part of the upper limb can have profound social and economic effects on a person. Studies have shown that people with upper limb amputations (ULA) often feel severely limited in physical and social activities and in finding and keeping jobs and represent a smaller portion of the work force than people without disabilities (3,5-7). They also tend to have a difficult time reintegrating into the workforce and find social interactions challenging after their injury (4). People can lose limbs for many reasons but, ultimately, they are left with a residual limb and a need to figure out how to interact with the world in a novel way.
2.2.1 Etiology
One of the best estimates of the number people with amputations found that there were 41,000 people in America with major ULA (proximal to hand) in 2008 (1). That number is expected to more than double by 2050. Trauma accounts for between 62 and 90% of all major ULA (1,22,43,44). Most of these traumatic amputations are the result of farm and industrial accidents. Operation Iraqi Freedom/Operation Enduring Freedom brought limb amputations to the forefront of the public perception although according to the most recent pentagon report, these conflicts only account for 737 American servicemembers suffering ULA (22).
Of acquired ULA, malignancy may account for up to 24% and dysvascular diseases account for 3% (43). Congenital limb deficiency, which can be caused by many factors, occurs in about 15 out of every 100,000 live births. While the risk of acquired ULA increases
14


with age, the majority of people with ULA are between the age of 20 and 40 and generally
considered active (44). The most common major ULA and congenital limb deficiency is trans-radial: between the wrist and elbow.
2.2.2 Anatomy and Surgery of the Upper Limb
Regardless of the cause of the amputation, the patient is left with a less-functional residual limb. This can make interacting with the world challenging, but it is by no means impossible. Indeed, many people with amputations are able to be fully active and do not need a prosthesis. In many cases, the residual limb will retain functional musculature, nerves, and vasculature. With care and training, the residual limb can remain useful for performing ADLs by itself or by controlling a prosthesis.
Surgical procedures for amputation vary by patient and underlying cause. Surgeons may perform a disarticulation at a joint or osteotomy across the bone. In either case they usually attempt to preserve as much length in the residual limb as possible (43,45). This maximizes the torque that can be applied by the limb and allows for better prosthesis attachment. Efforts may also be made to preserve prominent bony features or long bone length in order to accommodate a socket for a prosthesis. Other considerations include techniques for reconnecting the musculature and vasculature to keep the limb healthy and functional. Neurons must also be managed by cutting them short or embedding them in other muscles to avoid painful neuromas or phantom limb sensations.
There are other surgical techniques that seek to accommodate a prosthesis after amputation. Targeted muscle re-enervation can attach nerves to new muscles so they can
15


be used for controlling a prosthesis via electromyography (described below) (45,46). Tunnel
cineplasty creates an external loop of muscle that can be attached to a cable that activates a prosthesis (14). Osseointegration is a technique where a metal rod is attached to the bone in the residual limb and extends through the skin where it can be attached to a prosthesis (47). Each of these techniques attempts to improve the outcome of the amputation surgery and provide the patient with options for prostheses.
2.2.3 Considerations for People with Upper Limb Amputations
Without the intrinsic, neuro-muscular, system of the body, people with ULA lack the functionality to perform many tasks (22,23). People are often able to overcome this functional deficiency through compensatory motions such as major shoulder abduction or torso bending (48,49). However, these motions can lead to overuse and repetitive motion injuries, poor body mechanics, and social challenges (50-52).
There is limited information about the social and economic consequences of acquiring a ULA. One of the few studies to specifically address this population found that as many as 69% of people with ULAs become unemployed or have to change work places after their injury (8). More generally, people with disabilities often have a difficult time reintegrating into the workforce and interacting in social situations (4,7). This is reflected in US employment rates in the most recently available census data which shows that a smaller percentage of people with disabilities are in the work force than people without disabilities (5-7). This can have a severe negative impact on quality of life (53).
16


This is not to say that people with ULAs are unable to live fulfilling lives. Prosthetic
devices have come a long way in the last century and are able to provide a range of functions to people. Even without prosthetic devices, many people with ULAs are capable of extraordinary achievements. Some are even world champion rock climbers3.
2.3 Current Prosthetics
For many people with an ULA, a prosthesis will provide much of the function of an anatomical limb. However, designing functional and aesthetically pleasing prostheses for people with ULA has historically proven to be a challenge (54). An ideal prosthetic arm would mimic all the functions of an intact human arm, but each additional feature, point of control, DOF, and increase in force, adds to the volume, mass, and complexity of the device (11). The design of current prosthetic terminal devices is driven by minimizing these factors and limiting how difficult it is to control.
From a user perspective, a device must be simple to use, comfortable, and functional in order for it to be accepted (55). Current devices fail these criteria, as evidenced by the persisting high rejection rates among persons with unilateral amputations (38%, 30%), particularly among people with wrist disarticulation amputations (94%, 80%)(22,56).
Overall, fewer than a third of people with ULA report being satisfied with their prosthesis (8). Designing an overly complex device will not alleviate the problem of device
3 http://www.paradoxsports.org/paradox-ambassador-spotlight-maureen-beck/
17


abandonment and an overly simple device will not help people with amputations be more
social or more able to perform ADLs.
2.3.1 Types of Prosthetics
The commercially available prosthetic device options can be divided into three groups: "cosmetic" or "passive", "body-powered" and "powered devices". Cosmetic devices (e.g. cosmetic hand from RSL Steeper (57)) are typically made of a colored silicone or PVC glove over a moulded foam hand. While these devices provide minimal functionality, they provide a high level of social acceptance which is important for many people with amputations (58).
Body powered devices typically utilize a Bowden cable running from the device to an intact part of the body which directly or indirectly articulates a single DOF in the device. These devices are highly functional, well-liked, and extensively used by people with amputations. They have the advantage of being durable, inexpensive, easy to use, and give a modicum of sensory feedback. However, there have only been a few recent advancements in body powered devices. These include the 3d printed e-NABLE hands (59) and the sport-specific terminal devices made by TRS (60). Otherwise, the basic design of body powered devices has hardly changed since the 1950s and these devices are essentially at the limit of their potential (11,61).
The state of the art in upper limb prosthetics is powered myoelectric devices. These devices incorporate surface electromyographic (EMG) electrodes to record electric signals from contracting muscles in the residual limb and translate them into motor movement. Myoelectric devices have the potential to provide an intuitive control scheme for users with
18


minimal delays between thought and movement (17). However, surface electrodes lack
specificity i.e. EMG electrodes placed on the volar forearm pick up signals from all of the flexor muscles while electrodes on the dorsal forearm pick up all signals from extensor muscles rather than recording from individual extensor or flexor muscles (10). This typically limits these devices to being controlled by only one or two one agonist/antagonist pairs. With only one or two control sites, switching mechanisms are necessary to control multiple degrees of freedom (17). Adding DOF can quickly make the device too difficult to use.
This lack of control inputs constitutes a major challenge to full functionality for myoelectric prosthetics. Although the mechanics of prosthetic devices can mimic many of the motions of the human hand, there is not yet a way to effectively control them (35). The low number of control sites in powered myoelectric devices necessitates that only the most valuable DOF be incorporated in prosthetic devices. Commercially, this means that a prosthesis has at most a gripping device (e.g. hand or split hook) and a wrist unit (almost always a rotator) (62). It is not clear whether rotation is the best DOF to include (50).
The myoelectric system will be discussed in greater depth in section 2.4.
2.3.2 Prosthetic Wrists
Articulated wrist devices improve positioning and reduce compensatory movements in many ADLtasks, such as food preparation and personal hygiene (50). Users score higher on functional tests when flexion or deviation capabilities are added to the wrist (23,36,63,64). In one case, a majority of users who were fitted with a passive, locking, F-E wrist (Multi-Flex Wrist) report that having F-E in their device improved the number of activities they were
19


able to perform with their device and made activities feel more natural (36,65). However,
commercially available prosthetic devices do not provide full functionality and adding any additional DOF necessarily complicates controlling a device (66). This added complexity may be acceptable if including a functional wrist in a prosthesis improves the user experience and increases functionality (67).
There are only a few commercially available options for wrist prosthetics. Most commonly, terminal devices are simply fixed to the socket or attached via a constant-friction device. These hold the terminal device in a fixed position with a set screw or spacer washer. A quick-disconnect system allows users to quickly change terminal devices and set them at fixed PS angles (54) (Figure 2.6). While some work has been done to determine the optimal fixed angle of the wrist, most commercial devices are simply set by the user to a neutral or convenient angle (68).
There exist prosthetic wrists which utilize one-, two- or three- DOF. These include devices with passive motions such as the 5-Function N-Abler wrists (Texas Assistive Devices, Brazoria, TX), the Robo-Wrist (Medical Bionics Inc. Spruce Grove, Alberta, Canada), and the TRS Flexible wrist (Therapeutic Recreational Systems, Boulder, CO). Each of these devices can be manually positioned to achieve flexion but this requires additional body movements and a series of locks or switches which most users find difficult to operate (36).
20


Figure 2.6: Passive wrists. Terminal devices may be attached to these devices and passively positioned to achieve flexion. From left to right: Flosmer constant friction wrist and Flosmer flexion wrist (Flosmer, Campbell, Cty, Texas Assistive Devices N-Abler wrist, Medical Bionics Robo-Wrist, and TRS passive flexible wrist.
A few powered wrist devices also exist. The most common of these are the wrist rotators from Otto Bock and Motion Control (Motion Control Inc., Salt Lake City, UT), and the VASI Children's Electric Wrist Rotator (Liberating Technologies Inc, Flolliston, MA)
(Figure 2.7). Motion Control makes a powered wrist rotator with a passively positioned FE in their powered hand and prehensor and will be releasing a powered FE device this year.
Figure 2.7: Examples of powered wrists. The left two devices (Otto Bock wrist rotator, VASI children's electric wrist rotator) only provide rotation. The right two devices (Centri Fland, Shanghai hand) combine rotation and flexion-extension.
The Centri Fland (CentriAB, Sollentuna, Sweden) uses a 4-bar linkage to combine finger flexion with wrist extension. The Shanghai Fland (Shanghai Kesheng Prostheses Co. Ltd, Shanghai, P.R. China) has powered flexion-extension but does so in the anatomical F-E plane and cannot be rotated to the DTM plane. The Defense Advanced Research Projects Agency
21


(DARPA) RP2009 APL Modular Prosthetic Limb (MPL) Arm and the DARPA RP2007 DEKA Arm (Figure 2.8) have multi-DOF wrists that included FE but cannot specifically mimic the DTM.
Figure 2.8: The Defense Advanced Research Projects Agency (DARPA) RP2009 APL Modular Prosthetic Limb (MPL) Arm developed as part of the DARPA RP2009 initiative
includes a multi-DOF wrist.
2.3.3 Limitations of Current Prosthetic Devices
In 1985, Childress wrote "Adequate replacement of the human hand and arm is one of the most difficult problems facing medical technology" (55). This sentiment is no less true today (69). Despite major advancements in machining, processing, and battery life, today's top of the line myoelectric devices still fall short of adequately replacing an anatomical limb.
One of the limiting factors for people with ULA is the lack of simple and effective prosthetic devices (1). Too often, people must choose between overly simple, body-powered devices or overly complicated powered myoelectric devices (70). While myoelectric devices hold the promise of emulating the intact human hand and wrist, they provide too few inputs to effectively control a complete limb.
Many prosthetic arm users have difficulty performing ADLs with their device. Opening bottles, using tools, and personal hygiene tasks are often listed as some of the most
22


challenging tasks to perform with a prosthesis (8). Farmers report their primary complaint is
the inability to use their device for work [71,72). These challenges often cause people with unilateral ULA to resign themselves to only using their prosthesis for bimanual tasks (73).
It is not only the high-end devices that are failing to meet user's demands. Rates of rejection for body powered hooks have been measured to be as high as 85% (13) and passive hands have a rejection rate around 50% (74). This is often due to factors other than functionality; Cost, appearance, socket comfort, durability, and weight are major factors that can lead people to abandon their devices (13,58,74). Research is being done to improve these physical factors and the more functionality a device has, the more likely users will tolerate using it (11).
2.4 Myoelectric Control
Myoelectric systems allow users to control their devices with (EMG) signals from muscles in the residual limb (Figure 2.9). That is, when the contractile signal from the brain reaches the muscle, the electric potential around that muscle changes. That change in potential can be detected via EMG electrodes placed on the skin. This signal can then be processed in hardware or software in order to control the device (11). The muscles are used rather than trying to record directly from the neurons because the muscles produce a much higher voltage and the signal can be detected through the skin.
23


Parts of a below-elbow myoelectric prosthesis
electrodes
control unit and battery pack
friction wrist
socket
2012 Encyclopedia Britan nica, Inc.
electric hand
Figure 2.9: Diagram of myoelectric control for a powered prosthesis. The neural signal from the brain innervates the muscle in the residual limb, changingthe surrounding electric potential and causing it to contract. This change is detected by electro myogram electrodes on the surface of the arm, electronically amplified and processed, and sent as a control signal to a motor which moves the prosthesis4.
Processing the EMG signal involves first amplifying the signal and then filtering it (75). The signal is then full-wave rectified and averaged. The majority of this processing is often done in hardware in the electrodes. This allows the signal to be recorded as a voltage by a data acquisition unit attached to a computer or built into a prosthesis (75).
The processed EMG signal can be used to control one or more motors. Most myoelectric systems use the signal to directly control the motors by equating amplitude of the signal to the position of the prosthesis in space (position control) or by relating amplitude of the signal to the velocity of the motor (velocity control) (10). It is also possible to create a system where combinations of EMG signals relate to specific postures (posture control) or algorithms detect features within the EMG signal (pattern recognition) so that a 4
4 www.Britannica.com
24


variety of positions can be achieved without requiring a set of electrodes for each DOF
(17,76).
25


Chapter III
User Surveys Suggest Incorporating the DTM into a Prosthetic Wrist Would Be Useful
Around thirty percent of people with upper limb amputations choose not to use a prosthetic device because these devices do not meet their needs (22,56). In order to reduce this rejection rate and improve user satisfaction, it is important to understand what functionality users are looking for in a device and why they feel unsatisfied (58). "The user has the most important perspective, and various goals of the user will determine what the most appropriate device would be to reach those goals" (4, p84). Without understanding the user perspective, any research risks making the mistake of creating a solution for a problem that does not exist.
An online survey was developed to specifically address whether people's dissatisfaction performing activities of daily living was related to limited wrist function in their prosthesis. Twenty-three people with trans-radial amputations completed the survey. Overall, respondents reported significantly lower satisfaction performing activities of daily living in tasks that utilize the Dart Thrower's Motion. A majority of respondents also reported that they wanted a more functional wrist device in their prosthesis. A subset of five people who completed the survey volunteered for a follow-up interview to discuss their responses in greater depth. They indicated the lack of ability to use tools and perform many activities of daily living as a major limitation in their device. They also indicated several physical characteristics (weight, cost, durability, and socket discomfort) as needing improvement.
26


The results from this study motivate further investigation into the mechanics of the
wrist with the goal of improving wrist function.
3.1 Methods: Development
3.1.1 Human Factors Testing
Human factors testing (survey questions and in-person interviews) have long been established as an effective means to determine the best product designs for users (78). One of the earliest and largest surveys of the needs of people with upper-limb amputations was conducted by Atkins et al. in 1996 (67). This study included responses from 2,477 one-page surveys and 1,575 responses to follow-up seven-page surveys. They found that users of both body-powered and electric devices wanted more functional wrists. They also found that most people used body-powered devices instead of myoelectric ones for more rugged tasks such as manual labor, construction, and farming.
Since Atkins' work, surveys have attempted to obtain a general idea of what challenges people face using their devices. However, in 2011, Ritchie analyzed fifteen studies that surveyed people with upper limb amputations and reported a great deal of variation in how studies were conducted and what questions were asked (58). Furthermore, the most commonly used surveys (e.g. OPUS & DASH) only ask about difficulty and do not address whether tasks are important or if the user is satisfied performing those tasks (24,25,79). Despite nearly twenty years of trying to identify the users' needs, the prosthetics industry
27


has been slow to incorporate these findings and the few powered wrist devices on the market have not produced satisfactory results (12).
3.1.2 Survey and Interview Questions
This study developed and administered an online survey and interview questions to determine user experiences performing ADLs. There are three important characteristics to successfully completing a task with a prosthesis: How satisfied the user is with the device's performance (satisfaction), how important is it to be able to complete the task (importance); and, how easy is it to complete the task (difficulty). Each of these metrics is part of the overall user experience.
It is necessary to select tasks that accurately represent the kinds of ADLs that a user encounters during the day. One of the challenges to creating a survey about ADL satisfaction and difficulty is people's ability to compensate for limitations of their device by making compensatory motions (low difficulty, low satisfaction). These motions allow the user to accomplish the task but they may lead to overuse injuries and the motions are often socially awkward (48,49). Users may also simply avoid performing certain difficult tasks by using their intact limb or asking others for help (high difficulty, low importance) (71).
In order to investigate the typical user experience and mitigate the effects of compensatory strategies, this survey included typical ADLtasks, tasks that are difficult to perform with compensatory motions, and two-handed tasks where use of a prosthesis is required. The task list was influenced by two validated surveys of upper limb function: Disabilities of the Arm, Shoulder, and Hand survey (DASH) and Orthotics and Prosthetics
28


Survey Upper Extremity Functional Status (OPUS-UEFS) (25,79). However, neither of these
surveys specifically asked about several tasks that incorporated the DTM or about satisfaction or importance. Therefore, a new list of tasks was created after discussion with a panel that included experts in prosthetics, rehabilitation, and assistive devices (Table 3.1).
Users responded to the survey questions using a 5-point Likert-like scale (e.g. from "1 -not satisfied at all" to "5 very satisfied"). The order of the tasks was randomized to minimize respondent fatigue (80). In addition, participants were also asked to rate their overall opinion of their device from 1 (do not like at all) to 5 (like a lot).
29


Table 3.1: List of activities of daily living (ADL) used in the online survey. Participants were asked to rate their level of satisfaction and difficulty performing these 30 tasks, and how important it is to them to be able to complete them on a 1-5 Likert-like scale. Tasks include DTM and non-DTM ADL activities. Tasks were presented in random order to reduce
the influence of respondent fatigue.
DTM tasks that are hard to compensate for Non-DTM tasks
Turn a key Do/undo trouser buttons
Tie shoes Push open a heavy door
Do heavy household chores
2-handed DTM tasks Change a light bulb
Use fork and knife Carry a heavy object
Drain pasta in strainer Use mouse to navigate internet
Put on a pullover sweater
DTM tasks that are easily compensated for Easy Recreational activities
Use a knife to cut food Dial a cell phone
Write Name Take $1 from wallet
Place an object on a shelf above your head Type sentence
Garden or do yard work Fold laundry
Difficult Recreational activities Transport hay
Recreational activities in which you move your arm freely Sweep floor
Open bottle cap Put on socks/pants
Open Jamjar
Close eye drop/lipstick case
Take pill from case
Do/undo trouser zipper
30


Table 3.2: List of proposed additions to the prosthesis. Participants were asked to rate each idea by how useful they thought it would be to have in their device and how difficult they thought it would be to use. Ratings were based on a 5-point Likert-like scale from 1 (not at all useful/difficult) to 5 (very useful/difficult).
Proposed Improvement:
Individual control of fmgers hook components (e.g. open an envelope/snap your fingers) Wrist with powered side to side motion (R/U deviation)
Rotating thumb to allow pinch motion (Ab/adduction of thumb) Wrist with powered combination of F/E and R/U
Sensory feedback from fingers/hook Wrist with powered combination of F/E and R/U and rotation
Faster finger/hook motion Fully functional wrist with rotation F/E and R/U
Stronger grip Passive wrist that can be positioned in anatomical positions
Wrist with only powered rotation (pro/supination) Faster elbow
Wrist with only up/down motion (flexion/extension) Elbow rotation (humeral rotation)
Wrist with rotation and F/E
3.1.3 Survey Validation
Validating a novel survey with a small population is a challenge. Nevertheless, the standard method for validation was followed as closely as possible (81-83). "Face validity" was first established by assembling a panel of experts to assess whether the list of tasks could accurately answer the research questions. This panel included Dr. Bodine (assistive technology), Dr. Weir (prosthetics), and Dr. Sullivan (physical medicine and rehabilitation) who found the questions to be appropriate. All questions were individually reviewed by the
31


subject matter experts to make sure none were leading, confusing, or included other common errors.
Table 3.3: Cronbach's alpha for Satisfaction, Importance, and Difficulty calculated on total responses grouped by DTM and non-DTM. The value of >0.90 indicates that questions in these groups were highly correlated and therefore gave consistent responses.
Cronbach's Alpha
Satisfaction Importance Difficulty
DTM 0.96902 0.91263 0.95148
non-DTM 0.97691 0.91894 0.96205
Due to the small population available for this convenience sample, an iterative pilot testing phase to determine if questions were relevant was not feasible. However, grouping the questions into DTM and non-DTM and assessing internal consistency with Cronbach's Alpha revealed that questions within each group were highly correlated (Table 3.3). This provided a good indication that the questions accurately identified the participants' views of DTM and non-DTM tasks.
3.1.4 Interview
Follow-up interviews were conducted to allow participants to give more in-depth answers to survey questions and to allow the researchers to ask about any apparent discrepancies in the survey. During the interviews, participants were asked to describe situations where they had difficulty performing tasks with their prosthesis or where their prosthesis excelled, to clarify their responses to the survey, and to describe what types of
32


improvements they would like to see (Table 3.4). Additionally, participants were asked
whether they would be willing to use a more complex device if it had functionality.
The use of open-ended questions gave participants the freedom to volunteer information about their devices. This was particularly important in determining in what ways participants felt limited by their devices and identifying new ways their devices could be improved.
33


Table 3.4: List of questions asked of interview participants.
General questions Specific to survey & DTM
Describe the most difficult situation youve ever encountered when using your device and how you managed. Several people in the survey, (including you), rated tasks as difficult and important, but also said they were satisfied with their device. Can you explain why?
Generally speaking, are there particular tasks, or groups of tasks, you have difficulty performing with your device? Many people said that wrist rotation was not important, yet it is usually included in devices. What do you think?
Are there ever any situations, or particular tasks or activities, when you simply choose not to use your device? The proposed improvement should specifically aid users with tool use (e.g. swinging a hammer). Do you have trouble using tools with your device?
Describe an aspect of your device you like. Dislike? The proposed device would have an additional degree of freedom (angle it could move). This will require a more
Describe changes you would like to see in your device. complicated control system. Would you be willing to use a more complicated device if it provided you with greater functionality? How much better would a device have to be in order to justify the added complexity of the control system? What would it have to do for you?
Would you prefer a high- or low-tech solution to this problem? Open Discussion:
How much would you be willing to pay for this improvement? Is there anything else you would like to add or that you have thought about in regard to your device, or die proposed new device?
34


3.1.5 Power Analysis
An a priori power analysis indicated that twenty-three participants were sufficient to detect a 10% difference between responses with 97% confidence (84). This was calculated by the equation
where n is the required number of participants, o is the standard deviation of the population (0.1), 6 is the minimum detectable difference (10%), P is the desired p-value (0.05) and v is the number of DOF (22, # participants -1) (84). A 10% difference was selected as a best-guess at clinical relevance. This calculation was consistent with the recommendation in Rubin that suggests that a study of 23 participants is sufficient for representative results (78).
35


1
Population/Power curve for a=0.05, v=22, a = 0.1, y = 0.1
%> 0.75 £ 0.7
0.65
0.6
0.55
0.95
0.9
0.85
P = 0.97293
0.5
10 15 20 25 30 35
Population size (n)
Figure 3.1: Power analysis indicating that 21 participants were sufficient to detect a 10% difference in responses with 97% confidence.
3.2 Methods: Experiment
A convenience sample of twenty-three adults (18-85 years of age) with trans-radial amputations were recruited. Participants completed the online survey described above using Survey Monkey or Google Forms. Demographic information (name, gender, date of birth, profession, type of amputation, and time since amputation (if applicable)) was also collected. Participants were invited to return for a follow-up interview. Those interviews were conducted and recorded over Skype or in person at The Assistive Technology Partners Product Testing Laboratory and recorded with a camera and microphone. Inclusion/exclusion criteria was selected to ensure that the sample could accurately represent the population of people with ULA and that it could address specific limitations of the prosthetic device.
3.2.1 Inclusion/Exclusion Criteria
To be eligible for this study, participants had to meet the following criteria:
36


Be between the ages of 18 and 85
Have a transradial amputation Be non-sedentary (>4 hours of activity per week)
Have experience using at least one prosthetic device
Have normal or corrected vision and hearing
Can read and understand English at a cognitively appropriate level
The following criteria excluded people from taking part in this study:
Age less than 18 or greater than 85
Have an amputation at a location other than transradial
Perform less than four hours of exercise per week
Have no experience using a prosthesis
Be unable to use a computer to answer the survey
There were no exclusions based on race/ethnicity, gender, or socioeconomic status. Participants answered questions to ensure they met the inclusion/exclusion criteria before beginning the survey. Responses were excluded from analysis if the participant indicated that they did not meet the inclusion/exclusion criteria.
People with only trans-radial amputations were surveyed in order to minimize complicating factors. While more proximal amputations also include the wrist and hand, the additional DOF of an articulated elbow, humerus, or shoulder would make it difficult to determine whether the participant's responses were the result of the wrist prosthesis or other factors. Likewise, people with distal amputations would have an anatomical wrist and not face the same limitations as someone with a prosthetic wrist. Modern prostheses are
37


designed to be modular so even though this survey only questioned people with trans-radial
amputations, the results are applicable to all levels of amputation. Improvements for the trans-radial level also apply to more proximal amputations.
A non-sedentary population was selected to reduce the likelihood that participants had physical limitations that would make it difficult to determine whether the survey results applied to issues with the wrist or because of disease or inactivity. This does not necessarily mean that the user is more proactive about their device, only that they make adequate use of their device. This study was primarily interested in improving the lives of people who use, or want to use, their devices for ADLs and so a non-sedentary population was appropriate. Similarly, participants with visual or auditory impairments may have difficulties performing the tasks that are unrelated to limitations to the wrist.
3.2.2 Recruitment
Participants were recruited via email, postings at related organizations, and word of mouth. The following organizations were particularly helpful in recruiting participants:
Paradox Sports (www.paradoxsports.org)
Orthotics and Prosthetics International (www.oandp.com/oandp-l)
National Amputee Golf Association (nagagolf.org)
38


Flyers and emails included a link to the survey that participants could follow without
making direct contact with the research team. Participants who chose to email or call directly were directed to the survey link. The initial page of the survey included information about how the data would be handled and explained that the data would be de-identified and stored on an encrypted, HIPAA compliant, password-protected computer. Participants were not compensated for completing the survey or the interview. This study was conducted with the approval of the Colorado Institutional Review Board (COMIRB #11-0674).
3.2.3 Analysis
The analysis of the survey responses was performed using Matlab (The MathWorks, Inc, Natick, MA). The difference between responses was calculated by a student's t-test on the difference between the average participant's ratings for DTM and non-DTM tasks. T-tests were also performed participants representing common population subgroups as described in chapter 2.
Interview recordings were analyzed for keywords using Morae (Techsmith, Okemos,
Ml) (Table 3.5). Keywords were reported by the percentage of total keyed comments. Comments about particular limitations or desired improvements were also reported. The interview process was valuable for clarifying any confusion that arose from the online survey results.
39


Table 3.5: List of keyed participant comments during interviews. A mark was made each time a participant mentioned each of these topics as well as a comment with further details. The list was generated by observing common topics in each interview.
A (cosmetic) I (finger control)
B (difficulty using tools) K (function over form)
C (cost) L (Myoelectric is good at this)
D (socket issues) N (weight)
E (safety issue) R (ADL issues)
F (electronics issue) U (User wants improvement)
G (durability) Y (BP is good at this)
3.3 Results
Twenty-three adult participants with trans-radial limb deficiency (mean age 46.8
10.6) responded to the survey. Each respondent had experience with powered prostheses, unpowered prostheses, or both (only powered = 5, only unpowered = 9, both = 6, no preference/no regular device = 3). The sample included both participants with congenital (n=3) and acquired (n=20) limb deficiencies. Respondents with acquired limb deficiency had used prosthetics for an average of 18 years (mean 18 17.8). Average age of participants with congenital deficiencies was 28.3 5.8 years. Two of the participants with congenital deficiencies used body powered devices and one used both. Overall, users reported that they found their devices to be somewhat easy to use (mean 3.71 on a scale of 1 = very difficult to 5 = very easy). There was no correlation between responses to satisfaction, importance, and difficulty.
40


Table 3.6: Significance values of one-sample t-test on the difference between the average response to DTM tasks and non-DTM tasks. Highlights indicate significantly lower satisfaction, higher importance, or higher difficulty between DTM and non-DTM tasks (p<0.05). Overall, the subset of users of body powered devices and those with acquired limb deficiencies reported lower satisfaction with DTM tasks than non-DTM tasks. The acquired deficiency subgroup also reported higher difficulty performing DTM tasks than non-DTM tasks. The sum of participants in body powered and myoelectric devices is greater than 15 because some participants used both types of devices.
One-sample t-test vs mean of 0 (p-values)
N A Satisfaction A Importance A Difficulty
All participants 23 0.017 0.399 0.068
Body powered 15 0.024 0.965 0.021
Myoelectric 12 0.086 0.198 0.454
Acquired 20 0.013 0.773 0.026
Congenital 3 0.884 0.269 0.510
The total participant population (n=23) and the subsets describing the most common populations (body powered device users (n=15) and acquired limb deficiency (n=20)) rated their satisfaction performing DTM tasks lower than their satisfaction performing non-DTM tasks (one-sample, 2-tailed t-test: p=0.017, 0.024, 0.013) (Table 3.6). Participants who used body powered devices and people with acquired amputations reported significantly higher difficulty performing DTM tasks than non-DTM tasks (p=0.021, p=0.026).
Intra-population analysis revealed that participants with acquired limb deficiency reported that it was significantly more important to them that they be able to perform all
41


tasks with their prosthesis than participants with congenital limb deficiency (p=0.011). No
other populations had significant differences in response.
When asked about potential improvements to existing prosthesis, a majority of participants thought individual control of fingers, ab/adducting thumb, stronger grip, and faster grip would be "useful" or "very useful" (15,19, 19, and 10 out of 15) (Table 3.7). Nineteen participants responded that adding flexion/extension would be "useful" or "very useful" and thought flexion/extension plus rotation would be "useful" or "very useful". Seventeen participants felt that a fully functional wrist with flexion/extension, radial/ulnar deviation, and rotation, would be "extremely useful".
42


Table 3.7: User responses to the question: "Please rate how USEFUL you think each option would be". Leading suggestions were individual control of fingers, ab/adducting thumb, stronger grip, faster grip, and increased degrees of freedom in the wrist. Majority
responses are highlighted in green.
Suggested Improvement Number of Responses
Not at all Slightly Somewhat Very Extremely
useful useful useful Useful Useful
Individual control of fingers/hook components 3 1 3 4 11
Rotating thumb to allow a pinch motion (Ab/adduction of thumb) 1 0 1 6 13
Sensory feedback from fingers/hook 4 2 3 2 11
Faster finger/hook motion 1 1 2 2 13
Stronger grip 2 0 2 4 15
Wrist with only powered rotation (Pro/Supination) 2 5 7 6 2
Wrist with only up/down motion (Flexion/Extension) 4 4 3 12 0
Wrist with rotation and 0 (2 13
Flexion/Extension z z b
Wrist with powered side to side motion (Radial/Ulnar deviation) 4 2 4 9 4
Fully functional wrist with rotation, 15
Flexion/Extension, and Radial/Ulnar deviation 2 2 2 2

Passive wrist that can be positioned in anatomical positions 4 1 5 5 7
Five of the participants volunteered for a follow-up interview. Four of these were conducted and recorded via Skype and one was conducted in person and recorded via camera and microphone. Of all the keyed comments in the interviews, the most common
43


issues that participants reported having were: Completing ADL tasks (15.49%), Difficulty
using tools (14.79%), Cosmetic (11.27%), Durability (11.27%), and Cost (10.56%) (Figure 3.2). All of the participants reported that they wanted improvements and that they would be willing to use a device with a more complicated control system in order to get increased functionality. When asked specifically if they thought a device that incorporated the DTM would be useful, all participants reported that they did and that it would be worth the increased complexity.
Figure 3.2: Pie Chart showing how often participants made each of the keyed comments during their interview. Overall, nearly one third of comments were about difficulty using tools or completing other ADL tasks. Another third addressed common complaints with prosthetics (e.g. durability, socket, cost, weight, and electronics).
3.4 Discussion
This study found that participants were less satisfied using their prosthetic devices to perform ADL tasks that involve the DTM than those that do not. In interviews, difficulty
44


performing ADLtasks and tasks that specifically use the DTM accounted for 30% of keyed
user comments.
Paradoxically, many survey participants rated tasks as difficult and important but also claimed that they were satisfied performing these tasks with their current device. During interviews, participants explained that this was because they had developed compensatory motions that allowed them to perform the task to their satisfaction even though they still found it difficult. For example, participants mentioned that they could use their intact limb to position an object before picking it up with their prosthesis. In many cases, participants stated that they simply avoid performing the task or ask someone else to do it for them. While these may be effective methods for completing tasks, compensatory motions can lead to long-term over-use injuries and avoiding or getting help with certain tasks limits the user's independence.
Of particular interest were responses to the questions about which additional DOF participants wanted to see in a prosthetic wrist. Most users said they would like FE or RU in a device but few requested PS. This sentiment is consistent in user surveys going back as far as 1985 (69). Flowever, it contradicts the standard thinking in prosthetic wrist design that PS should be the second DOF included after a gripping unit. Interviewees explained this seeming inconsistency by stating that they don't need PS because they can effectively simulate it by abducting the shoulder.
In most cases, interviewees volunteered much more information than the question specifically asked. This led to four of the five participants stating, without prompting, that they would like some sort of wrist that included FE and RU. Other suggestions included: a
45


smartphone app to adjust the gain in a myoelectric device throughout the day, customizable grip shapes in a hand, and thinner fingers (e.g. for reaching into pockets).
Other complaints about prosthetic devices were consistent with previous literature (22,67). Everyone interviewed mentioned problems such as: uncomfortable and insecure socket, inconsistent myoelectric response, weight, cost, and durability. Combined with the fact that a third of comments referred to difficulty performing ADL and using tools, these responses support the idea that a powered wrist unit that incorporates the DTM would be beneficial to this population.
These results could be interpreted as showing that users simply want a more functional device. However, this study was designed to determine specific improvements users were looking for. Notably, users in the survey and interviews were strongly interested in having a device with powered FE or RU. Additionally, given that any increase in functionality will make the device more difficult to use, the focus should be on the improvements that users are excited by and motivated to try.
A limitation of this study was the fact that self-selected and self-reporting populations carry an inherent bias: people who choose to fill out online surveys tend to be the people with stronger positive or negative opinions. This is a known limitation of this type of survey (83). Efforts were made to recruit from a diverse range of sources and questions were written in a neutral voice so as to avoid leading.
46


3.5 Conclusion
The limited functionality of current prosthetic devices negatively affects the quality of life of people with amputations. While it is sometimes possible to compensate for this limitation, such movement patterns ultimately lead to additional health problems (50). Simply adding all the missing DOF would be immensely difficult to control by users with today's technology.
This study determined what improvements are desired by the people with upper limb amputations in their prosthetic arm and hand devices. Specifically, DTM was identified as a DOF which users have particular difficulty mimicking with their device. The results suggest that a device incorporating the DTM would improve functionality for most users. Additionally, people were generally willing to accept a more complicated device in return for greater functionality. This motivates the development of a new prosthetic wrist device that incorporates this motion. Based on the interviews, a device like this would be welcomed by this population as long as it provided improved functionality, was durable, did not cost too much, and was easy to use.
47


Chapter IV
Measurement of the Dart Thrower's Motion in the Context of a Prosthetics-Specific Functional Test
The study in the last chapter identified the lack of the DTM in prosthetics as a potential factor limiting users' ability to perform ADL tasks. These survey results and the DTM's presence in many ADLtasks indicate that the DTM is a strong candidate for creating an improved, simple wrist (27). Previous studies have identified the DTM as motion in a plane rotated 20-40 degrees from pure flexion-extension but a more precise measurement is needed to build a prosthesis (20,27,37). To determine this rotation angle, able-bodied participants completed a series of standardized tasks while equipped with electrogoniometers to measure wrist motion.
Measurement of the wrist determined that FE and RU were coupled in the DTM plane in a majority of tasks and were rarely coupled in the anti-DTM plane. The DTM accounted for 46% of the overall motion during the grasp phase with rotation accounting for another 28% of the variance. In the 2-dimensional FERU plane, 67% of the variance in the functional range of motion of the wrist was coupled at an of angle 21.8 degrees rotated from pure FE. These findings suggest that a prosthetic wrist that combines these DOF would improve functionality and only minimally increase control complexity.
The results presented here are intended to give insight into the motion of the wrist during a standard of care assessment tool for upper limb prosthetics. The focus is on improving the functionality of prosthetic devices by combining two or more of the motions
48


identified by this study. The results are, potentially, also useful to people studying nonamputation disabilities of the arm (e.g. arthritis) as well as to medical professionals assessing their patients.
4.1 Methods: Development of a Wrist Measurement System
Measuring the motion of the wrist is a challenging task. Although the wrist can be approximated by a 3 DOF joint containing two pin joints and a rotation, FE and RU are complex motions of the eight carpal bones and PS is the function of the radius and ulna, not the wrist itself. Attempts to accurately measure the wrist with an active-marker motion capture system were unsuccessful. Flowever, electro-goniometers consisting of flexion-sensitive resistors proved an effective way to measure the wrist during this study.
4.1.1 Motion Capture
Motion capture is one of the more common methods for measuring physiological motion. Markers placed on anatomical landmarks are recorded by cameras and their position in 3D space is calculated by software. From these measurements, it is possible to calculate the positions of body segments and joint angles (85). Initial attempts to use a motion capture system (Optotrak Certus, Northern Digital Inc. Ontario, CAN) proved unsuccessful. After an initial bout of hardware and software issues, pilot testing revealed several limitations to this system.
The Optotrak system uses a single tower of three cameras to locate "active markers" instead of the commonly used infrared (IR) reflective markers. These markers have IR LEDs
49


that flash at distinct frequencies that the system can automatically identify. This avoids the
time-consuming need to post process the data and identify the markers in software manually. This is an advantage in some situations where the markers are directly facing the camera but, in this case, markers were often obscured from the camera during wrist rotation making this system ineffective for this study.
Attempts were made to remedy this issue by creating rigid plates with four markers in fixed positions (Figure 4.1, top). These plates were attached to the proximal and distal dorsal forearm, and to the dorsal hand. This setup allowed the Optotrak software to determine the location of the body segments. However, camera position was still a limiting factor: from wrist pronation to supination the dorsal hand plate rotated out of view of the camera and the position was lost.
50


Figure 4.1: Attempts at overcoming the Optotrak limitations. Rigid plates (above) were attached to body segments which correctly located the segment but could only be seen from a fixed vantage point. Multiple markers attached to each segment and defined as a "rigid body" (below). Lighter lines and circles indicate connections and markers on the opposite side of the limb. This method was also ineffective because the Optotrak software was unable to account for soft tissue artefacts (i.e. minor motion between markers from finger abduction or forearm skin stretch) in determining the location of the rigid body.
Another attempt at solving this issue involved placing 8-10 markers on the forearm and hand and defining a "rigid body" in the system (Figure 4.1, bottom). This allowed the Optotrak software to calculate the position of the other markers based on the position of at least three visible markers. However, this calculation is distortion-intolerant and even the slight change in relative position of the markers due to soft tissue artefact (e.g. skin motion from finger abduction) were enough to prevent the Optotrak software from accurately determining the position of the body segments.
Given these limitations, this form of measurement was abandoned. Consideration was given to using the motion capture facilities at The Center for Gait and Movement Analysis and the Interdisciplinary Movement Sciences Laboratory at the University of Colorado, Anschutz Medical Campus. However, these options were dismissed in favor of using
51


electrogoniometers which are easier to transport, more customizable, and produce data
that is easier to process.
4.1.2 Flex Sensitive Resistors
A system was built using 4.5" flex sensitive resistors (FSR) (Spectra Symbol, Salt Lake City, UT) as a proof of concept for a commercial goniometer system. The design was based on the system described by Wang et. al. (86). Unlike Wang's design, only one FSR was necessary to measure each axis in this case as the FSRs were equally responsive in each direction. Voltages were read from the FSRs via an Nl DAQ2006 board (National Instruments Corporation, Austin, TX) and recorded in LabView (National Instruments Corporation, Austin, TX).
The idealized voltage response from the FSR circuit (Figure 4.2) is
where Vin is the voltage supplied by the Nl DAQ, Vout is recorded by the Nl DAQ, Rfsr is the value of the FSR, and R2 is a fixed resistor.
(4.1)
52


i
d
IMPEDANCE vn-V ( |
BUFFER v out v y r, + r /
Figure 4.2: Circuit diagram for the flex sensitive resistor setup. Vjn and Voutwere generated and recorded in Labview.
The FSR system was tested by attaching the FSR to a manual goniometer and recording the voltage in Labview and the indicated angle (Figure 4.3). Measurements were made three times in each direction. Minimal hysteresis was found to be 0.04 V over a 0.59 V range, a 6.7% variation. Based on these measurements, the corresponding anatomical angle (in degrees) was determined from voltage by
9 = 232.6 V+ 121. (4.2)
53


Angle vs Voltage (FSR)
Figure 4.3: Angle vs Voltage in flex sensitive resistor (FSR) testing. FSR was attached to a plastic goniometer and the voltage was recorded via Labview at various angles and moving in the forward (black) and backward (white) directions.
Figure 4.4: Mockup of the FSRs attached to the dorsal and ulnar sides of the wrist.
After positive bench test results, a pilot test was performed by attaching the FSRs to the dorsal and ulnar wrist (Figure 4.4). The pilot test revealed that the FSRs could accurately measure both FE and RU (Figure 4.5). Based on these findings, it was concluded that electrogoniometers based on FSRs would be an effective method for measuring wrist motion.
54


Figure 4.5: Plot of the wrist angles flexion-extension (FE) and radial-ulnar deviation (RU) recorded with the FSR system during simulated dart throwing.
4.1.3 Commercial Electrogoniometers
Commercially available electrogoniometers (Motion Labs Systems Inc., Baton Rouge,
LA) were tested by attaching them to a manual goniometer and recording the output voltages were in LabView. A torsionmeter (rotation sensitive resistor) was tested by attaching it to two pieces of pipe marked with angles and rotating. The voltage was again recorded with LabView. Measurements were taken three times in each direction and averaged. The goniometer and torsionmeter had a linear relationship between voltage and angle within the operating range of -90 degrees to 90 degrees (Figure 4.6). Flysteresis was minor and accounted for less than 2.5 degrees over 180 degrees (average of 2.1 0.7 for FE, 1.7 0.5 for RU, 1.9 0.7 for PS).
55


Goniometer and Torsionmeter Calibration Curves
FE (- to +y = -102.25X + 37.14, R2 = 0.9996 # FE (+ to = -103.53x + 40.043, R2 = 0.9994 RU ( to = '107-7x n-216' R2 = 0.9999 RU (+ toV-f '10776x 9-1336, R2 = 0.9998 PS (- to +y = -117.06X + 3.722, R2 = 0.9987 PS (+ toY)= -116.6lx +6.1324, R2 = 0.9988
Voltage (mV)
Figure 4.6: Results of calibration test of the goniometers and torsion meter over 180 degree operating range. Flysteresis was found to be <2.5 degrees over 180 degrees. Relationship between output voltage and angle is highly linear.
4.2 Methods: Standardized Motions
The Southampton Fland Assessment Procedure (SHAP) is a validated, standardized measure of hand functional ability that is becoming a standard assessment tool in prosthetics research (Figure 4.7) (28). It was developed by Colin Light, Paul Chappell and Peter Kyberd in 2002 at the University of Southampton (87). The SFIAP also has high test-retest reliability (88).
56


Figure 4.7: SHAP design showing the abstract objects (ball, cylinder, etc.) and simulated ADL objects (cup pouring, key turning, etc.). Participants perform timed movements of the objects from which the assessor can calculate the SHAP score (39).
During the SHAP, participants manipulated six light abstract objects, six heavy abstract objects (e.g. a ball or cylinder), and completed fourteen simulated ADL tasks {e.g. pouring water from a jug or turning a door handle). A list of all SHAP tasks can befound in Table4.1. Participants start a timer with ther active hand, move the target object using the instructed grip type an d techn iqu e, and th en return to stop th e ti mer. The S HAP retu m s a score based on the time to complete each task. This score ranges fromO (no function or could not completetasksJtolOO (normal function). While time to completion is not the only important measure of a prosthesis, the SHAP has been shown to be able to identify clinically relevant differences in prosthetic functionality (23).
57


Table 4.1: List of tasks involved in the Southampton Hand Assessment Procedure.
Light Abstract Objects Simulated ADL Tasks
L Sphere Place coins in Jar
L tripod Undo buttons
L cylinder Cut food
L Lateral Turn page
L tip Remove jar lid
L extension Pour water from Jar
Pour water from carton
Heavy Abstract Objects Move heavy object
H sphere Move light object
H tripod Lift and move tray
H cylinder Turn key
Hlateral Undo zipper
Htip Turn screw with screwdriver
H extension Turn door handle
4.3 Methods: Principal Component Analysis
Principal component analysis (PCA) was first introduced by Pearson in 1901 and later by Hotelling in 1933 (90). It is used to find correlations between multiple variables by finding a new basis into which the data can be rotated to maximize the variance in the first orthogonal dimension (91). PCA is useful for identifying trends in large data sets and reducing the dimensionality of data sets with many DOF. On biomechanics data, PCA is
58


capable of "identifying functional units of coordination in the form of synergies or coordinative structures at both the kinematic and the muscular level" (91, p416). These features make PCA potentially useful in identifying a way to reduce the wrist from three DOF to one.
The following discussion of PCA and its underlying mathematics rely heavily on Daffershofer, 2004; Demsar, 2012; and Schlens, 2003 (90-92).
PCA rotates an m x n matrix, X, into a new basis via a rotation matrix, P, with the goal of maximizing the variance in the first dimension. This is written as
X = TPT (4.3)
and
T = PX (4.4)
where P is an orthonormal matrix and T is the projection of X into the new basis. That is, the matrix X represents the collected data and T represents those data rotated such that the maximum variance is along the first axis. T is referred to as the "score" and P is referred to as the "loading".
The matrix P is calculated in such a way that its columns are ordered in the descending order of variance. This can be calculated from the covariance matrix of X which is defined by
59


(4.5)
Covx = XXT
A n-1
X must have zero mean in this case. It is important to note that the PCs can also be calculated from the correlation matrix of X. These terms are interchangeable in this derivation.
P is calculated by
PXPT = Covx = XXT (4.6)
where X is a diagonal matrix of ordered Eigenvalues. The Eigenvectors and Eigenvalues are typically calculated by a computer programs such as Matlab or Mathematica. Anyone interested in finding the Eigenvectors by hand are encouraged to read chapter 10 of Mathematical Methods in the Physical Sciences by Boas (93).
The angle of rotation from the original data to the PC space for the nth PC can be found by
where PCn i and PCnj are the ith and jth values of the nth principal component. These could, for instance, indicate the ratio of contributions of FE and RU to the DTM in the first PC. Physically, this represents the angle of rotation from the pure anatomical motion planes (FE, RU, PS).
60


P and X are useful for many reasons. P describes the rotation matrix that transforms the
original data into a new basis where the first variable accounts for the most variance. This identifies how the different variables are coupled and in what direction and also whether there is redundancy in the variables. The sum of the Eigenvalues, X, is related to the total variance of the system. Therefore, the % of variance explained, e, by any principal component (PC), k, is given by
*k= ^-*100% (4.8)
That is, the variance explained by each PC is the associated Eigenvalue divided by the sum of the Eigenvalues. This value can inform a reduction in dimensionality of the system. For example, if there are five variables but the first three PCs account for 80% of the variance, then the last two PCs can be dropped while only losing 20% of the information.
This is the application of PCA that was used in this study. Measurements were made on the three DOF of the wrist. PCA was applied to the resulting data with the goal of determining if any of the DOF were coupled and if a majority of the variance was contained in one or two DOF. If this is the case, it was determined that the wrist could be effectively reduced from three DOF to two or one DOF.
4.3.1 Spatial vs. Temporal PCA
PCA was originally developed to describe spatial data but can be expanded to give the same insights into temporal data. This can be particularly useful in biomechanics data. Flowever, this is most easily done on cyclical data, such as the angles of the leg joints during
61


walking. This was not the case in this study because each of the SHAP activities are single
actions and the motions from multiple participants do not have a specific event to use to synchronize it in time (such as heel strike in gait). Therefore, the positions of the wrist during each task were concatenated and plotted as three-dimensional spatial data (FE vs RU vs PS). This effectively removes any temporal information. Therefore, the result of this analysis was effectively a functional range of motion in the SHAP.
4.3.2 A Statement About Error Reporting in PCA
There has been a great deal of discussion about how to report error in PCA. One of the disquieting things about PCA is that "there are no parameters to tweak and no coefficients to adjust based on user experience the answer is unique and independent of the user" (12, pl2). That is, PCA simply finds the basis that maximizes variance. The results of PCA are simply reported as the Eigenvectors and the % variance explained by each. There is no "goodness of fit" value to report as with linear regression and random error is not propagated through the analysis because it is accounted for in the Eigenvalue. Indeed, if the error is evenly distributed about the mean, it should have no effect on the PC calculation. If this is not the case, the raw data should be used.
One way to quantify how well the PCA describes the data is to report the signal-to-noise ratio (SNR) in the first PC. This method was suggested by Shlens, who described the SNR as
62


SNR =
signal ___ G pci
(4.9)
where o2 is the variance and SNR1 indicates data with high precision (92).
While this is an interesting approach, it doesn't add any more information than is already given by the Eigenvalues. That is, if a PC is already known from the Eigenvalues to account for, say, 80% of the variance, no more insight is gained by calculating that 80/20l.
Because of this, the PCA in this chapter will be reported as the angle calculated from the PCs and the variance described by each of those components.
4.4 Methods: Experiment
A convenience sample of fifteen adult volunteers (18-85 years of age) with intact upper limbs was recruited to participate in this study. Participants completed the SHAP as described by the SHAP protocol. Wrist position was recorded using electrogoniometers and processed in MatLab. Experiments were conducted at the Bioengineering Department's Biomechatronics Laboratory at the Colorado Children's Hospital. Demographic information (gender, age, height, weight) was also collected. Inclusion/exclusion criteria were chosen to ensure that the sample represented the wrist motion of people with normal physiology.
4.4.1 Inclusion/Exclusion Criteria
Participants were required to meet the following inclusion criteria:
Between the ages of 18 and 85 years
63


Intact upper limb physiology
No complicating cognitive or physical impairments
Normal or corrected vision and hearing
Ability to read and understand English at a cognitively appropriate level
Ability to understand and follow instructions in English
Ability to read, understand, and give consent to participate
Participants were excluded from taking part in this study based on the following criteria:
Below the age of 18 or above the age of 85
Pathologic upper limb physiology
Pathological cognitive or physical impairment
Inability to read, understand, or follow directions in cognitively appropriate English
Unable or unwilling to give consent
There were no exclusions based on race/ethnicity, gender, or socioeconomic status.
Participants were screened to ensure they met the inclusion/exclusion criteria before beginning the study. These inclusion criteria were required to ensure that measurements represented anatomical motion without complications from impairment.
64


4.4.2 Recruitment
Participants were recruited via email, word of mouth, and flyers posted at the Anschutz Medical Campus. Participants contacted the researchers via email or phone to schedule a time to come to the Biomechatronics Laboratory at the Colorado Children's Hospital. The test took less than an hour to complete. Each participant read and signed an approved consent form before beginning the study. No personally identifiable information was recorded.
Participants were not compensated for taking part in this study. This study was conducted with the approval of the Colorado Multiple Institution Review Board (COMIRB #14-0838).
4.4.3 Electrogoniometers
Participants were equipped with commercial electrogoniometers (Motion Labs Systems Inc., Baton Rouge, LA) on the dorsal wrist to measure FE and RU, and a torsionmeter on the volar forearm to measure pronation-supination (PS) during the tasks. Voltages were read from the goniometers via an Nl DAQ 2006 board and recorded in LabView.
Linear fit equations relating voltage to angle were incorporated into the LabView VI. Angle measurements were zeroed with wrist neutral, forearm neutral, and elbow at 90 degrees. Positive motion was defined as extension, radial deviation, and supination.
Voltage, angle calculator, calibration data were recorded in comma separated value formatted files.
65


4.4.4 Assessment
Participants performed the SHAP in accordance with the defined protocol (28). Participants were seated with the target object directly in front of their active arm and with the chair height adjusted so that their forearm was parallel with the table when the elbow was at 90 degrees. Participants began each task with their hand flat on a timer, pressed the timer, manipulated the object, and returned to the timer. This created a distinctive "reach phase", "grasp phase", and "return phase". Recording began before the participant began the task and continued recording after they had ended the task. This resulted in a "rest phase" before and after each trial. Participants also simulated throwing a dart to visualize how closely this motion related to the physiological DTM.
4.4.5 Processing
Data were processed in Matlab (The MathWorks, Inc., Natick, Massachusetts). Each trial was filtered using a 4th order, zero-phase, Butterworth filter with a 10 Hz cut-off frequency to remove noise and preserve biological motion (94). This cut-off frequency is supported by the literature (91,95,96) and is confirmed by frequency analysis of pilot data showing that the majority of motion information is contained in the <10 Hz range (Figure 4.8).
66


Single-sided amplitude spectrum of wrist angles
Figure 4.8: Frequency spectrum of sample wrist motion showing that the majority of information is below 10 Hz. This justifies the use of a filter with a 10 Hz cutoff frequency.
The beginning of the reach phase was identified by a deviation of greater than 5 degrees from the initial rest phase in any of the three angles. The end of the reach phase (beginning of grasp phase) was manually selected by a visual comparison of the time plot of the trial to a recording of participants only performing the reach phase of each task. The end of the grasp phase was manually selected by visual inspection at the beginning of the movement back to the final rest position.
67


Figure 4.9: Demonstration of the phases of a SHAP task during a single trial. The initial rest phase results from the need to start recording before the participant beginsthe task. The reach phase typically involves the greatest motion as the hand is positioned at the object. The Grasp phase is typically characterized by the least motion and determined by recordings of the participant only holdingthe object. The final rest phase beginsonce the object is released and continues beyond the end of the task until the recording can be
manually stopped.
The grasp phase is the more important of the two for interacting with objects (97). The redundant nature of the DOF of the upper arm allows people to take many different paths to reach an object. This means the reach phase includes more variance than grasp. More importantly, the course motions people use to approach an object matters far less than how the hand and wrist are positioned for object manipulation. Therefore, while this study also analyzed the motion during reach phase, the conclusions and next steps are based on the results of the grasp phase.
4.4.6 Principal Component Analysis
PCAwas performed in Matlab using the pca() function. Each reach and grasp phase was interpolated to create a phase plot of 0-100% (101 points). PCA was first performed on the z-score (mean subtracted, variance divided) on the collection of all SHAP trialsto find the correlations between DOF. Since PCA measures variance, centeringthe data and dividing by
68


the variance prevents DOF with large motions from dominating the results and skewing the
fit (91). Without dividing by the variance, rotation dominates all of the components and any understanding of how the three DOF move relative to each other is lost. Performing PCA on the z-score measures how the three DOF move relative to each other and reveals coupling between the motions. This provided an assessment of the functional range of motion of the wrist during the SFIAP.
Coupling was assessed in each of the SFIAP tasks by performing PCA on the z-score of all participants in each task. "Strong coupling" between DOF in the PCA components (PCs) A and B was defined as | A| >0.5 & | B|>0.5 and % contribution to variance >50. This indicated that at least half of the information in the motion was contained in that PCA component (91). Opposite coupling (-) was defined as Sign(A) Sign(B), while parallel coupling (+) was defined as Sign(A) = Sign(B). In this configuration, the DTM was represented by positive FE and negative RU (FERU-) while the anti-DTM motion was represented by FE and RU with the same sign (FERU+). A random sample would show strong coupling in one direction (opposite or parallel) 12.5% (1/16) of the time.
Once coupling was found between two of the three DOF, PCA was performed on the centered data. In this case, the position was not divided by variance so as to preserve the proportionality of the motion of the two variables rather than simply the correlation between them.
69


4.4.7 Coupling Counts and Position Analysis
The percentage of tasks which used each type of coupling (FERU, RUPS, and PSFE) and in which direction there was coupling was determined. Of chief concern was that some tasks had positive coupling and others had negative coupling for a given combination of DOF. Creating a device that coupled those DOF in only the positive or negative direction would result in a device that was helpful some of the time but detrimental the rest of the time.
Finally, how often the wrist was in each of the possible positions (which of 8 quadrants on a FE-RU-PS plot) was counted to determine whether the wrist has a preferred position. This allowed for the determination of the typical orientation of the wrist during each phase. This was yet another method for determining whether there was a particular combination of DOF that could be combined in a device.
4.5 Results
Fifteen adults (aged 30.7 8.8 years) performed the SFIAP while wearing electrogoniometers to track wrist FE, RU, and PS. Each participant successfully completed the entire SFIAP with no complications. Participants average SFIAP score was 97.6 3.9 out of 100, indicating that each participant had normal hand functionality and provided an accurate measure of intact hand and wrist motion during the SFIAP.
70


4.5.1 PCA on Reach and Grasp
Reach and grasp motions were similar across all participants. Although the starting points were different between participants, the difference in motion (motion error) was small.
The average motions of each task showed a consistent pattern with minimal error across the task (mean error = 0.83 degrees) as seen in (Figure 4.10). The first principal component of the reach phase was PCi = [0.59, -0.57, 0.57] (FE, RU, PS) with a % contribution to variance of 45.61. The first principal component of the grasp phase was PCi = [-0.64, 0.68, -0.36] with an average % total variance of 45.12.
71


Figure 4.10: Motion of the wrist during the reach phase (top) and grasp phase (bottom) of each task in the SFIAP averaged across participants. Error bars indicate standard error. For each task, a distinctive motion from the rest position to the beginning of the grasp position is visible. During the reach phase, there is a visible coupling between FE and PS. During the grasp phase, many of the tasks have similar motions while the more complex tasks (coins, buttons, pouring) have distinctive profiles. During grips, FE and RU are commonly correlated. Simulated dart throwing (DTM) was included as a comparison.
72


Table 4.2: Results of PCA on the z-score of all SHAP tasks. The first component accounted for about 45% of the motion in both the reach and grasp phases. In the reach phase, the three motions were equally coupled and FE and RU were coupled in the DTM direction. In the grasp phase, FE and RU were strongly coupled in the DTM direction while rotation made up nearly all of the motion in the second PC. Note that the sum of the squares of the components of each PC equals to one.
PCI PC2 PC3
FE 0.5884 0.0082 0.8085
Reach RU -0.5715 0.7116 0.4087
PS 0.5719 0.7026 -0.4234
PCA % Variance 45.6101 27.5587 26.8311
FE -0.6391 -0.3844 0.6662
Grasp RU 0.6797 0.1230 0.7231
PS -0.3599 0.9149 0.1827
% Variance 45.1243 31.9224 22.9533
The calculated offset angles in two dimensions for the reach phases (Figure 4.12 (left)) and for grasp (Figure 4.12 (right)) phases are shown below.
4.5.2 2-Dimensional PCA on Averaged Data
Initially, the offset angle in each 2D plane was calculated on the wrist data averaged over participants. In this method, the largest offset angle in grasp phase was in the axial (FERU, i.e. DTM) plane at 26.72 degrees accounting for 76% of the variance. Offsets in the PSFE and RUPS planes were 8.3 degrees with 89% variance and 6.8 degrees with 80% variance, respectively.
73


Magnitude of wrist angles in FERU plane (grasp) 0 = 26.7, % latency = 75.8
Flexion
30
Extension
Figure 22: 2-dimensional radial plot in the FERU plane on the averaged wrist data. Red and green lines represent the first and second principal component. Their lengths represent
the percent contribution of each PC.
4.5.3 2-Dimensional PCAon Full Data Set
As noted in section 4.3.2, there are consequences to making the assumption that error is evenly distributed. In the initial analysis of the angles in each 2D plane, the data were averaged across participants before PCA was performed (section 4.5.2). This was done with the intent of smoothing and simplifying the data. However, this assumption is only accurate if the error is normally distributed and, in this case, it is not. Makingthis assumption was the equivalent of taking an intermediate mean of the data.
Recalculating PCA on the 2D data without averaging, the largest offset angle during the reach phase was 28 degrees accounting for 68% of the variance in the PSFE plane. Minor angles were 10 degrees in the FERU plane accounting for 73% of the variance and 9 degrees in the RUPS plane accounting for 81% of the variance. In the grasp phase, the largest offset angle was 22% in the FERU phase accounting for 67% of the variance followed by 4 degrees at 72% variance in the PSFE plane and 2.5 degrees at 81% variance in the RUPS plane.
74


Magnitude of wrist angles in FERU plane (reach) 0 = 9.93, % latency = 72.9
Extension
Flexion
Magnitude of wrist angles in FERU plane (grasp) 0 = 21.8, % latency = 66.8
Extension
Magnitude of wrist angles in RUPS plane (reach) 6 = 8.9, % latency = 80.7
Ulnar
Magnitude of wrist angles in RUPS plane (grasp) 0 = 2.4, % latency = 81.2
Ulnar
00
60
Magnitude of wrist angles in PSFE plane (reach) 0 = 28, % latency = 68
Pronation
Supination
ixtension
Flexion
Magnitude of wrist angles in PSFE plane (grasp) 0 = 3.9, % latency = 72.5
Pronation
Extension
Flexion
Supination
Figure 4.12: Two-dimensional radial plots of wrist angle in the three planes during the reach phase (left) and grasp phase (right). Red and green lines represent the first and second principle components, respectively. Lengths indicate the percent contribution to the variance of each PC. Boxed plots show the largest deviation in each of the phases.
75


4.5.4 Coupling
The coupling during the reach phase was significantly greater than random in the FERU-(DTM), RUPS-, PSFE+ and and between all three (Figure 4.24). The coupling during grasp phase was greater than random in all but the FERU- (anti-DTM) mode. The levels of coupling
only differed between reach and grasp phase in the PSFE- mode.
90 80 70 So 60
| 50 8 40 | 30 20 10 0
Percentage of SHAP Tasks with Strong Coupling
JiUji
FERU+ FERU- RUPS+ RUPS- PSFE+ PSFE- FERUPS (DTM)
Coupling type
Random (12.5%)
Reach
Grasp
Figure 4.24: Coupling in the three degrees of freedom of the wrist. Error bars indicate one standard deviation from the mean. We found significant coupling in the reach phases of FERU- (DTM), RUPS-, PSFE+/- and in all three degrees of freedom. Greater coupling in the grasp phase was found in all but the FERU+ (Anti-DTM) mode.
The majority of the grasp phases of SFIAP tasks took place with the wrist in either flexion, ulnar deviation, and pronation or in extension, radial deviation, and supination (Figure 4.25).
76


Quadrant distribution of wrist angles during SHAP (reach)
sS
Quadrant distribution of wrist angles during SHAP (grasp) 25-----------,--------,--------,---------,--------,---------,--------,--------
20 -
15 -
10 -
5 -
FRP FRS FUP FRP ERP ERS EUP ERP Quadrant
Figure 4.25: Percentage of SHAP task reach (left) and grasp (right) phases in given wrist orientation (F-flexion, E-extension, R-radial deviation, U-ulnar deviation, P-pronation, S-supination). During the reach phase, wrist position is present in all combinations with FUP the most common position. A plurality of grasps takes place with the wrist in flexion, ulnar deviation, and pronation or in extension, radial deviation, and supination. This presentation
is consistent with the DTM.
4.6 Discussion
Wrist motion was coupled during both the reach and grasp phase. In the reach phase, the first PCA accounted for 46% of the variance and was made of roughly equal parts FE,
RU, and PS, indicating no preferred direction in this phase. While this does not suggest an appropriate reduction in DOF, the FE and RU components did have different signs indicating a DTM is part of the overall motion during reach
In the grasp phase, the DTM accounted for 45% of the motion in the first PC, while the second PC, at 28% was almost entirely PS. This suggests that a simplified prosthetic wrist could accomplish almost half of the functionality by including the DTM and over two-thirds of the anatomical motion by including both the DTM and PS. In fact, the third PC, accounting for 22% of the variance, appears to be made up of the anti-DTM motion,
77


suggesting that a 3-DOF wrist might perform better with PS, DTM and anti-DTM, rather than
the anatomical, PS, RU, and FE (i.e. PS plus an offset flexion and deviation unit). The fact that the DTM accounted for 76% of the variance in the FERU plane is a strong indicator that it would be appropriate for reducing the wrist to 1 DOF. The RUPS and PSFE plots in Figure 4.12 (middle and bottom right) show little correlation and would be poor candidates for dimension reduction.
A second way to determine which combination of DOF would be most effective is to measure how the DOF were coupled in the wrist motion during grasp phase. The DTM (FERU-) was the most coupled motion in the wrist and there was little anti-DTM (FERU) coupling (Figure 4.24). There was also significant opposite coupling between RU and PS, between PS and FE and between all three motions. While these results indicated that there are several options for combining DOF in a prosthetic wrist, the other couplings were strong in both the and + directions. Therefore, about half of the time a coupled RU-PS or PS-FE would be helpful but the rest of the time it would be a hindrance.
The DTM combination is the most promising because it had a high negative correlation (DTM) and low positive correlation (anti-DTM). That is, most of the time, the motion was in the DTM direction. Therefore, including the DTM would help more than half of the time and would rarely be detrimental. In addition, there were many trials that had strong coupling in all three motions. This is encouraging but a prosthetic wrist that incorporates all of these motions has never been commercially available and building such a wrist would not solve the issue of not having enough control inputs to control all of the DOF.
78


As a final argument for using the DTM to reduce the dimensionality of the wrist, this
study investigated in what position the wrist spent the most time during the reach and grasp phases (Figure 4.25). During the reach phase, there was not a preferred position of the wrist. However, during the grasp phase, a plurality of the time the wrist was in either FUP or ERS. These positions are consistent with motion in the DTM plane.
4.7 Conclusion and Next Steps
This study demonstrated that FE and RU are coupled in both the reach and grasp phase of ADL tasks (as represented by the SHAP). This suggests that a single DOF, based on the DTM, included in a prosthetic wrist, would be helpful to most users. This hypothesis is tested in the next chapter.
In the future, this study could be replicated using other measurement systems. Although the goniometers were determined to be appropriate for this study, motion capture or radiography could potentially provide a more precise measurement. This could indicate how the bones of the wrist move during this motion which could be useful for a different wrist design or for informing wrist surgery practices. However, these methods are difficult to use and resource intensive.
Although it was not the focus of this thesis, the angles of the hand joints were also recorded using a Cyber Glove II (CyberGlove Systems LLC., San Jose, CA) in eight of the participants. This was done to help design a new postural controller for the prosthetic hand and wrist based on the work done in 1998 by Santello etal. (97). Santello showed that the 15 motions of the hand could be effectively described using only two degrees of freedom
79


(DOF) while maintaining more than 80% of the information. While this has allowed for the
creation of drastically more effective prosthetic hand control schemes, little work has been done to correlate wrist positions with these postures. This means that users must still perform compensatory motions in order to properly position the hand for grasping. The work done here shows that the wrist can be reduced to two DOF and incorporating the hand position data has the potential to identify correlation between the hand postures and wrist positions, expanding the functionality of postural controllers.
80


Chapter V
Testing a Novel Prosthetic Wrist that Incorporates the Dart Thrower's Motion
The last chapter demonstrated that a plurality of the motion of the anatomical wrist could be preserved in a prosthetic wrist with a single degree of freedom using the DTM. This study sought to determine whether this finding would translate to a single degree of freedom prosthetic wrist. Ten people with intact limbs completed the SHAP with a simulated amputation and prosthesis with a wrist in the DTM, flexion, and fixed configurations. Three people with intact limbs completed the SHAP with their hands and five people with transradial amputations completed the SHAP with a custom socket and prosthesis in the DTM configuration as a comparison. Shoulder motion was measured with electrogoniometers to determine whether the different wrist configurations affected the compensatory motions. After the test, participants completed the NASA task load index (NASA-TLX) to give a subjective measure of how difficult it was to use each device.
The DTM wrist significantly improved SHAP scores over a standard FE wrist (p< 10"7). However, shoulder ab/adduction did not appear to be reduced by the DTM wrist (p>0.05). About half of the participants did display reduced shoulder motion though and it was observed that participants used several other compensatory strategies that were not measured by this study. Using the DTM wrist was no more difficult than using the FE wrist or no wrist according to the NASA-TLX (p>0.05). This supports the hypothesis that reducing the dimensionality of the wrist using the DTM will improve performance while not substantially increasing the difficulty to use it. The secondary hypothesis, that the DTM wrist would reduce compensatory shoulder motions, was not confirmed.
81


Full Text

PAGE 1

D EVELOPMENT OF A NOVE L PROSTHETIC WRIST D EVICE INCORPORATING THE D ART T HROWER S M OTION by M ATTHEW L EE D AVIDSON B.S., Reed College, 2006 M.S., University of Colorado at Denver, 2011 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfilment of the requirements for the degree of Doctor of Philosophy Bioengineering Program 2017

PAGE 2

ii 2017 MATTHEW LEE DAVIDSON ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Doctor of Philosophy degree by Matthe w Lee Davidson has been approved for the Bioengineering Program by Cathy Bodine, Chair Richard F. ff. Weir Advisor James J. Carollo Kendall S. Hunter William Sullivan Date: July 29, 2017

PAGE 4

iv Davidson, Matthew Lee ( Ph.D. Bioengineering Program ) De velopment o f a Novel Prosthetic Wrist Device Incorporating t he Thesis directed by Research Associate Professor Richard F. ff Weir A BSTRACT The purpose of this research was to identify limitations people with arm amputations face comp leting daily l iving tasks and to design a new prosthesis which alleviates these deficiencies State of the art prosthetic devices can mimic many of the motions of an intact limb but are controlled by a limited number of signals from the muscles in the resi dual limb. The majority of current research is focused on improving the control of these devices by increasing the number of inputs or using software to interpret the limited inputs in a more meaningful way. This research instead determined that the mechan ics of the prosthesis could be simplified while maintaining functionality and a simple control system. Specifically, this research tested the hypothesis that the three degrees of freedom in the wrist (flexion extension, radial ulnar deviation, and rotation ), could be combined into a single degree of motion and functionality and could be controlled with a simple input method There are currently no commercially available wrist flexion devices which utilize this motion. The studies presented in this dissertation surveyed people with upper limb amputations and with their prosthesis. T bodied individuals to be 22 degrees offset from the anatomical flexion extension plane. Finally, a new prosthetic wrist device was developed based on this angle. This new

PAGE 5

v prosthesis improve d functionality over a traditional flexion wrist and was no more difficult to use than a device without a wrist. This research helps to alleviate many of the barriers to inclusion which people living with upper limb deficiency regularly face. The form and content of this abstract are approved. I recommend its publication. Approved: Richard F. ff Weir

PAGE 6

vi To my parents, Niza and Richard Davidson, for making this possible.

PAGE 7

vii A CKNOWLEDGEMENTS Thank you to my committee for their support, encouragement, and patience as I pursued this research. Their contributions helped steer this work to a successful completion. I am especially grateful to Richard Weir for his invaluable insights into the world of prosthetics and to Cathy Bodine for making me feel like I could accomplish this work. Thanks also to the other members of the biomechatronics lab for their help, both machine shop for building parts for me. A very big thank s goes to all my friends who supported me and volunteered their time to participate in my experiments: your volunteered time was invaluable. Finally, I want to thank Jyll Tuggle at Under the Umbrella Caf for feeding me and providing a safe, c omfortable sp ace for me to work countless bringing me delicious food so many times when I was too busy or tired to feed myself.

PAGE 8

viii T ABLE OF C ONTENTS I MOTIVATION AND RESEARCH GOALS ................................ ................................ ................................ 1 1.1 S UMMARY OF G OALS ................................ ................................ ................................ ................................ 3 1.1.1 Specific Aim 1: Identify what ADL tasks people with amputations find challenging via online survey and determine what improvements they are interested in seeing in their devices. .................. 3 Specific Functional Test ................................ ................................ ................................ ......................... 4 .. 5 II BACKGROUND ................................ ................................ ................................ ................................ ... 7 2.1 P HYSIOLOGY OF THE U PPER L IMB ................................ ................................ ................................ ................ 7 2.1.1 The Wrist ................................ ................................ ................................ ................................ ..... 9 ................................ ................................ ................................ ....... 11 2.2 T HE A MPUTATED U PPER L IMB ................................ ................................ ................................ .................. 14 2.2.1 Etiology ................................ ................................ ................................ ................................ ...... 14 2.2.2 Anato my and Surgery of the Upper Limb ................................ ................................ .................. 15 2.2.3 Considerations for People with Upper Limb Amputations ................................ ......................... 16 2.3 C URRENT P ROSTHETICS ................................ ................................ ................................ ........................... 17 2.3.1 Types of Prosthetics ................................ ................................ ................................ ................... 18 2.3.2 Prosthetic Wrists ................................ ................................ ................................ ........................ 19 2.3.3 Limitations of Current Prosthetic Devices ................................ ................................ .................. 22 2.4 M YOELECTRIC C ONTROL ................................ ................................ ................................ .......................... 23 III USER SURVEYS SUGGES T INCORPORATING THE DTM INTO A PROSTHETI C WRIST WOULD BE USEFUL 26 3.1 M ETHODS : D EVELOPMENT ................................ ................................ ................................ ....................... 27 3.1.1 Human Factors Testing ................................ ................................ ................................ .............. 27

PAGE 9

ix 3.1.2 Survey and Interview Questions ................................ ................................ ................................ 28 3.1.3 Survey Validation ................................ ................................ ................................ ....................... 31 3.1.4 Interview ................................ ................................ ................................ ................................ .... 32 3.1.5 Power Analysis ................................ ................................ ................................ ........................... 35 3.2 M ETHODS : E XPERIMENT ................................ ................................ ................................ .......................... 36 3.2.1 Inclusion/Exclusion Criteria ................................ ................................ ................................ ....... 36 3.2.2 Recruitment ................................ ................................ ................................ ............................... 38 3.2.3 Analysis ................................ ................................ ................................ ................................ ...... 39 3.3 R ESULTS ................................ ................................ ................................ ................................ ............... 40 3.4 D ISCUSSION ................................ ................................ ................................ ................................ .......... 44 3.5 C ONCLUSION ................................ ................................ ................................ ................................ ......... 47 IV MEASUREMENT OF THE N IN THE CONTEXT OF A PROSTHETICS SPECI FIC FUNCTIONAL TEST ................................ ................................ ................................ ........................... 48 4.1 M ETHODS : D EVELOPMENT OF A W RIST M EASUREMENT S YSTEM ................................ ................................ ..... 49 4.1.1 Motion Capture ................................ ................................ ................................ ......................... 49 4.1.2 Flex Sensitive Resistors ................................ ................................ ................................ .............. 52 4.1.3 Commercial Electrogoniometers ................................ ................................ ............................... 55 4.2 M ETHODS : S TANDARDIZED M OTIONS ................................ ................................ ................................ ........ 56 4.3 M ETHODS : P RINCIPAL C OMPONENT A NALYSIS ................................ ................................ ............................. 58 4.3.1 Spatial vs. Temporal PCA ................................ ................................ ................................ ........... 61 4.3.2 A Statement About Error Reporting in PCA ................................ ................................ ............... 62 4.4 M ETHODS : E XPERIMENT ................................ ................................ ................................ .......................... 63 4.4.1 Inclusion/Exclusion Criteria ................................ ................................ ................................ ....... 63 4.4.2 Recruitment ................................ ................................ ................................ ............................... 65 4.4.3 Electrogoniometers ................................ ................................ ................................ ................... 65 4.4.4 Assessment ................................ ................................ ................................ ................................ 66

PAGE 10

x 4.4.5 Processing ................................ ................................ ................................ ................................ .. 66 4.4.6 Principal Component Analysis ................................ ................................ ................................ ... 68 4.4.7 Coupling Coun ts and Position Analysis ................................ ................................ ...................... 70 4.5 R ESULTS ................................ ................................ ................................ ................................ ............... 70 4.5.1 PCA on Reach and Grasp ................................ ................................ ................................ ........... 71 4.5.2 2 Dimensional PCA on Averaged Data ................................ ................................ ...................... 73 4.5.3 2 Dimen sional PCA on Full Data Set ................................ ................................ .......................... 74 4.5.4 Coupling ................................ ................................ ................................ ................................ ..... 76 4.6 D ISCUSSION ................................ ................................ ................................ ................................ .......... 77 4.7 C ONCLUSION AND N EXT S TEPS ................................ ................................ ................................ .................. 79 V TESTING A NO VEL PROSTHETIC WRIST THAT INCORPORATES T TION ...... 81 5.1 M ETHODS : D EVELOPMENT ................................ ................................ ................................ ....................... 82 5.1.1 Development of the DTM Wrist ................................ ................................ ................................ 82 5.1.2 Hand Modifications ................................ ................................ ................................ ................... 87 5.1.3 Functional Characterist ics of the Prosthetic Wrist and Hand ................................ .................... 88 5.2 M ETHODS : E XPERIMENT ................................ ................................ ................................ .......................... 89 5.2.1 Population ................................ ................................ ................................ ................................ 90 5.2.2 Inclusion Criteria ................................ ................................ ................................ ........................ 91 5.2.3 EMG and Prosthesis Fitt ing ................................ ................................ ................................ ....... 92 5.2.4 Myoelectric Control ................................ ................................ ................................ ................... 95 5.2.5 Limited SHAP ................................ ................................ ................................ ............................. 99 5.2.6 Compensatory Motions ................................ ................................ ................................ ........... 100 5.2.7 NASA TLX ................................ ................................ ................................ ................................ 102 5.3 R ESULTS ................................ ................................ ................................ ................................ ............. 102 5.3.1 SHAP Results ................................ ................................ ................................ ............................ 103 5.3.2 Shoulder Results ................................ ................................ ................................ ...................... 104

PAGE 11

xi 5.3.3 NASA TLX Results ................................ ................................ ................................ ..................... 106 5.4 D ISCUSSION ................................ ................................ ................................ ................................ ........ 107 5.4.1 SHAP ................................ ................................ ................................ ................................ ........ 108 5.4.2 Shoulder ................................ ................................ ................................ ................................ ... 109 5.4.3 NASA TLX ................................ ................................ ................................ ................................ 111 5.4.4 Corrected Angle ................................ ................................ ................................ ....................... 111 5.5 C ONCLUSION ................................ ................................ ................................ ................................ ....... 112 VI CONCLUSIONS AND FUT URE WORK ................................ ................................ .............................. 114 6.1 C ONCLUSIONS ................................ ................................ ................................ ................................ ...... 114 6.1.1 Survey ................................ ................................ ................................ ................................ ...... 114 6.1.2 DTM angle ................................ ................................ ................................ ............................... 115 6.1.3 DTM Prosthetic Wrist ................................ ................................ ................................ .............. 115 6.2 F UTURE WORK ................................ ................................ ................................ ................................ ..... 116 6.2.1 Rotator vs DTM ................................ ................................ ................................ ........................ 116 6.2.2 Correction of the DTM Angles ................................ ................................ ................................ 117 6.2.3 Assessment of Compensatory Motions ................................ ................................ ................... 117 6.2.4 Combination of Wrist and Hand Position in Postural Controller ................................ ............. 119 6.3 F INAL T HOUGHTS ................................ ................................ ................................ ................................ 120 REFERENCES 122

PAGE 12

xii L I S T OF F IGURES Figure 2.1: Homunculus drawing by Penfield showing that neurons dir ecting the control of the hand dominate the motor cortex (34). ................................ ................................ ....... 9 Figure 2.2: Bones of the hand and wrist. ................................ ................................ ................ 10 Figure 2.3: Range of motion of the anatomical wrist ((14), modified). ................................ .. 11 Figure 2.4: An example of The Dart Thrower's Motion. ................................ ......................... 12 Figure 2.5: Radial plo t of the average maximum passive range of motion of 6 cadaver wrists as described by Crisco. The DTM plane is visible as a plane rotated 20 40 degrees from the plane of pure flexion extension (20). ................................ ................................ ....... 13 Figure 2.6: Passive wrists. Terminal devices may be attached to these devices and passively positioned to achieve flexion. From left to right: Hosmer constant friction wrist and Hosmer flexion wrist (Hosmer, Campbell, CA), Texas Assistive Devices N Abler wrist, Medical Bionics Robo Wrist, and TRS passive flexible wrist. ................................ .......... 21 Figure 2.7: Examples of powered wrists. The left two devices (Otto Bock wrist rotator, VASI children's electric wris t rotator) only provide rotation. The right two devices (Centri Hand, Shanghai hand) combine rotation and flexion extension. ................................ ... 21 Figure 2.8: The Defense Advanced Research Projects Agency (DARPA) RP2009 APL Modular Prosthetic Limb (MPL) Arm developed as part of the DARPA RP2009 initiative includes a multi DOF wrist. ................................ ................................ ................................ .............. 22 Figure 2.9: Diagram of myoelectric control for a powered prosthesis The neural signal from the brain innervates the muscle in the residual limb, changing the surrounding electric

PAGE 13

xiii potential and causing it to contract. This change is detected by electromyogram electrodes on the surface of the arm, electronically amplified and processed, and sent as a control signal to a motor which moves the prosthesis. ................................ .......... 24 Figure 3.1: Power analysis indicating that 21 participants were sufficient to detect a 10% difference in re sponses with 97% confidence. ................................ ............................... 36 Figure 3.2: Pie Chart showing the how often participants made each of the keyed comments during their interview. Overall, nearly one third of comments were about difficulty using tools or completing other ADL tasks. Another third addressed common complaints with prosthetics (e.g. durability, socket, cost, weight, and electronics). .... 44 Figure 4.1: Attemp ts at overcoming the Optotrak limitations. Rigid plates (above) were attached to body segments which correctly located the segment but could only be seen from a fixed vantage point. Multiple markers attached to each segment and defined as elow). Lighter lines and circles indicate connections and markers on the opposite side of the limb. This method was also ineffective because the Optotrak software was unable to account for soft tissue artefacts (i.e. minor motion between markers from finge r abduction or forearm skin stretch) in determining the location of the rigid body. ................................ ................................ ................................ ................. 51 Figure 4.2: Circuit diagram for the flex sensitive resistor setup. V in and V out were generated and recorde d in Labview. ................................ ................................ ................................ 53

PAGE 14

xiv Figure 4.3: Angle vs Voltage in flex sensitive resistor (FSR) testing. FSR was attached to a plastic goniometer and the voltage was recorded via Labview at various angles and m oving in the forward (black) and backward (white) directions. ................................ .. 54 Figure 4.4: Mockup of the FSRs attached to the dorsal and ulnar sides of the wrist. ........... 54 Figure 4.5: Plot of the wrist angles flexion extension (FE) and radial ulnar deviation (RU) recorded with the FSR system during simulated dart throwing. ................................ .... 55 Figure 4.6: Res ults of calibration test of the goniometers and torsion meter over 180 degree operating range. Hysteresis was found to be <2.5 degrees over 180 degrees. Relationship between output voltage and angle is highly linear. ................................ .. 56 Figure 4.7: SHAP design showing the abstract objects (ball, cylinder, etc.) and simulated ADL objects (cup pouring, key turning, etc.). Participants perform timed movements of the objects from which the assessor can calculate the SHAP score (89). ............................. 57 Figure 4.8: Frequency spectrum of sample wrist motion showing that the majority of information is below 10 Hz. This justifies the use of a filter with a 10 Hz cutoff freque ncy. ................................ ................................ ................................ ....................... 67 Figure 4.9: Demonstration of the phases of a SHAP task during a single trial. The initial rest phase results from the need to start recording before the participant begins the task. The reach phase typically involves the greatest motion as the hand is positioned at the object. The Grasp phase is typically characterized by the least motion and determined by recordings of the participant only holding the object. The final rest phase begins

PAGE 15

xv once the object is released and continues beyond the end of the task until the recording can be manually stopped. ................................ ................................ .............. 68 Figure 4.10: Motion of the wrist during the reach phase (top) and grasp pha se (bottom) of each task in the SHAP averaged across participants. Error bars indicate standard error. For each task, a distinctive motion from the rest position to the beginning of the grasp position is visible. During the reach phase, there is a visible coupling between FE and PS. During the grasp phase, many of the tasks have similar motions while the more complex tasks (coins, buttons, pouring) have distinctive profiles. During grips, FE and RU are commonly correlated. Simulated dart throwing (DTM) was included as a comparison. ................................ ................................ ................................ .................... 72 Figure 22: 2 dimensional radial plot in the FERU plane on the averaged wrist data. Red and green lines represent the first and second principal component. Their len gths represent the percent contribution of each PC. ................................ ............................. 74 Figure 4.12: Two dimensional radial plots of wrist angle in the three planes during the reach phase (left) and grasp phase (right). Red an d green lines represent the first and second principle components, respectively. Lengths indicate the percent contribution to the variance of each PC. Boxed plots show the largest deviation in each of the phases. .... 75 Figure 4.13: Coupling in the three degrees of freedom of the wrist. Error bars indicate one standard deviation from the mean. We found significant coupling in the reach phases of FERU (DTM), RUPS PSFE+/ and in all three degrees o f freedom. Greater coupling in the grasp phase was found in all but the FERU+ (Anti DTM) mode. .......................... 76

PAGE 16

xvi Figure 4.14: Percentage of SHAP task grasp phases in given wrist orientation (F flexion, E exte nsion, R radial deviation, U ulnar deviation, P pronation, S supination). During the reach phase, wrist position is present in all combinations with FUP the most common position. A plurality of grasps takes place with the wrist in flexion, ulnar deviation, and pronation or in extension, radial deviation, and supination. This presentation is consistent with the DTM. ................................ ................................ ................................ 77 Figure 5.1: The transmetacarpal hand used as the starting point for the DTM wr ist flexion device used in this study (99). The fingers were removed and replaced with an adjustable flexion part and the gears replaced with a commercial gear box. ............... 84 Figure 5.2: Design of the DTM wrist from CAD (left), to plastic 3D printed (middle), to a machined device (right). ................................ ................................ ................................ 84 Figure 5.3: CAD view of the final DTM wrist (left) and exploded view (right) created for this study. Custom parts for this study are colored: Flexion Part (red), Adjustable Adapter Plate (green), Securing Annulus (yellow), and Wedge (teal). With the wedge removed and the adapter plate rotated to 0, this wrist functions as a standard flexion wrist. .. 85 .......... 86 Figure 5.5: Modified transmetacarpal hand used in this study. Cotton padding and rubber gloves were added to the fingers to improve grip. The wedge and adapter plate of the DTM wrist can be seen at the bottom. ................................ ................................ ........... 88

PAGE 17

xvii Figure 5.6: Process for sim ulating an amputation in an intact participant. Attach electrodes and EMG sensors (Top left), apply Coban (top right), apply padding (bottom left), splint with prosthesis (bottom right). ................................ ................................ ....................... 94 Fig ure 5.7: Process for building a custom myoelectric socket for a participant with an amputation. Bare residual limb (top left), apply electrodes and EMG sensors (top right), wrap in cotton and fiberglass casting material (bottom Left), apply second layer of fiberglass cast to secure the prosthesis to the inner cast (bottom right). ..................... 94 Figure 5.8: Illustration of the hybrid velocity position control system. Top left: raw EMG, Top right: thresholded EMG, Bottom: hand and wrist positions. Participant activates flexors (red) to flex the wrist, co contracts for the required time to switch to hand control, activates flexors which now open the hand, moves to the target (gap in top right), and then activat es extensors (white) to open the hand. The resultant movements can be seen in the bottom plots. ................................ ................................ ................................ 96 Figure 5.9: Co contraction display showing the controlled variables (threshold, refractory tim e, and co contract time). Participant has held a co contraction for 6 out of 10 required bins but missed one in the process, probably by dropping below threshold. Each green bin represents 10ms of co contraction time. Participants could see this display du ring the training session but not while performing the SHAP. ....................... 96 Figure 5.10: Index of function scores showing significant differences between the No Wrist condition and all other conditions and between the DTM wrist and the FE wrist (black). There was significant difference between the FE and DTM wrists (red). The score for

PAGE 18

xviii the DTM wrist used by people with transradial amputations was not significantly different than the DTM wrist used by people with simulated amputations (blue). Error bars represent standard error. ................................ ................................ ..................... 104 Figure 5.11: Comparison of shoulder ab/adduction ROM during the SHAP tasks. No significant differences were found be tween the different prosthetic wrist conditions. Error bars represent standard error. ................................ ................................ ............ 106 Figure 5.12: Compensatory motions employed by participants during the SHAP. A wide variety of compensat ory motions were used, including: large abduction (left), shoulder lift (middle), and torso bending (right). ................................ ................................ ........ 105 Figure 5.13: Comparison of NASA TLX scores. Only the No Wrist and AMP (DTM) cond itions were significantly different. Higher score indicates greater difficulty. ........................ 107

PAGE 19

xix L IST OF A BBREVIATIONS ADL Activity o f Daily Living BP Body Powered COMIRB Colorado Multiple Institution Review Board DAQ Data A cquisition Board DASH Disabilities o f The Arm Shoulder a nd Hand O utcome M easure DOF Degree Of Freedom DTM Dart Thrower's Motion FE Wrist Flexion Extension FSR Flexion Sensitive Resistor IFS Index Of Function Score IR Infrared LED L ight Emitting Diode ME Myoelectri c OPUS Orthotics And P rosthetics U ser S urvey PC Principal Component PCA Principal Component Analysis PS Wrist Pronation Supination RU Wrist Radial Ulnar Deviation SHAP Southampton Hand Assessment Proc edure SL Scaffoid Lunate Joint

PAGE 20

1 C HAPTER I M OTIVATION AND R ESEARCH G OALS There are an estimated 41,000 Americans currently living with major upper limb loss (proximal to the hand) and this number is expected to more than double in the next 30 years (1) Over three fourths of upper limb amputations are the result of traumatic injury and the majority of people with upper limb amputations are under the age of 65 (1,2) This would suggest an active and adaptable population, but according to the World Health ICF), this population ofte n feels severely limited in their daily lives and report difficulty participating in social activities, finding fulfilling employment, and living independently (3) This is consistent with older studies such as the 1998 Harris survey which found that people with dis abilities have a difficult time reintegrating into the workforce and with social interactions after an injury (4) It is also reflected in US employment rates: recent census data show that a smaller percentage of people with disabilities are in the work force than people without disabilities (5 7) Furthermore, as many as 69% of people with upper limb amputations become unemployed or have to change their career after an injury (5 8) A prosthetic arm enables people with upper limb amputa tions to be better able to perform activities of daily living (ADL), such as personal hygiene and food preparation, and gain employment in positions that require specific motions (2) Body powered prosthetics that use a cable to actuate a hook or simple hand are fairly limited, but modern myoelectric powered prosthetics can pick up muscle signals from the residual limb and translate that into multiple motions (9 11) Many of these ne w devices feature multiple degrees of

PAGE 21

2 freedom (DOF) in the hand and wrist in an attempt to mimic the anatomical arm (12) However, the control systems for these devices lack the ability to translate the myoelectric signals into complex movem ents. This lack of control means that prosthetic devices often upper limb amputations reject their prosthesis (13) A great deal of prosthetics research is focused on solving thi s limitation by increasing the level of control by adding more surface electrodes, implanting electrodes into individual muscles, or designing control schemes to translate patterns of signals into specific motions (14 18) These soluti ons are promising, but have not yet had great commercial success The research presented in this dissertation instead focused on simplifying the mechanical properties of the prosthetic wrist with the goal of improving functionality while avoiding complicat ing the control system. This was done by d etermining which DOF are most important for performing ADLs and including those while removing the others This research was inspired by the work done by Palmer who, in 1985, showed that the motion of the wrist is offset from the anatomical flexion extension plane in what is known (19) More recently, Wolfe and Crisc o found that the DTM is an important DOF to preserve wrist function when surgically fusing bones of the wrist (20,21) The specific goal of this research was to determine whether a simple prosthetic wrist using the DTM could accomplish most of the function of an anatomical wrist.

PAGE 22

3 1.1 Summary of Goals The pur pose of this research was to test the hypothesis that people with upper limb amputations face particular difficulty with ADL tasks that require wrist mobility due to the limited degrees of freedom found in most existing prosthetic wrists; and that limitati on could be alleviated by a prosthetic wrist that incorporates the DTM controlled with a simple myoelectric system. To this end, this dissertation tested three specific aims: 1.1.1 Specific Aim 1: Identify what ADL tasks people with amputations find challenging via online survey and determine what improvements they are interested in seeing in their devices. It is most important to identify the needs of a population before designing any new device for them. Performing human factors testing with people with upper limb amputations and integrating their input into the design from the beginning improves the chances of making a successful device. Several studies have surveyed people about their relationship with their prosthesis and found that they often have difficul ty performing ADL, especially those that involve tool use (8,22,23) Howe ver, none of these surveys specifically inquired about tasks that utilize the DTM. Twenty three people completed a novel, online survey developed for this study to determine the levels of difficulty, satisfaction, and importance of performing DTM and non D TM tasks with their prosthesis. The survey was based on existing surveys (24,25) and supported by a team of experts in the field of human factors testing and reha bilitation. A five person subset of the people surveyed were also interviewed to clarify their responses and add additional feedback. The survey and

PAGE 23

4 interviews indicated that people were less satisfied performing DTM tasks with their prosthesis than non DT M tasks and that people were interested in having a more fu nctional wrist in their device. This motivated the precise measurement of the DTM and development of a prosthetic wrist that incorporated the DTM. 1.1.2 Specific Aim 2: Motion in the Context of a Prosthesis Specific Functional Test The goal of creating a simple prosthetic wrist was informed by the results of specific aim 1. In order to design a wrist that incorporates the DTM it was necessary to first precisely identify the angle of the DTM. Previous studies on the DTM have described a range of angles rotated from pure flexion extension (20,26,27) Many of these studies have limited application to prosthetic design because they used cadaver arms, or examined full range of motion, or measured non standardized motions. To create a clinically relevant measure of the DTM, fiftee n intact participants performed the Southampton Hand Assessment Protocol (SHAP) while wearing goniometers to record the angle of the wrist (28) The SHAP is the current standard of care measurement of prosthetic function and involves moving several abstract objects (e.g. ball, cylinder, tab) and simulated ADL objects (e.g. turning a key, picking up coins, pouring water from a cup). The results indicated that a prosthetic wrist could mimic the DTM by combining flexion extension and radial ulnar deviation at an angle of 22 degrees during the grasp phase.

PAGE 24

5 1.1.3 Specific Aim 3: Motion A novel prosthetic wrist incorporating t he results from specific aim 2 was designed and tested. The wrist was based on an existing wrist flexion unit rotated to the DTM angle. The prosthesis was controlled by a simple modified position control myoelectric system. The goal of this study was to d etermine whether the DTM wrist would perform better than a flexion wrist and would still be as easy or easier to use than a prosthesis without a wrist. To test this hypothesis, a group of 10 able bodied participants performed the SHAP using this device wi th a splint to simulate an amputation. Each person in this group completed the SHAP using the DTM wrist, a flexion wrist, and a fixed wrist (no wrist) The results from this group were compared to three able bodied people performing the SHAP with their int act limbs and to five people with upper limb amputations using a prosthesis with the DTM wrist. Improvement in function was measured by the SHAP Index of Function Score (related to time to completion) and shoulder compensatory motion (shoulder angle measur ed with goniometers as in aim 2). Ease of use was measured by the NASA Task Load Index (NASA TLX), a subjective questionnaire filled out at the end of each session. The results of this study indicated that the DTM wrist performed better than the flexion w rist on the SHAP although both active wrist conditions were slower than the no wrist condition. The shoulder data measuring compensatory motions did not indicate a reduction in ab/adduction compensatory motion but a range of other unmeasured compensatory m otions were observed during the test The NASA TLX indicated that neither the DTM wrist nor the flexion wrist were more difficult to control than a device without a wrist. This is a

PAGE 25

6 strong indica tion that the DTM wrist improve s prosthetic wrist function wi thout being too difficult to control

PAGE 26

7 C HAPTER II B ACKGROUND This chapter contains the background necessary to unde rstand the challenges facing anyone who wants to design a better prosthetic device for people with upper limb amputations (ULA) This include s: the physiology of the intact and amputated upper limb, specifically focusing on the function of the anatomical wrist; a discussion of electromyogram signals from the muscles and how they can be used to control powered prosthetic devices; and an overview of existing upper limb prosthetic devices, particularly commercial and research grade powered wrist devices. This research was focused on the mechanics of the anatomical wrist and design of a new prosthetic wrist. Therefore, a majority of this space is de voted to describing those systems. 2.1 Physiology of the Upper Limb The human arm evolved over millions of years to perform both dexterous and gross movements using 36 muscles in the hand and forearm, 27 bones, and 18 joints, resulting in 27 possible motions, or degrees of freedom (DOF) in the hand and wrist and an additional 7 DOF in the arm (11) The hand alone contains 22 DOF while the wrist has three functional DOF arising from movement of the eight metacarpal bones and the relative motion of the radius and ulna. These motions are primarily actuated by muscles in the proximal forearm. They allow the arm to perform a wide array of motions from large, course movements, such as swinging and grasping, to precise positioning and fine motor control (29)

PAGE 27

8 The upper limb consists of the shoulder joint, upper arm, elbow, forearm (consisting of the radius and ulna bones), wrist (made up of the me tacarpal bones), and the hand. The role of the proximal arm is primarily to produce course movements for positioning and power while the distal components of the arm are typically responsible for fine control and precision It has been understood for over a century that a great amount of brain power is devoted to control ling the arm and hand (30) The function of the upper limb is the result of mental planning and neural signaling, proprioceptive and sensory feedback loops, and the biomechanics of the muscles, tendons, and bones (31) These actions must be timed and coordinated precisely to interact with the world. Penfield famously illustrated the outsized proportion of the motor cortex devoted to the hand in his homunculus drawing ( Figure 2.1 ) (32) The evolution of the hand and motor cortex in early hominids likely developed simultaneously and set humans apart from other primates (33)

PAGE 28

9 Figure 2 1 : Homunculus drawing by Penfield showing that neurons directing the control of the hand dominate the motor cortex (34) 2.1.1 The Wrist The w rist is used to alter the angle of attack of the hand to allow efficient grasping (35) Without a functional wrist, people would need to make large compensatory motions with the upper arm to position the hand and perform ADLs. There are redundancies in the muscles that activate the wrist but the primary activators are described below. The first two DOF of the wrist, flexion extension (FE) and radial ulnar deviation (RU), are functions of the movement of the carpal bones relative to the radius and ulna ( Figure 2.2 ) These are activated by several muscles in the forearm. Specifically: the flexors carpi ra dialis and ulnaris (flexion), extensor s carpi radialis longus and brevis (extension), flexor carpi radialis (radial deviation), and flexor carpi ulnaris (ulnar deviation). The third DOF of the wrist, rotation or pronation/supination (PS), takes place in th e forearm as result of the

PAGE 29

10 radius and ulna twisting about each other. This motion is activated by the pronators teres and quadratus (pronation) and supinator (supination) (29) Figure 2 2 : Bones of the hand and wrist 1 The functional motio ns of the wrist allow up to 135 degrees of FE, 40 degrees of RU, and 180 degrees of PS ( Figure 2.3 ) (14) Most tasks, however, only utilize about 40 degrees FE and less than 30 degrees RU (19) Additional flexibility and compliance in the wrist is provided by movement of the carpal bones against each other. 1 www. anatomy diagram.info

PAGE 30

11 Figure 2 3 : Range of motion of the anatomical wrist ( (14) modified) Wrist motion is important for performing many activities of daily living (ADLs). Palmer measured the range of motion (ROM) of wrist F E and R U of normal participants performing a series of ADLs ranging from personal hygiene and food preparation to tool use and secretarial work (19) He found that m any ADLs especially tasks that involve tools (e.g. carpentry, culinary, and some personal hygiene) require a signif icant amount of wrist FE and RU. The wrist is also valuable for positioning the hand at the midline for ADL tasks such as feeding, cooking, and grooming. 2.1.2 The a combin ation of wrist FE and RU in in an oblique plan e rotated between 20 and 40 degrees from pure FE (20) This motion has becom As the name suggests, the DTM is the motion of the hand and wrist travel from an extended and radial ly deviated position to a flexed and ulnar deviated position when a dart is thrown ( Figure 2.4 ). It is estimated that more than half of ADLs make use of this motion (36)

PAGE 31

12 Figure 2 4 : An example of T he Dart Thrower's Motion 2 The combination of these two DOF allows for efficient positioning of the hand for prehension. In intact wrists, the DTM is a motion guided by the distal s urface of the scaphoid bone and minimizes scaphoid lunate and spheroid lunate motion ( Figure 2.2 ) (37) This mo tion allows for the greatest range of wrist circumduction while minimizing required muscle activity and maximizing muscle force output (2 7,38) The DTM is particularly important for the power stroke of tool use and throwing. S tudies have suggested that the development of the bone and ligament structure s involved in this motion were one of the major evolutionary advantages that set humans a part from other primates (33,39,40) Several studies have attempted to identify the specific angle of rotation for the DTM (19,27,41,42) These methods have measured passive and active range of motion, live par ticipant s and cadavers, and used goniometers, MRI, and CT scans to try to pin down the DTM angle. One of the best studies was done by Crisco in 2011 measuring range of motion on cadaver arms mounted to a custom jig ( Figure 2.5 ) This showed the passive range of motion of the wrist to be an envelope rotated about 30 degrees from pure FE. 2 www.nicedarts.com

PAGE 32

13 Figure 2 5 : Radial plot of the average maximum passive range of motion of 6 cadav er wri sts as described by Crisco. The DTM plane is visible as a plane rotated 20 40 degrees from the plane of pure flexion extension (20) The concept of the DTM is currently in use in the medical field among hand surgeons. Some diseases of the wrist, including severe arthritis and tumors, require surgeons to fuse some or all of t he carpal bones. When these procedures are performed, surgeons often are forced to choose a single DOF that will minimize motion while preserving function. Scott Wolfe, a physician and one of the leading researchers on the DTM, advocates that when wrist mo bility must be limited the DTM should be preserved to retain the majority of functionality (21,23) The designer of a pros thetic wrist faces a similar choic e as the surgeons: the DOF in a prosthetic wrist are limited by weight power, cost, and ease of use. Therefore, it is prudent to find a motion that maximizes functionality while limiting complexity. The DTM is an excellent candidate for such a motion.

PAGE 33

14 2.2 The Amputated Upper Limb The loss of a major part of the upper limb can have profound social and economic effects on a person. Studies have shown that people with upper limb amputations (ULA) often feel severely limited in physical and social activities and in finding and keeping jobs and represent a smaller po rtion of the work force than people without disabilities (3,5 7) They also tend to have a difficult time reintegrating into the workforce and find social interactions challenging after their injury (4) People can lose limbs for many reasons but, ultimately, they are left with a residual limb and a need to figure out how to i nteract with the world in a novel way. 2.2.1 Etiology One of the best estimates of the number people with amputations found that there were 41,000 people in America with major ULA (proximal to hand ) in 2008 (1) That number is expected to more than double by 2050. Trauma accounts for between 62 and 90% of all major ULA (1,22,43,44) Most of these traumatic amputations are the result of farm and industrial accidents. Operati on Iraqi Freedom/Operation Endu ring Freedom brought limb amputations to the forefront of the public percepti on although according to the most recent pentagon report, these conflict s only account for 737 American servicemembers suffering ULA (22) Of acquired ULA, malignancy may account for up to 24% and dysvascular diseases account for 3% (43) Congenital limb deficiency, which can be cau sed by many factors, occurs in about 15 out of every 100,000 live births. While the risk of acquired ULA increases

PAGE 34

15 with age, the majority of people with ULA are between the age of 20 and 40 and generally considered active (44) The most common major ULA and congenital limb deficiency is trans radial : between the wrist a nd elbow. 2.2.2 Anatomy and Surgery of the Upper Limb Regardless of the cause of the amputation, the patient is left with a less functional residual limb. This can make interacting with the world challenging, but it is by no means impossible. Indeed, many peopl e with amputations are able to be fully active and do not need a prosthesis. In many cases, the residual limb will retain functional musculature, nerves, and vasculature. With care and training, the residual limb can remain useful for performing ADLs by it self or by controlling a prosthesis. Surgical procedures for amputation vary by patient and underlying cause. Surgeons may perform a disarticulation at a joint or osteotomy across the bone. In either case they usually attempt to preserve as much length in the residual limb as possible (43,45) This maximizes the torque that can be applied by the limb and allows for better prosthesis attachment. Efforts may also be made to preserve prominent bony features o r long bone length in order to accommodate a socket for a prosthesis. Other considerations include techniques for reconnecting the musculature and vasculature to keep the limb healthy and functional. Neurons must also be managed by cutting them short or em bedding them in other muscles to avoid painful neuromas or phantom limb sensations. There are other surgical techniques that seek to accommodate a prosthesis after amputation. Targeted muscle re enervation can attach nerves to new muscles so they can

PAGE 35

16 be us ed for controlling a prosthesis via electromyography (described below) (45,46) Tunnel cineplasty creates an external loop of muscle that can be attached to a cable that acti vates a prosthesis (14) Osseointegration is a technique where a met al rod is attached to the bone in the residual limb and extends through the skin where it can be attached to a prosthesis (47) Each of these techniques attempts to improve the outcome of the amputation surger y and provide the patien t with options for prostheses. 2.2.3 Considerations for People with Upper Limb Amputations Without the intrinsic, neuro muscular, system of the body, people with ULA lack the functionality to perform many tasks (22,23) People are often able to overcome this functional deficiency through compensatory motions such as major shoulder abduction or torso bending (48,49) However, these motions can lead to overuse and repetitive motion inj uries, poor body mechanics, and social challenges (50 52) There is limited information about the social and economic consequences of acquiring a ULA. One of th e few studies to specifically address this population found that a s many as 69% of people with ULAs become unemployed or have to change work places after their injury (8) More generally, p eople with disabilities often have a difficult time reintegrating into t he workforce and interacting in social situations (4,7) This is reflected in US employment rates in the most recently available census data which shows that a smaller percentage of people with disabilities are in the work force than peop le without disabilities (5 7) This can have a severe negative impact on quality of life (53)

PAGE 36

17 This is not to say that people with ULAs are unable to live fulfilling lives. Prosthetic devices have come a long way in the last century and are able to provide a range of functions to people. Even without prosthetic devices, many people with ULAs are capable of extraordinary achievements. Some are even world champion rock climbers 3 2.3 Current P rosthetics For many people with an ULA a prosthesis will provide much of the function of an anatomical limb Howeve r, designing functional and aesthetically pleasing prostheses for people with ULA has historically proven to be a challenge (54) An ideal prosthetic arm would mimic all the functions of an intact human arm, but each additional feature, point of control, DOF, and increase in force, adds to the volume, mass, and complexity of the device (11) The design of current prosthetic terminal devices is driven by minimizing these factors and limiting how difficult it is to control. From a user perspective, a device must be simple to use, comfortable, and functional in order for it to be accepted (55) Current devices fail these criteria, as evidenced by the persisting high rejection rates among persons with unilateral amputations (38%, 3 0%), particularly among people with wrist disarticulation amputations (94%, 80%) (22,56) Overall, fewer than a third of people with ULA report being satisfied with their prosthesis (8) Designing an overly complex device will not alleviate the problem of device 3 http://www.para doxsports.org/paradox ambassador spotlight maureen beck/

PAGE 37

18 abandonment and an overly simple device will not help people with amputations be more social or more able to perform ADLs. 2.3.1 Types of Prosthetics The commercially available prosthetic device options can be divided into three groups: tic devices (e.g. cosmetic hand from RSL Steeper (57) ) are typically made of a colored silicone or PVC glove over a mo u lded foam hand. While these devices provide min imal functionality, they provide a high level of social acceptance which is important for many people with amputations (58) Body powered devices typically utilize a Bowden cable running from the device to an intact part of the body which directly or indirectly articulates a single DOF in the device. These devices are highly functional, well liked and extensively u sed by peopl e with amputations. They have the advantage of being durable, inexpensive, easy to use, and give a modicum of sensory feedback. However, there have only been a few recent advancements in body powered devices These include the 3d printed e NABLE hands (59) and the sport specific terminal devices made by TRS (60) Otherwise, the basic design of body powered devices has hardly changed since the 1950s and th ese devices are essentially at the limit of their potential (11,61) The state of the art in upper limb prosthetics is powered myoelectric devices. These devices incorporate surface electromyographic (EMG) electrodes to record electric signals from contracting muscles in the residual limb and translate them into motor mo vement Myoelectric devices have the potential to provide an intuitive control scheme for users with

PAGE 38

19 minimal delays between thought and movement (17) However, surface electrodes lack specificity i.e. EMG electrodes p laced on the volar forearm pick up signals from all of the flexor muscles while electrodes on the dorsal forearm pick up all signals from extensor muscles rather than recording from individual extensor or flexor muscles (10) This typically limits these devices to being controlled by only one or two one agonist/antagonist pairs. With only one or two control sites, switching mechanisms are necessary to control multiple degrees of freedom (17) Adding DOF can quickly make the device too difficult to use This lack of control inputs constitutes a major challenge to full functionality for myoelectric prosthetics. Although the mechanics of prosthetic devices can mimic many of the motions of the human hand there is not yet a way to effectively control them (35) The low number of control sites in powered myoelectric devices necessitates that only the most valuable DOF be incorporated in prosthet ic devices. Commercially, this means that a prosthesis has at most a gripping device (e.g. hand or split hook) and a wrist unit (almost always a rotator) (62) It is not clear whether rotation is the best DOF to include (50) The myoelectric system will be discussed in greater depth in section 2.4 2.3.2 Prosthetic Wrists Articulated wrist devices improve positioning and reduce compensatory movements in many ADL tasks, such as food preparation and personal hygiene (50) Users score higher on functional tests when flexion or deviation capabilities are added to the wrist (23,36,63,64) In one case, a majority of users who were fitted with a passive, locking F E wrist (Multi Flex Wrist) report that having F E in their device improved the number of activities they were

PAGE 39

20 able to perform with their device and made activities feel more natural (36,65) However, c ommercially available prosthetic devices do not provide full functionality and adding any additional DOF necessaril y complicates controlling a device (66) This added complexity may be acc eptable if i ncluding a functional wrist in a prosthesis improves the user experience and increases functionality (67) There are only a few commercially available options for wrist prosthetics Most commonly, terminal devices are simply fixed to the socket or attached via a constant friction device These hold the terminal device in a fixed position with a set screw or spacer washer A quick disconnect system allows users to quickly change terminal devices and set them at fixed PS angles (54) ( Figure 2.6 ). While some work has been done to determine the optimal fixed angle of the wrist, most commercial devices are simply set by the user to a neutral or convenient angle (68) There exist prosthetic wrists which utilize one two or three DOF. These include devices with passive motions such as the 5 Function N Abler wrists (Texas Assistive Devices, Brazoria, TX), t he Robo Wrist (Medical Bionics Inc. Spruce Grove, Alberta, Canada), and the TRS Flexible wrist (Therapeutic Recreational Systems, Boulder, CO). Each of these devices can be manually positioned to achieve flexion but this requires additional body movements and a series of locks or switches which most users find difficult to operate (36)

PAGE 40

21 Figure 2 6 : Passive wrists. Terminal devices may be attached to these devices and passively positioned to achieve flexion. From left to right: Hosmer constant friction wrist and Hosmer flexion wrist (Hosmer, Campbell, CA), Texa s Assistive Devices N Abler wrist, Medical Bionics Robo Wrist, and TRS passive flexible wrist. A few p owered wrist devices also exist. T he most common of these are the wrist rotators from Otto Bock and Motion Control (Motion Control Inc., Salt Lake City, U T), and the VASI Children's Electric Wrist Rotator (Liberating Technologies Inc, Holliston, MA) ( Figure 2.7 ). Motion Control makes a powered wrist rotato r with a passively positioned F E in their powered hand and prehen sor and will be releasing a powered FE device this year Figure 2 7 : Examples of powered wrists. The left two devices (Otto Bock wrist rotator, VASI children's electric wrist rotator) only provide rotation The right two devices (Centri Hand, Shanghai hand) combine rotation and flexion extension. The Centri Hand (Centri AB, Sollentuna, Sweden) uses a 4 bar linkage to combine fing er flexion with wrist extension. T he Shanghai Hand (Shanghai Kesheng Prostheses Co. Ltd, Shanghai, P.R. China) has powered flexion extension but does so in the anatomical F E plane a nd cannot be rotated to the DTM plane. T he Defense Advanced Research Projects Agency

PAGE 41

22 (DARPA) RP2009 APL Modular Prosthetic Limb (MPL) Arm and the DARPA R P2007 DEKA Arm ( Figure 2.8 ) have multi DOF wrists that included FE but cannot specifically mimic the DTM. Figure 2 8 : The Defense Advanced Research Projects Agency (DA RPA) RP2009 APL Modular Prosthetic Limb (MPL) Arm developed as part of the DARPA RP2009 initiative includes a multi DOF wrist 2.3.3 Limitations of C urrent P rosthetic D evices In 1985, Childress wrote "Adequate replacement of the human hand and arm is one of the most difficult pro blems facing medical technology (55) This sentiment is no less true today (69) Despite major advanc top of the line myoelectric devices still fall short of adequately replacing an anatomical limb. One of the limiting factors for people with ULA is the lack of simple and effective prosthetic devi ces (1) Too often, people must choose between overly simple, body powered devices or overly complicated powered myoelectric devices (70) While myoelectric devices hold the promise of emulating the intact human hand and wrist, they provide too few inputs to effectively control a complete limb. Many prosthetic arm users have difficu lty performing ADLs with their device. Opening bottles, using tools, and personal hygiene tasks are often listed as some of the most

PAGE 42

23 challenging tasks to perform with a prosthesis (8) F armers report their primary complaint is the inabili ty to use their device for work (71,72) These challenges often cause people with unilateral ULA to resign themselves to only using their prosthesis for bimanual tasks (73) It is not only the high Rates of rejection for body powered hooks have been measured to be as high as 85% (13) and passive hands have a rejection rate around 50% (74) This is often due to factors other than f unctionality; Cost, appearance, socket comfort, durability, and weight are major factors that can lead people to abandon their devices (13,58,74) Research is being done to improve these physical factors and the more functionality a device has, the more likely use rs will tolerate using it (11) 2.4 Myoelectric C o ntrol Myoelectric systems allow users to control their devices with (EMG) signals from muscles in the residual limb ( Figure 2.9 ). That is, w hen the contractile signal from the brain reaches the muscle, the electric potential around that muscle changes. That change in potential can be detected via EMG electrodes placed on the skin. This signal can then be processed in hardware or software in order to control the device (11) The muscles are used rather than trying to record directly from the neurons because the muscles produce a much higher voltage and the signal can be detected through the skin.

PAGE 43

24 Figure 2 9 : Diagram of myoelectric control for a powered prosthesis. The neural signal from the brain innervates the m uscle in the residual limb, changing the surrounding electric potential and causing it to contract. This change is detected by electromyogram electrodes on the surface of the arm, electronically amplified and processed, and sent as a control signal to a mo tor which moves the prosthesis 4 Processing the EMG signal involves first amplifying the signal and then filtering it (75) The signal is then full wave rectified and averaged. The majority of this proces sing is often done in hardware in the electrodes. This allows the signal to be recorded as a voltage by a data acquisition unit attached to a computer or built into a prosthesis (75) The processed EMG sig nal can be used to control one or more motors. Most myoelectric systems use the signal to directly control the motors by equating amplitude of the signal to the position of the prosthesis in space (position control) or by relating amplitude of the signal t o the velocity of the motor (velocity control) (10) It is also possible to create a system where combinations of EMG signals relate to specific postures (posture control) or algorithms detect features within the EMG signal (pattern recognition) so that a 4 www.Britannica.com

PAGE 44

25 varie ty of positions can be achieved without requiring a set of electrodes for each DOF (17,76)

PAGE 45

26 C HAPTER III U SER S URVEYS S UGGEST I NCORP ORATING THE DTM INTO A P ROSTHETIC W RIST W OULD B E U SEF UL Around thirty percent of people with upper limb amputations choose not to use a prosthetic device because these devices do not meet their needs (22,56) In order to reduce this rejection rate and improve user satisfaction, it is important to und erstand what functionality users are looking for in a device and why they feel unsatisfied (58) has the most import ant perspective, and various goals of the user will determine what the (4 p84 ) Withou t understanding the user perspective, any research risks making the mistake of creating a solution for a problem that does not exist. performing activities of daily li ving was related to limited wrist function in their prosthesis. Twenty three people with trans radial amputations completed the survey. Overall, respondents reported significantly lower satisfaction performing activities of daily living in tasks that utili they wanted a more functional wrist device in their prosthesis. A subset of five people who completed the survey volunteered for a follow up interview to discuss their responses in greater depth. They indicated the lack of ability to use tools and perform many activities of daily living as a major limitation in their device. They also indicated several physical characteristics (weight, cost, durability, and socket discomfort) as need ing improvement.

PAGE 46

27 The results from this study motivate further investigation into the mechanics of the wrist with the goal of improving wrist function. 3.1 Methods: Development 3.1.1 Human Factors Testing Human factors testing (survey questions and in person interv iews) have long been established as an effective means to determine the best product designs for users (78) One of the earliest and largest surveys of the needs of people with upper limb amputations was conducted by Atki ns et al. in 1996 (67) This study included responses from 2,477 one page surveys and 1,575 responses to follow up seven page surveys. They found that users of both body powered and electric devices wanted more functional wrists. They also found that most people used body powered devices instead of myoelectric ones for more rugged tasks such as manual labo r, construction, and farming. people face using their devices. However, in 2011, Ritchie analyzed fifteen studies that surveyed people with upper limb amputations and r eported a great deal of variation in how studies were conducted and what questions were asked (58) Furthermore, the most comm only used surveys (e.g. OPUS & DASH) only ask about difficulty and do not address whether tasks are important or if the user is satisfied performing those tasks (24,25,79) Despite nearly t

PAGE 47

28 has been slow to incorporate these findings and the few powered wrist devices on the market have not produced satisfactory results (12) 3.1.2 Survey and Interview Questi ons This study developed and administered an online survey and interview questions to determine user experiences performing ADLs. There are three important characteristics to successfully completing a task with a prosthesis: How satisfied the user is with performance (satisfaction), how important is it to be able to complete the task (importance); and, how easy is it to complete the task (difficulty). Each of these metrics is part of the overall user experience. It is necessary to select tasks that accurately represent the kinds of ADLs that a user encounters during the day. One of the challenges to creating a survey about ADL making compensatory mo tions (low difficulty, low satisfaction). These motions allow the user to accomplish the task but they may lead to overuse injuries and the motions are often socially awkward (48,49) Users may also simply avoid performing certain difficult tasks by using their intact limb or asking others for help (high difficulty, low importance) (71) In order to investigate the typical user experience and mitigate the effects of comp ensatory strategies, this survey included typical ADL tasks, tasks that are difficult to perform with compensatory motions, and two handed tasks where use of a prosthesis is required. The task list was influenced by two validated surveys of upper limb func tion: Disabilities of the Arm, Shoulder, and Hand survey (DASH) and Orthotics and Prosthetics

PAGE 48

29 Survey Upper Extremity Functional Status (OPUS UEFS) (25,79) However, neither of these surveys specifically asked about several tasks that incorporated the DTM or abou t satisfaction or importance. Therefore, a new list of tasks was created after discussion with a panel that included experts in prosthetics, rehabilitation, and assistive devices ( Table 3.1 ). Users responded to th e survey questions using a 5 point Likert like scale ( e.g. 1 not 5 minimize respondent fatigue (80) In addition, participants were also asked to rate their overall opinion of their device from 1 (do not like at all) to 5 (like a lot).

PAGE 49

30 Table 3 1 : List of activities of daily living (ADL) used in the online survey. Participants were asked to rate their level of satisfaction and difficulty performing these 30 tasks, and how important it is to them to be able to complete them on a 1 5 Li kert like scale. Tasks include DTM and non DTM ADL activities. Tasks were presented in random order to reduce the influence of respondent fatigue. DTM tasks that are hard to compensate for Non DTM tasks Turn a key Do/undo trouser buttons Tie shoes Push open a heavy door Do heavy household chores 2 handed DTM tasks Change a light bulb Use fork and knife Carry a heavy object Drain pasta in strainer Use mouse to navigate internet Put on a pullover sweater DTM tasks that are easily compensated for Easy Recreational activities Use a knife to cut food Dial a cell phone Write Name Take $1 from wallet Place an object on a shelf above your head Type sentence Garden or do yard work Fold laundry Difficult Recreational activities Transport tray Recre ational activities in which you move your arm freely Sweep floor Open bottle cap Put on socks/pants Open Jam jar Close eye drop/lipstick case Take pill from case Do/undo trouser zipper

PAGE 50

31 Table 3 2 : List of proposed additions to the prosthesis. Participants were asked to rate each idea by how useful they thought it would be to have in their device and how difficult they thought it would be to use. Ratings were based on a 5 point Likert like scale f rom 1 (not at all useful/difficult) to 5 (very useful/difficult). Proposed Improvement: Individual control of fingers/hook components (e.g. open an envelope/snap your fingers) Wrist with powered side to side motion (R/U deviation) Rotating thumb to al low pinch motion (Ab/adduction of thumb) Wrist with powered combination of F/E and R/U Sen sory feedback from fingers/hook Wrist with powered combination of F/E and R/U and rotation Faster finger/hook motion Fully functional wrist with rotation F/E and R/ U Stronger grip Passive wrist that can be positioned in anatomical positions Wrist with only powered rotation (pro/supination) Faster elbow Wrist with only up/down motion (flexion/extension) Elbow rotation (humeral rotation) Wrist with rotation and F/E 3.1.3 Survey Validation Validating a novel survey with a small population is a challenge. Nevertheless, the standard method for validation was followed as closely as possible (81 83) was first established by assembling a panel of experts to assess whether the list of tasks could accurately answer the research questions. This panel included Dr. Bodine (assisti ve technology), Dr. Weir (prosthetics), and Dr. Sullivan (physical medicine and rehabilitation) who found the questions to be appropriate. All questions were individually reviewed by the

PAGE 51

32 subject matter experts to make sure none were leading, confusing, or included other common errors. Table 3 3 : Cronbach's alpha for Satisfaction, Importance, and Difficulty calculated on total responses grouped by DTM and non DTM. The value of >0.90 indicates that questions in these groups were highly correlated and therefore gave consistent responses. Cronbach's Alpha Satisfaction Importance Difficulty DTM 0.96902 0.91263 0.95148 non DTM 0.97691 0.91894 0.96205 Due to the small population available for this convenienc e sample, an iterative pilot testing phase to determine if questions were relevant was not feasible. However, grouping the questions into DTM and non Alpha revealed that questions within each group wer e highly correlated ( Table 3.3 ) This iews of DTM and non DTM tasks. 3.1.4 Interview Follow up i nterviews were conducted to allow pa rticipants to give more in depth answers to survey questions and to allow the researchers to ask about any apparent discrepancies in the survey. During the interviews, participants were asked to describe situations where they had difficulty performing task s with their prosthesis or where their prosthesis excelled, to clarify their responses to the survey, and to describe what types of

PAGE 52

33 impro vements they would like to see ( Table 3.4 ) Additionally, participants were a sked whether they would be willing to use a more complex device if it had functionality. The use of open ended questions gave participants the freedom to volunteer information about their devices. This was particularly important in determining in what way s participants felt limited by their devices and identifying new ways their devices could be improved.

PAGE 53

34 Table 3 4 : List of questions asked of interview participants. General questions Specific to survey & DTM Describe the most difficult situ ation when using your device and how you managed. Several people in the survey, (including you), rated tasks as difficult and important, but also said they were satisfied with their device. Can you explain why? Generally speaking, are there particular tasks, or groups of tasks, you have difficulty performing with your device? Many people said that wrist rotation was not important, yet it is usually included in devices. What do you think? Are there ever any situations, or particular tasks or activities, when you simply choose not to use your device? The proposed improvement should specifically aid users with tool use (e.g. swinging a hammer). Do you have trouble using tools with your de vice? Describe an aspect of your device you like. Dislike? The proposed device would have an additional degree of freedom (angle it could move). This will require a more complicated control system. Would you be willing to use a more complicated devi ce if it provided you with greater functionality? How much better would a device have to be in order to justify the added complexity of the control system? What would it have to do for you? Describe changes you would like to see in your device. Wo uld you prefer a high or low tech solution to this problem? Open Discussion: How much would you be willing to pay for this improvement? like to add or that you have thought about in regard to your device, or the pro

PAGE 54

35 3.1.5 Power A nalysis An a priori power analysis indicated that twenty three participants were sufficient to detect a 10% difference between responses with 97% confidence (84) This was calculated by the equation ( 3 1 ) P is the desired p value (0.05) and v is the number of DOF (22, # participants 1) (84) A 10% difference was selected as a best guess at clinical relevance. This calculation was co nsistent with the recommendation in Rubin that suggests that a study of 23 participants is sufficient for representative results (78)

PAGE 55

36 Figure 3 1 : Power analysis indicati ng that 21 participant s were sufficient to detect a 10% difference in responses with 97% confidence. 3.2 Methods : Experiment A convenience sample of twenty three adults (18 85 years of age) with trans radial amputations were recruited. Participants completed the online survey described above using Survey Monkey or Google Forms. Demographic information ( name, gender, date of birth, profession, type of amputation, and time since amputation (if applicable)) was also collected. Participants were invited to return for a follow up interview. Those interviews were conducted and recorded over Skype or in person at The Assistive Technology Partners Product Testing Laboratory and recorded with a camera and microphone. Inclusion/exclusion criteria was selected to ensure that the sample could accurately represent the population of people with ULA and that it could address specific limitations of the prosthetic device. 3.2.1 Inclusion/ E xclusion C riteria To be eligible for this study, participants had to meet the following criter ia:

PAGE 56

37 Be be tween the ages of 18 and 85 Have a transradial amputation Be n on sedentary (>4 hours of activity per week) Have e xperience using at least one prosthetic device Have n ormal or corrected vision and hearing Can read and understand English at a cogn itively appropriate level The following criteria excluded people from taking part in this study: Age less than 18 or greater than 85 Have an a mputation at a location other than transradial Perform less than four hours of exercise per week Have n o experienc e using a prosthesis Be u nable to use a computer to answer the survey There were no exclusions based on race/ethnicity, gender, or socioeconomic status. Participants answered questions to ensure they met the inclusion/exclusion criteria before beginning t he survey. Responses were excluded from analysis if the participant indicated that they did not meet the inclusion/exclusion criteria. P eople with only trans radial amputations were surveyed in order to minimize complicating factors. While more proximal am putations also include the wrist and hand, the additional DOF of an articulated elbow, humerus, or shoulder would make it difficult to other factors. Likewise, people with distal amputations would have an anatomical wrist and not face the same limitations as someone with a prosthetic wrist. Modern prostheses are

PAGE 57

38 designed to be modular so even though this survey only questioned people with trans radial amputations, the r esults are applicable to all levels of amputation. Improvements for the trans radial level also apply to more proximal amputations. A non sedentary population was selected to reduce the likelihood that participants had physical limitations that would make it difficult to determine whether the survey results applied to issues with the wrist or because of disease or inactivity. This does not necessarily mean that the user is more proactive about their device, only that they make adequate use of their device. This study was primarily interested in improving the lives of people who use, or want to use, their devices for ADLs and so a non sedentary population was appropriate. Similarly, participants with visual or auditory impairments may have difficulties perfo rming the tasks that are unrelated to limitations to the wrist. 3.2.2 Recruitment Participants were recruited via email, postings at related organizations, and word of mouth. The following organizations were particularly helpful in recruiting participants: Para dox Sports (www.paradoxsports.org) Orthotics and Prosthetics International ( www.oandp.com/oandp l) National Amputee Golf Association (nagagolf.org)

PAGE 58

39 Flyers and emails included a link to the survey that participants could follow without making direct contac t with the research team. Participants who chose to email or call directly were directed to the survey link. The initial page of the survey included information about how the data would be handled and explained that the data would be de identified and stor ed on an encrypted, HIPAA compliant, password protected computer. Participants were not compensated for completing the survey or the interview. This study was conducted with the approval of the Colorado Institu tional Review Board (COMIRB #11 0674). 3.2.3 Analysi s The analysis of the survey responses was performed using Matlab (The MathWorks, Inc, test on the DTM tasks. T tests were also performed participants representing common population subgroups as described in chapter 2 Interview recordings were analyzed for keywords using Morae (Techsmith, Okemos, MI) ( Table 3.5 ) Keywords were reported by the percentage of total keyed comments. Comments about particular limitations or desired improvements were also reported. The interview process was valuable for clarifying any confusion that arose from the o nline survey results.

PAGE 59

40 Table 3 5 : List of keyed participant comments during interviews. A mark was made each time a participant mentioned each of these topics as well as a comment with further details. The list was generated by observing common topics in each interview. A (cosmetic) I (finger control) B (difficulty using tools) K (function over form) C (cost) L (Myoelectric is good at this) D (socket issues) N (weight) E (safety issue) R (ADL issues) F (ele ctronics issue) U (User wants improvement) G (durability) Y (BP is good at this) 3.3 Results Twenty three adult participants with trans radial limb deficiency (mean age 46.8 10.6) responded to the survey. Each respondent had experience with powered prosth eses, unpowered prostheses, or both (only powered = 5, only unpowered = 9, both = 6, no preference/no regular device = 3). The sample included both participants with congenital (n=3) and acquired (n=20) limb deficiencies. Respondents with acquired limb def iciency had used prosthetics for an average of 18 years (mean 18 17.8). Average age of participants with congenital deficiencies was 28.3 5.8 years. Two of the participants with congenital deficiencies used body powered devices and one used both. Overa ll, users reported that they found their devices to be somewhat easy to use (mean 3.71 on a scale of 1 = very difficult to 5 = very easy). There was no correlation between responses to satisfaction, importance, and difficulty.

PAGE 60

41 Table 3 6 : Significance values of one sample t test on the difference between the average response to DTM tasks and non DTM tasks. Highlights indicate significantly lower satisfaction, higher importance, or higher difficulty between DTM a nd non DTM tasks (p<0.05). Overall, the subset of users of body powered devices and those with acquired limb deficiencies reported lower satisfaction with DTM tasks than non DTM tasks. The acquired deficiency subgroup also reported higher difficulty perfor ming DTM tasks than non DTM tasks. The sum of participants in body powered and myoelectric devices is greater than 15 because some participants used both types of devices. One sample t test vs mean of 0 (p values) N Satisfaction Importance ifficulty All participants 23 0.017 0.399 0.068 Body powered 15 0.024 0.965 0.021 Myoelectric 12 0.086 0.198 0.454 Acquired 20 0.013 0.773 0.026 Congenital 3 0.884 0.269 0.510 The total participant population (n=23) and the subsets describing the m ost common populations (body powered device users (n=15) and acquired limb deficiency (n=20)) rated their satisfaction performing DTM tasks lower than their satisfaction performing non DTM tasks (one sample, 2 tailed t test: p=0.017, 0.024, 0.013) ( Table 3.6 ) Participants who used body powered devices and people with acquired amputations reported significantly higher difficulty performing DTM tasks than non DTM tasks (p=0.021, p=0.026). Intra population analysis r evealed that participants with acquired limb deficiency reported that it was significantly more important to them that they be able to perform all

PAGE 61

42 tasks with their prosthesis than participants with congenital limb deficiency (p=0.011). No other populations had significant differences in response. When asked about potential improvements to existing prosthesis, a majority of participants thought individual control of fingers, ab/adducting thumb, stronger grip, and ( Table 3.7 ) Seventeen participants felt that a fully functional wrist with flexion/extension, radial/ulnar

PAGE 62

43 Table 3 7 : User responses to the question: "Please rate how USEFUL you think each option would be". Leading suggestions were individual control of fingers, ab/adducting thumb, stronger grip, faster grip, and increased degrees of freedom in the wrist. Majority responses are highlighted in green. Suggested Improvement Number of Responses Not at all useful Slightly useful Somewhat useful Very Useful Extremely Useful Individual control of fingers/hook components 3 1 3 4 11 Rotating thumb to allow a pinch motion (Ab/adduction of thumb) 1 0 1 6 13 Sensory feedback from fingers/hook 4 2 3 2 11 Faster finger/hook motion 1 1 2 2 13 Stronger grip 2 0 2 4 15 Wrist with only powered rotation (Pro/Supination) 2 5 7 6 2 Wrist with only up/down motion (Flexion/Extension) 4 4 3 12 0 Wrist with rotation and Fl exion/Extension 2 2 0 6 13 Wrist with powered side to side motion (Radial/Ulnar deviation) 4 2 4 9 4 Fully functional wrist with rotation, Flexion/Extension, and Radial/Ulnar deviation 2 2 2 2 15 Passive wrist that can be positioned in anatomical positi ons 4 1 5 5 7 Five of the participants volunteered for a follow up interview. Four of these were conducted and recorded via Skype and one was conducted in person and recorded via camera and microphone. Of all the keyed comments in the interviews, the mos t common

PAGE 63

44 issues that participants reported having were: Completing ADL tasks (15.49%), Difficulty using tools (14.79%), Cosmetic (11.27%), Durabili ty (11.27%), and Cost (10.56%) ( Figure 3.2 ) All of the participant s reported that they wanted improvements and that they would be willing to use a device with a more complicated control system in order to get increased functionality. When asked specifically if they thought a device that incorporated the DTM would be usef ul, all participants reported that they did and that it would be worth the increased complexity. Figure 3 2 : Pie Chart showing how often participant s made each of the keyed comments during their interview. Overall, nearly one third of comments were about difficulty using tools or completing other ADL tasks. Another third addressed common complaints with prosthetics (e.g. durability, socket, cost, weight, and electronics). 3.4 Discussion This study found that par ticipants were less satisfied using their prosthetic devices to perform ADL tasks that involve the DTM than those that do not. In interviews, difficulty R (ADL issues) 15.49% B (difficulty using tools) 14.79 14.79% A (cosmetic) 11.27% G (durability) 11.27% C (cost) 10.56% D (socket issues) 9.86% U (User wants improvement) 9.15% F (eletronics issue) 7.75% Y (BP is good at this) 2.82% L (Myo is good at this) 2.11% N (weight) 2.11% I (finger control) 1.41% E (safety issue) 0.70% K (function over form) 0.70% Keyed responses from all participant interviews

PAGE 64

45 performing ADL tasks and tasks that specifically use the DTM accounted for 30% of keyed user comments. Paradoxically, many survey participants rated tasks as difficult and important but also claimed that they were satisfied performing these tasks with their current device. During interviews, participants explained that this was because they had developed c ompensatory motions that allowed them to perform the task to their satisfaction even though they still found it difficult. For example, participants mentioned that they could use their intact limb to position an object before picking it up with their prost hesis. In many cases, participants stated that they simply avoid performing the task or ask someone else to do it for them. While these may be effective methods for completing tasks, compensatory motions can lead to long term over use injuries and avoiding or getting help with certain tasks limits the Of particular interest were responses to the questions about which additional DOF participants wanted to see in a prosthetic wrist. M ost users said they would like FE or RU in a device bu t few requested PS. This sentiment is consistent in user surveys going back as far as 1985 (69) However, it contradicts the standard thinking in prosthetic wrist design that PS should be the second DOF included after a gripping unit. Interviewees explained this seeming inconsis simulate it by abducting the shoulder. In most cases, interviewees volunteered much more information than the question specifically asked. This led to four of the five participants stati ng, without prompting, that they would like some sort of wrist that included FE and RU. Other suggestions included: a

PAGE 65

46 smartphone app to adjust the gain in a myoelectric device throughout the day, customizable grip shapes in a hand, and thinner fingers (e.g for reaching into pockets). Other complaints about prosthetic devices were consistent with previous literature (22,67) Everyone interviewed mentioned problems such as: uncomfortable and insecure socket, inconsistent myoelectric res ponse, weight, cost, and durability. Combined with the fact that a third of comments referred to difficulty performing ADL and using tools, these responses support the idea that a powered wrist unit that incorporates the DTM would be beneficial to this pop ulation. These results could be interpreted as showing that users simply want a more functional device. However, this study was designed to determine specific improvements users were looking for. Notably, users in the survey and interviews were strongly i nterested in having a device with powered FE or RU Additionally, given that any increase in functionality will make the device more difficult to use, the focus should be on the improvements that users are excited by and motivated to try. A limitation of this study was the fact that self selected and self reporting populations carry an inherent bias: people who choose to fill out online surveys tend to be the people with stronger positive or negative opinions. This is a known limitation of this type of sur vey (83) Efforts were made to recruit from a diverse range of sources and questions were written in a neutral voice so as to avoid leading.

PAGE 66

47 3.5 Conclusion The limited functionality of current prosthetic devices negatively affects the quality of life of people with amputations. While it is sometimes possible to compensate for this limitation, such movement patterns ultimately lead to additional health problem s (50) Simply adding all the missing DOF would be immensely d ifficult to control by users with This study determined what improvements are desired by the people with upper limb amputations in their prosthetic arm and hand devices. Specifically, DTM was identified as a DOF which users have partic ular difficulty mimicking with their device. The results suggest that a device incorporating the DTM would improve functionality for most users. Additionally, people were generally willing to accept a more complicated device in return for greater functiona lity. This motivates the development of a new prosthetic wrist device that incorporates this motion. Based on the interviews, a device like this would be welcomed by this population as long as it provided improved functionality, was durable, did not cost t oo much, and was easy to use.

PAGE 67

48 C HAPTER IV M EASUREMENT OF THE D ART T HROWER S M OTION IN THE C ONTEXT OF A P ROST HETICS S PECIFIC F UNCTIONAL T EST The study in the last chapter identified the lack of the DTM in prosthetics as a potential ability to perform ADL tasks. These survey results and the DTM presence in many ADL tasks indicate that the DTM is a strong candidate for creating an improved, simple wrist (27) Previous studi es have identified the DTM as motion in a plane rot ated 20 40 degrees from pure flexion extension but a more precise measurement is needed to build a prosthesis (20,27,37) To determine this rotation angle able bodied participants completed a series of standardized tasks while equipped wit h electrogoniometers to measure wrist motion. Measu rement of the wrist determined that FE and RU were coupled in the DTM plane in a majority of tasks and were rarely coupled in the anti DTM plane The DTM accounted for 46% of the overall motion during the grasp phase with rotation accounting for another 28 % of the variance. In the 2 dimensional FERU plane, 67 % of the variance in the functional range of motion of the wrist was coupled at an of angle 21.8 degrees rotated from pure FE These findings suggest that a prosthetic wrist that combines these DOF woul d improve functionality and only minimally increase control complexity. The results presented here are intended to give insight into the motion of the wrist during a standard of care assessment tool for upper limb prosthetics. The focus is on improving th e functionality of prosthetic devices by combining two or more of the motions

PAGE 68

49 identified by this study. The results are, potentially, also useful to people studying non amputation disabilities of the arm (e.g. arthritis) as well as to medical professionals assessing their patients. 4.1 Methods: Development of a W rist M easurement S ystem Measuring the motion of the wrist is a challenging task. Although the wrist can be approximated by a 3 DOF joint containing two pin joints and a rotation, FE and RU are complex motions of the eight carpal bones and PS is the function of the radius and ulna, not the wrist itself. Attempts to accurately measure the wrist with an active marker motion capture system were unsuccessful. However, electro goniometers consisting of flexio n sensitive resistors proved an effective wa y to measure the wrist during this study. 4.1.1 Motion Capture Motion capture is one of the more common methods for measuring physiological motion. Markers placed on anatomical landmarks are recorded by cameras and th eir position in 3D space is calculated by software. From these measurements, it is possible to calculate the position s of body segments and joint angles (85) Initial attempts to use a motion capture system (Optotrak Certus, Northern Digital Inc. Ontario, CAN) proved unsuccessful. After an initial bout of hardware and software issues, pilot testing revealed several limitations to this system. The Optotrak system uses a single tower of three cameras to locat e instead of the commonly used infrared (IR) reflective markers. These markers have IR LEDs

PAGE 69

50 that flash at distinct frequencies that the system can automatically identify. This avoids the time consuming need to post process the data and ide ntify the markers in software manually. This is an advantage in some situations where the markers are directly facing the camera but, in this case, markers were often obscured from the camera during wrist rotation making this system ineffective for this st udy. Attempts were made to remedy this issue by creating rigid plates with four markers in fixed positions ( Figure 4.1 top) These plates were attached to the proximal and distal dorsal forearm, and to the dorsal hand. This setup allowed the Optotrak software to determine the location of the body segments However, camera position was still a limiting factor: from wrist pronation to supination the dorsal hand plate rotated out of view of the camera and the positio n was lost.

PAGE 70

51 Figure 4 1 : Attempts at overcoming the Optotrak limitations Rigid plates (above) were attached to body segments which correctly located the segment but could only be seen from a fixed vantag e point. Multiple markers attached to each segment and defined as a opposite side of the limb. This method was also ineffective because the Optotrak software was unable to account for soft tissue artefacts ( i.e. minor motion between markers from finger abduction or forearm skin stretch) in determining the location of the rigid body. Another attempt at solving this issue involved placing 8 10 markers on the forearm and ha ( Figure 4.1 bottom) This allowed the Optotrak software to calculate the position of the other markers based on the position of at least three visible markers. However, this calculation is distortion intolerant and even the slight change in relative position of the markers due to soft tissue artefact ( e.g. skin motion from finger abduction) were enough to prevent the Optotrak software from accurately determining the posi tion of the body segments. Given these limitations, this form of measurement was abandoned. Consideration was given to using the motion capture facilities at The Center for Gait and Movement Analysis and the Interdisciplinary Movement Sciences Laboratory at the University of Colorado, Anschutz Medical Campus. However, these options were dismissed in favor of using

PAGE 71

52 electrogoniometers which are easier to transport, more customizable, and produce data that is easier to process. 4.1.2 Flex Sensitive Resistors A sys City, UT) as a proof of concept for a commercial goniometer system. The design was based on the system described by Wang et. al. (86) s design, only one FSR was necessary to measure each axis in this case as the FSRs were equally responsive in each direction. Voltages were read from the FSRs via an NI DAQ 2006 board (National Instruments Corporation, Austin, TX) and recorded in LabView ( National Instruments Corporation, Austin, TX). The idealized voltage response from the FSR circuit ( Figure 4.2 ) is ( 4 1 ) where V in is the voltage supplied by the NI DAQ, V out is recorded by the NI DAQ, R FSR is the value of the FSR, and R 2 is a fixed resistor.

PAGE 72

53 Figure 4 2 : Circuit diagram for the flex sensitive res istor setup. V in and V out were generated and recorded in Labview. The FSR system was tested by attaching the FSR to a manual goniometer and recording the voltage in Labview and the indicated angle ( Figure 4.3 ) Me asurements were made three times in each direction. Minimal hysteresis was found to be 0.04 V over a 0.59 V range, a 6.7% variation. Based on these measurements, the corresponding anatomical angle (in degrees) was determined from voltage by ( 4 2 )

PAGE 73

54 Figure 4 3 : Angle vs Voltage in flex sensitive resistor (FSR) testing. FSR was attached to a plastic goniometer and the voltage was recorded via La bview at various angles and moving in the forward (black) and backward (white) directions. Figure 4 4 : Mockup of the FSRs attached to the dorsal and ulnar sides of the wrist. After positive bench test resu lts, a pilot test was performed by attaching the FSRs to the dorsal and ulnar wrist ( Figure 4.4 ) The pilot test revealed that the FSRs could acc urately measure both FE and RU ( Figure 4.5 ) Based on these findings, it was concluded that electrogoniometers based on FSRs would be an effective method for measuring wrist motion. y = 0.0043x + 0.5201 R = 0.9907 y = 0.003x + 0.4813 R = 0.9619 0 0.2 0.4 0.6 0.8 -100 -50 0 50 100 Voltage Angle Angle vs Voltage (FSR) Toward Black Toward White Linear (Toward Black) Linear (Toward White)

PAGE 74

55 Figure 4 5 : Plot of the wrist angles flexion exten sion (FE) and radial ulnar deviation (RU) recorded with the FSR system during simulated dart throwing. 4.1.3 Commercial Electrogoniometers Commercially available e lectrogoniometers (Motion Labs Systems Inc., Baton Rouge, LA) were tested by attaching them to a ma nual goniometer and recording the output voltages were in LabView. A torsionmeter (rotation sensitive resistor) was tested by attaching it to two pieces of pi pe marked with angles and rotating. T he volta ge was again recorded with LabView. Measurements were taken three times in each direction and averaged. The goniometer and torsionmeter had a linear relationship between voltage and angle within the operating range of 90 degrees to 90 degrees ( Figure 4.6 ) Hysteresi s was -150 -100 -50 0 50 100 150 0 1000 2000 3001 4001 Angle (deg) Time (ms) FE and RU plots with FSR F-E angle R-U angle

PAGE 75

56 Figure 4 6 : Results of calibration test of the goniome ters and t orsion meter over 180 degree operating range. Hysteresis was found to be <2.5 degrees over 180 degrees. Relationship between output voltage and angle is highly linear. 4.2 Methods: Standardized Motions The Southampton Hand Assessment Procedure (SHAP) is a vali dated, standardized measure of hand functional ability that is becoming a standard assessment tool in prosthetics research ( Figure 4.7 ) (28) It was developed by Colin Light, Paul Chappell and Peter Kyberd in 2002 at the University of Southampton (87) The SHAP also has high test retest reliability (88) y = 102.25x + 37.14, R = 0.9996 y = 103.53x + 40.043, R = 0.9994 y = 107.7x 11.216, R = 0.9999 y = 107.76x 9.1336, R = 0.9998 y = 117.06x + 3.722, R = 0.9987 y = 116.61x + 6.1324, R = 0.9988 -90 -60 -30 0 30 60 90 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 Angle (deg) Voltage (mV) Goniometer and Torsionmeter Calibration Curves FE (to +) FE (+ to -) RU (to +) RU (+ to -) PS (to +) PS (+ to -)

PAGE 76

57 Figure 4 7 : SHAP design showing the abstract objects (ball, cylinder, etc.) and simulated ADL objects (cup pouring, key turning, etc.). Participants perform timed movements of the objects from which the assessor can calculate the SHAP score (89) During the SHAP, participants manipulated six light abstract objects, six heavy abstract objects (e.g. a ball or cylinder) and completed fourteen simulated ADL tasks ( e.g. pouring water from a jug or turning a door ha ndle). A list of all SHAP tasks can be found in Table 4.1 Participants start a timer with their active hand, move the target object using the instructed grip type and technique, and then return to stop the timer. The SHAP returns a score based on t he time to complete each task. This score ranges from 0 (no function or could not complete tasks) to 100 (normal function). While time to completion is not the only important measure of a prosthesis, the SHAP has been sho wn to be able to identify clinically relevant differences in prosthetic functionality (28)

PAGE 77

58 Table 4 1 : List of tasks involved in the S outhampton Hand Assessment Procedure. Light Abstract Objects Simulated ADL Tasks L Sphere Place coins in Jar L tripod Undo buttons L cylinder Cut food L Lateral Turn page L tip Remove jar lid L extension Pour water from Jar Pour water from ca rton Heavy Abstract Objects Move heavy object H sphere Move light object H tripod Lift and move tray H cylinder Turn key H lateral Undo zipper H tip Turn screw with screwdriver H extension Turn door handle 4.3 Methods: Principal Component Analysi s Principal component analysis (PCA) was first introduced by Pearson in 1901 and later by Hotelling in 1933 (90) It is used to find correlations between multiple variables by finding a new basis into which the data can be rotated to maximize the variance in the first orthogonal dimension (91) PCA is useful for identifying trends in large data sets and reducing the dimensionality of data set s with many DOF. On biomechanics data, PCA is

PAGE 78

59 (91 p416 ) These features make PCA potentially useful in identifying a way to reduce the wrist from three DOF to one. The following discussion of PCA and its underlying mathematics rely heavily on Daffershofer, 2004; Demsar, 2012; and Schlens, 2003 (90 92) PCA rotates an m x n matrix, X, into a new basis via a rotation matrix, P, with the goal of maximizing the variance in the first dimension. This is written as ( 4 3 ) and ( 4 4 ) where P is an orthonormal matrix and T is the projection of X into the new basis. That is, the matrix X represents the collected data and T represents those data rotated such that the maximum variance is along the first axis. The matrix P is calculated in such a way that its columns are ordered in the d escending order of variance. This can be calculated from the covariance matrix of X which is defined by

PAGE 79

60 ( 4 5 ) X must have zero mean in this case. It is important to note that the PCs can also be calculated from the correlation matrix of X. These terms are interchangeable in this derivation. P is calculated by ( 4 6 ) where is a diagonal matrix of ordered Eigenvalues. The Eigenvectors and Eigenvalues are typically calculated by a computer programs such as Matlab or Mathematica. Anyone interested in finding the Eigenvectors by hand are encouraged to read chapter 10 of Mathematical Methods in the Physical Sciences by Boas (93) The a ngle of rotation from the original data to the PC space for the n th PC can be found by ( 4 7 ) where and are the i th and j th values of the n th principal component These could, for instance, indicat e the ratio of contribution s of FE and RU to the DTM in the first PC. Physically, this represents the angle of rotation from the pure anatomical motion planes (FE, RU, PS).

PAGE 80

61 any reasons. P describes the rotation matrix that transforms the original data into a new basis where the first variable accounts for the most variance. This identifies how the different variables are coupled and in what direction and also whether there is redundancy in the variables. component (PC), k, is given by ( 4 8 ) That is, the variance explained by each PC is the associated Eigenvalue divided by the sum of the Eigenvalues. This value can inform a reduction in dimensionality of the system. For example, if there are five variables b ut the first three PCs account for 80% of the variance, then the last two PCs can be dropped while only losing 20% of the information. This is the application of PCA that was used in this study. Measurements were made on the three DOF of the wrist. PCA wa s applied to the resulting data with the goal of determining if any of the DOF were coupled and if a majority of the variance was contained in one or two DOF. If this is the case, it was determined that the wrist could be effectively reduced from three DOF to two or one DOF. 4.3.1 Spatial vs. T emporal PCA PCA was originally developed to describe spatial data but can be expanded to give the same insights into temporal data. This can be particularly useful in biomechanics data. However, this is most easily done on cyclical data, such as the angles of the leg joints during

PAGE 81

62 walking. This was not the case in this study because each of the SHAP activities are single actions and the motions from multiple participants do not have a specific event to use to synchronize it in time (such as heel strike in gait). Therefore, the positions of the wrist during each task were concatenated and plotted as three dimensional spatial data (FE vs RU vs PS). This effectively removes any temporal information Therefore, the result of this analysis was effectively a functional range of motion in the SHAP. 4.3.2 A S tatement A bout E rror R eporting in PCA There has been a great deal of discussion about how to report error in PCA. One of the rs to tweak and no coefficients to adjust based on user exp erience the answer is unique and independent of the user (12 p12 ) That is, PCA simply finds the basis that maximizes variance. The results of PCA are simply reported as the Eigenvectors and the % variance e xplained by each. There is no propagated through the analysis because it is accounted for in the Eigenvalue. Indeed, if the error is evenly distributed about the mean, it s hould have no effect on the PC calculation. If this is not the case, the raw data should be used. One way to quantify how well the PCA describes the data is to report the signal to noise ratio (SNR) in the first PC. This method was suggested by Shlens, wh o described the SNR as

PAGE 82

63 ( 4 9 ) where 2 is the variance and SNR>>1 indicates data with high precision (92) alr eady given by the Eigenvalues. That is, if a PC is already known from the Eigenvalues to account for, say, 80% of the variance, no more insight is gained by calculating that 80/20>>1. Because of this, the PCA in this chapter will be reported as the angle c alculated from the PCs and the variance described by each of those components. 4.4 Methods: Experiment A convenience sample of fifteen adult volunteers (18 85 years of age) with intact upper limbs was recruited to participate in this study. Participants comple ted the SHAP as described by the SHAP protocol. Wrist position was recorded using electrogoniometers and graphic information (gender, age, height, weight) was also collected. Inclusion/exclusion criteria were chosen to ensure that the sample represented the wrist motion of people with normal physiology. 4.4.1 Inclusion/ E xclusion C riteria Participants were required to meet the following inclusion criteria: Between the ages of 18 and 85 years

PAGE 83

64 Intact upper limb physiology No complicating cognitive or physical impairments Normal or corrected vision and hearing Ability to read and understand English at a cognitively app ropriate level Ability to understand and follow instructions in English Ability to read, understand, and give consent to participate Participants were excluded from taking part in this study based on the following criteria: Below the age of 18 or above th e age of 85 Pathologic upper limb physiology Pathological cognitive or physical impairment Inability to read, understand, or follow directions in cognitively appropriate English Unable or unwilling to give consent There were no exclusions based on race/eth nicity, gender, or socioeconomic status. Participants were screened to ensure they met the inclusion/exclusion criteria before beginning the study. These inclusion criteria were required to ensure that measurement s represented anatomical motion without c omplications from impairment.

PAGE 84

65 4.4.2 Recruitment Participants were recruited via email, word of mouth, and flyers posted at the Anschutz Medical Campus. Participants contacted the researchers via email or phone to schedule a time to come to the Biomechatronics La test took less than an hour to complete. Each participant read and signed an approved consent form before beginning the study No personally identifiable information was recorded. Participants were not comp ensated for taking part in this study. This study was conducted with the approval of the Colorado Multiple Institution Review Board (COMIRB #14 0838). 4.4.3 Electrogoniometers Participants were equipped with commercial electrogoniometers (Motion Labs Systems Inc ., Baton Rouge, LA) on the dorsal wrist to measure FE and RU, and a torsionmeter on the volar forearm to measure pronation supination (PS) during the tasks. Voltages were read from the goniometers via an NI DAQ 2006 board and recorded in LabView. Linear fi t equations relating voltage to angle were incorporated into the LabView VI. Angle measurements were zeroed with wrist neutral, forearm neutral, and elbow at 90 degrees. Positive motion was defined as extension, radial deviation, and supination. Voltage, a ngle calculator, calibration data were recorded in comma separated value format ted files

PAGE 85

66 4.4.4 Assessment Participants performed the SHAP in accordance with the defined protocol (28) P articipants were seated with the target object directly in front of their active arm and with the chair height adjusted so that their forearm was parallel with the table when the elbow was at 90 degrees. Participants began each task with their hand flat on a timer, pressed the Recording began before the participant began the task and continued recording after they had ended visualize how closely this motion related to the physiological DTM. 4.4.5 Processing Data were processed in Matlab (The MathWorks, Inc., Natick, Massachusetts). Each trial was filtered using a 4 th order zero phase, Butterworth filter with a 10 Hz cut off frequency to remove noise and preserve biological motion (94) This cut off frequency is supported by the literature (91,95,96) and is confirmed by frequency analysis of pilot d ata showing that the majority of motion information is contained in the <10 Hz range ( Figure 4.8 )

PAGE 86

67 Figure 4 8 : Frequency spectrum of sample wrist motion showing th at the majority of information is below 10 Hz. This justifies the use of a filter with a 10 Hz cutoff frequency. The beginning of the reach phase was identified by a deviation of greater than 5 degrees from the initial rest phase in any of the three angles The end of the reach phase (beginning of grasp phase) was manually selected by a visual comparison of the time plot of the trial to a recording of participants only performing the reach phase of each task. The end of the grasp phase was manually selected by visual inspection at the beginning of the movement back to the final rest position.

PAGE 87

68 Figure 4 9 : Demonstration of the phases of a SHAP task during a single trial The initial rest phase results from the need to start recording before the participant begins the task. The reach phase typically involves the greatest motion as the hand is positioned at the object. The Grasp phase is typically characterized by the least motion and determined by recordings of the participant only holding the object. The final rest phase begins once the object is released and continues beyond the end of the task until the recording can be manually stopped. The grasp phase is the more important of the two for interacting with obj ects (97) The redundant nature of the DOF of the upper arm allows people to take many different paths to reach an object. This means the reach phase includes more variance than gras p. More importantly, the course motions people use to approach an object matters far less than how the hand and wrist are positioned for object manipulation. Therefore, while this study also analyzed the motion during reach phase, the conclusions and next steps are based on the results of the grasp phase. 4.4.6 Principal Component Analysis PCA was performed in Matlab using the pca() function Each reach and grasp phase was interpolated to create a phase plot of 0 100% (101 points). PCA was first performed on the z score (mean subtracted, variance divided) on the collection of all SHAP trials to find the correlations between DOF Since PCA measures variance, centering the data and dividing by

PAGE 88

69 the variance prevents DOF with large motions from dominating the results and skewing the fit (91) Without dividing by the variance, rotation dominates all of the components and any understanding of how t he three DOF move relative to each other is lost. Performing PCA on the z score measures how the three DOF move relative to each other and reveals coupling between the motions. This provided an assessment of the functional range of motion of the wrist duri ng the SHAP. Coupling was assessed in each of the SHAP tasks by performing PCA on the z score of all participants in each task. and B was defined as |A|>0.5 & |B|>0.5 and % contribution to varian ce >5 0. This indicated that at least half of the information in the motion was contained in that PCA component (91) Opposite coupli ng ( ) was was defined as Sign(A) = Sign(B). In this configuration, the DTM was represented by positive FE and negative RU (FERU ) while the anti DTM motion was represented by FE and RU with the sam e sign (FERU+). A random sample would show strong coupling in one direction (opposite or parallel) 12.5% (1/16) of the time. Once coupling was found between two of the three DOF, PCA was performed on the centered data. In this case the position was not d ivided by variance so as to preserve the proportionality of the motion of the two variables rather than simply the correlation between them.

PAGE 89

70 4.4.7 Coupling C ounts and P osition A nalysis T he percentage of tasks which used each type of coupling (FERU, RUPS, and PS FE) and in which direction there was coupling was determined Of chief concern was that some tasks had positive coupling and others had negative coupling for a given combination of DOF. C reating a device that coupled those DOF in only the positive or negat ive direction would result in a device that was helpful some of the time but detrimental the rest of the time. Finally, how often the wrist was in each of the possible positions (which of 8 quadrants on a FE RU PS plot) was counted to determine whether th e wrist has a preferred position. This allowed for the determination of the typical orientation of the wrist during each phase. This was yet another method for determining whether there was a particular combination of DOF that could be combined in a device 4.5 Results Fifteen adults (aged 30.7 8.8 years) performed the SHAP while wearing electrogoniometers to track wrist FE, RU, and PS. Each participant successfully completed the entire SHAP with no complications. P articipants average SHAP sco re was 97.6 3 .9 out of 100, indicating that each participant ha d normal hand functionality and provided an accurate measure of intact hand and wrist motion during the SHAP.

PAGE 90

71 4.5.1 PCA on Reach and Grasp R each and grasp motions were similar across all participants Although t he starting points were different between participants, the difference in motion (motion error) was small. The average motions of each task show ed a consistent pattern with minimal error across the task (mean error = 0.83 degrees ) as seen in ( Figure 4.10 ) The first principal component of the reach phase was PC 1 = [0.59, 0. 57, 0.57] (FE, RU, PS) with a % contribution to variance of 45.61. The first principal component of the grasp phase was PC 1 = [ 0.64, 0.68, 0. 36] with an average % total variance of 45.12.

PAGE 91

72 Figure 4 10 : Motion of the wrist during the reach phase (top) and grasp phase (bottom) of each task in the SHAP averaged across participants. Error bars ind icate standard error. For each task a distinctive motion from the rest position to the beginning of the grasp position is visible During the reach phase there is a visible coupling between FE and PS. During the grasp phase, many of the tasks have simila r motions while the more complex tasks (coins, buttons, pouring) have distinctive profiles. During grips, FE and RU are commonly correlated. Simulated dart throwing (DTM) was included as a comparison.

PAGE 92

73 Table 4 2 : Results of PCA on the z score of all SHAP tasks. The first component accounted for about 45% of the motion in both the reach and grasp phases. In the reach phase, the three motions were equally coupled and FE and RU were coupled in the DTM direction. In the grasp phase, FE and RU were strongly coupled in the DT M direction while rotation made up nearly all of the motion in the second PC. Note that the sum of the squares of the components of each PC equal s to one. PC1 PC2 PC3 PCA Reach FE 0.5884 0.0082 0.8085 RU 0.5715 0.7116 0.4087 PS 0.5719 0.7026 0.4234 % Variance 45.6101 27.5587 26.8311 Grasp FE 0.6391 0.3844 0.6662 RU 0.6797 0.1230 0.7231 PS 0.3599 0.9149 0.1827 % Variance 45.1243 31.9224 22.9533 T he calculated offset ang les in two dimensions for the reach phases ( Figure 4.12 (left)) and for grasp ( Figure 4.12 (right)) phases are shown below 4.5.2 2 Dimensional PCA on Averaged Data Initially, t he offset angle in each 2D plane was calculated on the wrist data averaged over participants. In this method, t he largest offset angle in grasp phase was in the axial (FERU, i.e. DTM) plane at 26.72 degrees accounting for 76% of the variance. Offsets in th e PSFE and RUPS planes were 8.3 degrees with 89% variance and 6.8 degrees with 80% variance respectively.

PAGE 93

74 Figure 22 : 2 dimensional radial plot in the FERU plane on the averaged wrist data. Red and green lines represent the first and second principal component. Their lengths represent the percent contribution of each PC. 4.5.3 2 Dimensional PCA on Full Data Set As noted in section 4.3.2 there are consequences to making the assumption that error is evenly d istributed. In the initial analysis of the angles in each 2D plane, the data were averaged across participant s before PCA was performed (section 4.5.2 ) This was done with the intent of smoothing and simplifying the data. Howev er, this assumption is only accurate if the error is normally distributed and, in this case, it is not. Making this assumption was the equivalent of taking an intermediate mean of the data. Recalculating PCA o n the 2D data without averaging, the largest o ffset angle during the reach phase was 28 degrees accounting for 68% of the variance in the PSFE plane. Minor angles were 10 degrees in the FERU plane accounting for 73% of the variance and 9 degrees in the RUPS plane accounting for 81% of the variance. In the grasp phase, the largest offset angle was 22% in the FERU phase accounting for 67 % of the variance followed by 4 degrees at 72% variance in the PSFE plane and 2.5 degrees at 81% variance in the RUPS plane.

PAGE 94

75 Figure 4 12 : Two dimensional radial plot s of wrist angle in the three planes during the reach phase (left) and grasp phase (right). R ed and green lines represent the first and second principle components, respectively. L engths indicate the perc ent contribution to the variance of each PC. Boxed plots show the largest deviation in each of the phase s

PAGE 95

76 4.5.4 Coupling The coupling during the reach phase was significantly greater than random in the FERU (DTM), RUPS PSFE+ and and between all three ( Figure 4.24 ) The coupling during grasp phase was greater than random in all but the FERU (anti DTM) mode. The levels of coupling only differed between reach and grasp phase in the PSFE mode. Figure 4 24 : Coupling in the three degrees of freedom of the wrist. Error bars indicate one standard deviation from the mean. We found significant coupling in the reach phases of FERU (DTM), RUPS PSFE+/ and in all three deg rees of freedom. G reater coupling in the grasp phase was found in all but the FERU+ (Anti DTM) mode. The majority of the grasp phases of SHAP tasks took place with the wrist in either flexion, ulnar deviation, and pronation or in extension, ra dial deviatio n, and supination ( Figure 4.25 ) 0 10 20 30 40 50 60 70 80 90 FERU+ FERU(DTM) RUPS+ RUPSPSFE+ PSFEFERUPS Percentage Coupling type Percentage of SHAP Tasks with Strong Coupling Random (12.5%) Reach Grasp

PAGE 96

77 Figure 4 25 : Percentage of SHAP task reach (left) and grasp (right) phases in given wrist orientation (F flexion, E extension, R radial deviation, U ulnar deviation, P pronation, S supination). During the reach phase, wrist position is present in all combinations with FUP the most common position. A plurality of grasps takes place with the wrist in flexion, ulnar deviation, and pron ation or in extension, radial deviation, and supination. This presentation is consistent with the DTM. 4.6 Discussion Wrist motion was coupled during both the reach and grasp phase. In the reach phase, the first PCA accounted for 46% of the variance and was ma de of roughly equal parts FE, RU, and PS indicating no preferred direction in this phase. While this does not suggest an appropriate reduction in DOF, the FE and RU components did have different signs indicating a DTM is part of the overall motion during reach In the grasp phase, t he DTM accounted for 45% of the motion in the first PC, while the second PC, at 28% was almost entirely PS. This suggests that a simplified prosthetic wrist could accomplish almost half of the functiona lity by including the DTM a nd over two thirds of the anatomical motion by including both the DTM and PS. In fact, the third PC, accounting for 22% of the variance, appears to be made up of the anti DTM motion,

PAGE 97

78 suggesting that a 3 DOF wrist might perform better with PS, DTM and anti DTM, rather than the anatomical, PS, RU, and FE (i.e. PS plus an offset flexion and deviation unit). The fact that the DTM accounted for 76% of the variance in the FERU plane is a strong indicator that it would be appropriate for reducing the wrist to 1 DO F. The RUPS and PSFE plots in Figure 4.12 (middle and bottom right) show little correlation and would be poor candidates for dimension reduction. A second way to determine which combination of DOF would be most eff ective is to measure how the DOF were coupled in the wrist motion during grasp phase. The DTM (FERU ) was the most coupled motion in the wrist and there was little anti DTM (FERU) coupling ( Figure 4.24 ). T here was also significant opposite coupling between RU and PS, between PS and FE and between all three motions. While these results indicated that there are several options for combining DOF in a prosthetic wrist, the other couplings were strong in both the and + directions. Therefore, about half of the time a coupled RU PS or PS FE would be helpful but the rest of the time it would be a hindrance. The DTM combination is the most promising because it ha d a high negative correlation (DTM) and low positive correla tion (anti DTM). That is, most of the time, the motion was in the DTM direction. Therefore, including the DTM would help more than half of the time and would rarely be detrimental. In addition, there were many trials that had strong coupling in all three m otions. This is encouraging but a prosthetic wrist that incorporates all of these motions has never been commercially available and building such a wrist would not solve the issue of not having enough control inputs to control all of the DOF.

PAGE 98

79 As a final a rgument for using the DTM to reduce the dimensionality of the wrist this study investigated in w hat position the wrist spent the most time during the reach and grasp phases ( Figure 4.25 ). D uring the reach phase, t here was not a preferred position of the wrist. However, during the grasp phase, a plurality of the time the wrist was in either FUP or ERS. These positions are consistent with motion in the DTM plane. 4.7 Conclusion and Next Steps This study demonstrated tha t FE and RU are coupled in both the reach and grasp phase of ADL tasks (as represented by the SHAP). This suggests that a single DOF, based on the DTM, included in a prosthetic wrist, would be helpful to most users. This hypothesis is tested in the next ch apter. In the future, this study could be replicated using other measurement systems. Although the goniometers were determined to be appropriate for this study, motion capture or radiography could potentially provide a more precise measurement. This could indicate how the bones of the wrist move during this motion which could be useful for a different wrist design or for informing wrist surgery practices. However, these methods are difficult to use and resource intensive. Although it was not the focus of this thesis, the angles of the hand joints were also recorded using a Cyber Glove II (CyberGlove Systems LLC., San Jose, CA) in eight of the participants. This was done to help design a new postural controller for the prosthetic hand and wrist based on the work done in 1998 by Santello et al (97) Santello showed that the 15 motions of the hand could be effectively described using only two degrees of freedom

PAGE 99

80 (DOF) while maintaining m ore than 80% of the information. While this has allowed for the creation of drastically more effective prosthetic hand control schemes, little work has been done to correlate wrist positions with these postures. This means that users must still perform com pensatory motions in order to properly position the hand for grasping. The work done here shows that the wrist can be reduced to two DOF and incorporating the hand position data has the potential to identify correlation between the hand postures and wrist positions, expanding the functionality of postural controllers.

PAGE 100

81 C HAPTER V T ESTING A N OVEL P ROSTHETIC W RIST T HAT I NCORPORATES THE D ART T HROWER S M OTION The last chapter demonstrated that a plurality of the motion of the anatomical wrist could be preserved in a prosthetic wrist with a single degree of freedom using the DTM. This study sought to determine whether this finding would translate to a single degree of freedom prosthetic wrist. Ten people with intact limbs completed the SHAP with a simulated amput ation and prosthesis with a wrist in the DTM, flexion, and fixed configurations. Three people with intact limbs completed the SHAP with their hands and five people with transradial amputations completed the SHAP with a custom socket and prosthesis in the D TM configuration as a comparison. Shoulder motion was measured with electrogoniometers to determine whether the different wrist configurations affected the compensatory motions. After the test, participants completed the NASA task load index (NASA TLX) to give a subjective measure of how difficult it was to use each device. The DTM wrist significantly improved SHAP scores over a standard FE wrist (p< 10 7 ). However, shoulder ab/adduction did not appear to be reduced by the DTM wrist (p>0.05). About half of the participants did display reduced shoulder motion though and it was observed that participants used several other compensatory strategies that were not measured by this study. Using the DTM wrist was no more difficult than using the FE wrist or no wris t according to the NASA TLX (p>0.05). This supports the hypothesis that reducing the dimensionality of the wrist using the DTM will improve performance while not substantially increasing the difficulty to use it. The secondary hypothesis, that the DTM wris t would reduce compensatory shoulder motions was not confirmed.

PAGE 101

82 5.1 Methods: Development Testing whether the DTM would improve functionality of the wrist involved three major development aspects. Frist, the new prosthesis was designed to meet the angles foun d in the previous chapter. Care was taken to make sure that the hand and wrist were fast enough and strong enough to be able to complete the SHAP tasks. Second, a myoelectric control system was written that allowed the user to effectively control the devic e. This control system was based on a position controlled state machine where the user could activate their muscle to move the device to a desired position and co contract to switch between controlling the hand or wrist. Third, goniometers were used to mea sure compensatory motions. Because of the limited number of goniometers available, a literature search was done to determine which DOF of the shoulder were most important to compensatory motion. 5.1.1 Development of the DTM W rist A single DOF wrist prosthesis that utilized the DTM was designed and built for this study using the results described in the previous chapter. The device was built with a variable angle that could be set to an appropriate angle. A tolerance of 5 degrees was deemed appropriate to reflec t the minimum clinically measurable angle (98,99) The DTM an gle used for the majority of this study was determined to be 2 7 degrees by finding the 2 dimensional principal components of the wrist motions averaged across participant s as described in section 4.5.2 It was later determined that the proper angle was 22 degrees as described in section 4.5.3 However, a difference of 5 degrees is less than the minimum

PAGE 102

83 clinically measurable angle and therefore within tolerance. A small, follow up study confirmed that the results from the 2 7 degree and 22 degree prosthesis were statistically similar. The wrist device was created by modifying a hand prosthesis designed for people with transmetacarpal amputations ( Figure 5.1 ) (100) The thumb of the transmetacarpal hand was removed and the fingers were replaced with a flexion part ( Figure 5.3 red) and an adjustable adapter plate ( Figure 5.3 green) that mated with the terminal device. The planetary gear stack in the hand was replaced with a commercial 415:1 gear box (Faulhaber, Schnaich Germany) which meshed with a 3:1 planetary gear system that ke yed into the flexion part. This was necessary to prevent the wrist from backdriving. The wrist was driven by a MicroMo 1724 006SR motor with IE2 16 encoder (Micromo, Clearwater, FL). Having an encoder on the Micromo motor allowed the wrist to be driven wit h velocity control (DC) or set to specific positions (position control). Specifics of this control system are discussed in section 5.2.4

PAGE 103

84 Figure 5 1 : The transmetacarpal hand us ed as the starting point for the DTM wrist flexion device used in this study (100) The fingers were removed and replaced with an adjustable flexion part and the gears replaced with a commercial gear box. Figure 5 2 : Design of the DTM wrist from CAD (left), to plastic 3D printed (middle), to the final machined device (right). The adapter plate could be positioned at 0 degrees to make the wrist move in the a natomical FE plane like a standard FE wrist or rotated to make the wrist move in the plane of the DTM. The wrist could also be disabled, in software or by disconnecting the motor controller cable, to produce a fixed wrist. This allowed the same wrist devic e to be used for each group in the experiment.

PAGE 104

85 A wedge was added to the DTM wrist to incorporate the PCA angles in the less significant planes found in section 4.5 ( Figure 5.3 teal). T he most prominent angle of the DTM, 27 degrees in the FERU plane, could be created by rotating the adapter plate. In order to include the RUPS and PSFE components of the DTM, a wedge was created and secured between the adapter plate and the hand. The wedge was angled at 6.8 degrees in the RUPS plane and 8.3 degrees in the PSFE plane. This allowed the wrist to move along the DTM angle with a single actuator. Note that the angle measurements on the wedge are more precise than the adapter plate. This is becaus e the wedge was designed in SolidWorks and precision printed on a metal 3D printer, an EOS M270 (EOS of North America, Inc., Novi, MI), whereas the adapter plate was set by hand using a manual goniometer. Figure 5 3 : CAD view of the final DTM wrist (left) and exploded view (right) created for this study. Custom parts for this study are colored: Flexion Part (red), Adjustable Adapter Plate (green), Securing Annulus (yellow), and Wedge (teal). With the wedg e removed and the adapter plate rotated to 0 this wrist functions as a standard flexion wrist. The adapter plate and wedge allowed the DTM wrist to emulate the motion of the anatomical DTM. This is shown in Figure 5.4

PAGE 105

86 Side Front Top Neutral Forward Backward Anatomical Figure 5 4 : Illustration ability to mimic the anatomical DTM

PAGE 106

87 This wrist was the culmination of sever al iterative designs. This process was aided by Solidworks CAD software and 3D printing. A Stratasys Connex350 was used for printing prototype plastic and rubber parts and an EOS M270 print ed the final wedge part Other parts were manufactured by the team engineering department machine shop in Denver, Colorado. 5.1.2 Hand M odifications There are many commercial and research grade multiple DOF hands available that allow independently articulated digits and specific grip types (70) However, these hands necessitate more complicated control systems (10) More complex hands and complicated control systems have the potential to obscure the benefit s of any new wrist mechanisms. Furthermore, there is doubt as to whether independent fingers provide useful benefits in prosthetic devices (101) Since the goal of t his study was to determine whether a DTM wrist would improve function, a simple, 1DOF, hand was used. The hand chosen for this study was an updated version of the transmetacarpal hand described in (100) The fingers were drilled out to be lighter, but otherwise the design was the same. The hard plastic finger tips in the original design slipped during grasping so cotton padding was added to the finger tips and covered with a rubber glove to improve grip. The planetary gear train was replaced with a commercial 1526:1 gear head. Both of these modifications were necessary to provide enough grip strength to effectively grasp the objects used in the SHAP. The hand was also driven by a MicroMo 1724 006SR motor with IE2 16 encoder (Micromo, Clearwater, FL). The modified hand is shown in Figure 5.5

PAGE 107

88 Figure 5 5 : Modified transmetacarpal hand used in this study. Cotton padding and rubber gloves were added to the fingers to improve grip. The wedge and adapter plate of the DTM wrist can be seen at the bottom. 5.1.3 Functional C haracteristics of the P rosthetic W rist and H and When completed, the wrist had a maximum range of 90 degrees a nd a rate of 1.0 rad/sec. The maximum opening distance was 10 cm (50 degrees) which could be closed in 3 seconds for a speed of 0.3 rad/s. Maximum grip strength was sufficient to lift SHAP objects but was not specifically measured. Likewise, product lifeti me was considered but decided to be unimportant for a research device in this case. Physical properties of the prosthesis components are summarized in Table 5.1

PAGE 108

89 Table 5 1 : Physical properties of the prosthesis used in this study. Mass (g) Excursion (degree) Speed (rad/s) Hand 190.44 50 0.3 Wedge 32.31 Wrist 175.47 90 1.0 Total ( assembled ) 401.51 The prosthetic wrist ha d a comparable ROM to the anatomic al wrist (102) speed of 1 rad/sec was on par with many prosthetic FE wrists that also move at speeds below 1 rad/sec and provide low torque (11) The hand was slower than the 3 rad/sec that the anatomical hand typically uses during ADL t asks and lower than the 1 4 rad/sec recommended for prosthetics (11,103) This slower speed was a result of the need to increase torque in the hand in order to lift SHAP objects. 5.2 Methods: Experiment A sample of eighteen people, aged 18 85, in three groups were recruited to this study. Participants performed the SHAP with their anatomical hand, or with a prosthetic hand and wrist in three different configurations. Participants were equipped with electrogoniometers to measure shoulder motion during the SHAP. After the test, partici pants filled out the NASA TLX to indicate how difficult it was to control the prosthesis. Testing was performed

PAGE 109

90 convenient to the participant. Inclusion/exclusion cri teria was selected to ensure that an accurate measure of the function of the DTM wrist could be obtained from the SHAP. 5.2.1 Population Three groups of people were recruited to take part in this study to perform the SHAP under different conditions: Group 1: P eople with normal upper limb physiology performed the test without a prosthesis Group 2: People with normal upper limb physiology performed the test using a splint that simulated an amputation and a prosthetic wrist in three configurations: o Group 2.a: Fix ed wrist o Group 2.b: Flexion Extension wrist o Group 3: People with trans radial amputations performed the test with a custom made socket and the prosthetic wrist in the DTM configuration. Three people from Group 2 vol unteered to return to confirm that the 22 degree wrist was similar to the 27 degree wrist. A priori power analysis indicated that, for an estimated variance of 4 points on the SHAP score and a just noticeable difference of 5 points, ten participants woul d provide a 91% confidence (84) These assumptions were based on the results from the previous study and other work done by the Biomechatronics Laboratory (16,104) Due to the small number of available people wit h trans radial amputations, the three wrist conditions were tested by people with simulated amputations and the results compared to people with genuine amputations, who only tested the DTM wrist configuration.

PAGE 110

91 5.2.2 Inclusion Criteria To be eligible for this st udy participants in all groups were required to meet the following criteria: Between the ages of 18 and 85 years No complicating cognitive impairments Normal or corrected vision and hearing Ability to read and understand English at a cognitively appropriat e level Ability to understand and follow instructions in English Ability to read, understand, and give consent to participate Participants in G roup 1 and 2 were required to meet the following additional criteria: Intact upper limb physiology No complicatin g physical impairments Participants in G roup 3 were required to meet the following additional criteria Have a transradial amputation (acquired or congenital) Have experience using at least one myoelectric prosthetic device Participants were excluded from taking part in this study based on the following criteria: Below the age of 18 or above the age of 85 Pathologic al upper limb physiology (for Group 1 or 2) Normal upper limb physiology (for G roup 3) No experience using a myoelectric prosthesis (for G roup 3)

PAGE 111

92 Pathological cognitive or physical impairment Inability to read, understand, or follow directions in cognitively appropriate English Unable or unwilling to give consent There were no exclusions based on race/ethnicity, gender, or socioeconomic status. Participants were screened to ensure they met the inclusion/exclusion criteria before beginning the study. These inclusion criteria were required to ensure that participants were able to provide appropriate data for their group. Each participant read and signed the consent form before testing began. No personally identifiable information was recorded for this study. Participants who spent more than two hours in the lab were given a $30 Amazon gift card. This research was approved by the Colorado Multiple Institutions Review Board (COMIRB # 14 0838 ). 5.2.3 EMG and Prosthesis Fitting Participants in G roup 2 (intact limb) and G roup 3 (limb deficiency) were fitted with a prosthesis that consisted of a single DOF hand and DTM wrist. Pairs of electrodes (Noraxon, Sco ttsdale, AZ ), were placed over the dorsal and ventral forearm muscles in the intact or residual limb. A ground electrode was placed over the olecranon. Electrode placement was determined by manually palpating the arm while the participant activated flexor and extensor muscles. Independent activation of muscle signals was confirmed in software before attaching the splint or prosthesis. Gains and thresholds were manually set by the tester in software to minimize the contraction strength required to activate t he device and

PAGE 112

93 to minimize unintended co contraction signals. These values were adjusted throughout the test as necessary to accommodate participant fatigue or movement of the electrodes. The arm was wrapped with Coban to prevent the electrodes from dislodg ing during the test. Participants in G roup 2 were fitted with a splint with the custom prosthesis attached to the end. The prosthesis extended about 10 cm distal to the hand ( Figure 5.6 ). Cotton padding was added around the forearm if necessary. This setup was the same for each of the three prosthetic configurations. The arm of people in G roup 3 was wrapped in cotton gauze before a fiberglass cast was created around the residual limb that captured the lateral and m edial epicondyles to create anatomical suspension. Once the cast was set, the prosthesis was attached to it with acrylic stays and a second layer of fiberglass cast ( Figure 5.7 ).

PAGE 113

94 Figure 5 6 : Process for simulating an amputation in an intact participant. Attach electrodes and EMG sensors (Top left) a pply Coban (top right) apply padding (bottom left) splint with prosthesis (bottom right) Figure 5 7 : Process for building a custom myoelectric socket for a participant with an amputation. Bare residual limb (top left) a pply electrodes and EMG sensors (top right) w rap in cotton and fiberglass casting material (bo ttom Left) apply second layer of fiberglass cast to secure the prosthesis to the inner cast (bottom right)

PAGE 114

95 5.2.4 Myoelectric C ontrol A hybrid velocity position control system with co contract to switch between hand and wrist control was created for this study ( Figure 5.8 ). To the participant, this system behaved like standard of care velocity control. This system had the benefit of preventing backdriving without the need for a higher gear ratio gear box that would move very slowly. Without this control system in place, the hand could not pick up several of the objects in the SHAP (e.g. heavy sphere, heavy lateral, and jug pour). The control system was written in Labview (National Instruments, Austin, TX) and the command signals were sent to the motors via custom motor controllers (Sigenics Inc. Chicago, IL ). Processing of the EMG signal (full wave rectifying, smoothing) was handled in hardware by the EMG electrodes. Programmatically, this system worked by creating a set point for the motor position and moving that set point in the positive or negative direction by a fixed amount (v) when the participant generated a contraction that exceeded a manually defined threshold. This was implemented in LabView with the following logic: if((EMG[0]>0 && EMG[1] == 0) || (EMG[0]==0 && EMG[1] > 0 )) wristPos[w] += sign(EMG[0] EMG[1])* v[w]; That is, when one of the EMG signals is non zero and the other is zero, the target position ( wristPos[w]) for the currently selec ted motor ( w) moves in the direction of the contraction by an amount ( v[w] ) defined for that motor Zero EMG signal could be the result of no muscle activity or because the level of activity was below the threshold. Independent speeds could be set for each motor. The range of motion of the motors could be set in software to prevent the prosthetic wrist and hand from jamming as they ran into their

PAGE 115

96 mechanical limits. No motion of the prosthesis was allowed while the participant was co contracting. Figure 5 8 : Illustration of the hybrid velocity position control system. Top left: raw EMG, Top right: thresholded EMG, Bottom: hand and wrist positions Participant activates flexors (red) to flex the wrist, co cont racts for the required time to switch to hand control, activates flexors which now open the hand, moves to the target (gap in top right), and then activates extensors (white) to open the hand. The resultant movements can be seen in the bottom plots. Fig ure 5 9 : Co contraction display showing the controlled variables (threshold, refractory time, and co contract time). Participant has held a co contraction for 6 out of 10 required bins but missed one in the p rocess (#10) probably by dropping below threshold. Each green bin represents 10ms of co contraction time. Participants could see this display during the training session but not while performing the SHAP. A proportional integral derivative (PID) loop com pared the desired set point to the current motor position and determined a pulse width modulated output signal (PWM) to be

PAGE 116

97 sent to the motor controller. The three constants of the PID loop could be independently controlled allowing tuning of the signal to the motors and mechanical system of the hand and wrist. The maximum PWM had to be limited to prevent the motors from drawing too much current and tripping the internal circuit breakers in the motor controllers. The PWM was sent from LabView to an Arduino N ano (Adafruit Industries, New York City, NY) running custom code to the motor controllers and then to the motors (16) The position was read back from the motor encoders via the same system. To switch between controlling the wrist and hand, participants co contracted for a set amount of time. This logic was implemented in Labview. The co contract threshold and required time to switch could be set by the experimenter. Once the co contraction was held for the designated time, motor control switched and a vibrating motor taped to the back of icipant of their success. A refractory time was implemented to prevent the user from immediately switching back to the previous motor if they continued to hold the co contraction. Participants were given 15 minutes of practice with the device and SHAP obj ects before beginning the test. During this time, the gain, PID controller, threshold, and co contract to switch settings were tuned by the experimenter. Participants were also given instruction on how to best control the device. Instruction included some of the following examples: Explaining how the threshold worked. Explaining how to co contract to switch between controlling the hand and wrist. Suggestion to move the device to the target location while in a more comfortable position rather than while exte nded to the target option.

PAGE 117

98 Explaining that since the motors hold position it is not necessary to continue contracting the muscle to continue holding the object and that relaxing the muscle would reduce overshooting the position. Suggestions on ideal ways to manipulate the SHAP objects. Participants with intact limbs were instructed to contract their extensor muscles to extend the wrist or open the hand, contract their flexor muscles to flex the wrist or close the hand, and to make a fist to co contract. P articipants with limb deficiencies were instructed to watch the output signals on the screen to determine the best way to create individual signals and the co contraction. Each of the participants with a ULA had a different strategy. For example, one perso n imagined extending to open the hand, making a fist to close the hand, and rotating to switch; while another, who had extensive experience using a envisioning doing, but could easily generate indep endent or co contraction signals. Participants were allowed to view the output signals and co contraction threshold indicators during the training phase but not during the testing phase. During both the training and testing phases, a 3 V micro vibration m otor (Radio Shack, Fort Worth, TX) taped to the back of the hand vibrated to inform the participant when they had successfully co contracted to switch control between the hand and wrist. After the initial training, participants were allowed up to 5 minute s of guided instruction and practice before each task. This included the SHAP instructions and suggestions for strategies to best perform each task. Optimal strategies were suggested to e task with each

PAGE 118

99 prosthesis configuration was being tested, and not their ability to figure out how to perform the task. During each session, the participant performed the SHAP as designed with one of the three prosthesis configurations. The order of the sessions was randomized to reduce the learning effects. Sessions were each performed on different days. At the beginning of each trial, the hand was moved to 50% open (5cm) and the wrist was placed in a neutral position (0 degrees flexion). Participants a lways began each trial controlling the wrist. 5.2.5 Limited SHAP All participants performed the SHAP as described in the previous chapter with the following modifications. D ue to limitations of the hand, participants were not able to complete the Button Board, Simulated Cutting, Tray, or Zipper tasks. T hese were excluded from the experiment In addition, to prevent potentially spilling water on the equipment, the Jug Pour and Carton Pour tasks were modified so that instead of pouring water, participants position ed the empty jug or carton as if to pour while counting four seconds before replacing the object on the table. This time was chosen based on the time required for an intact individual to pour out the water. Participants perform ed all remaining tasks with t he prosthesis or their hand in accordance with the standard SHAP protocol while sitting upright with the task object in front of their prosthetic hand. Due to these changes, the maximum score possible on the SHAP was not 100. In accordance with the SHAP p rotocol, times for the four tasks that were not completed were

PAGE 119

100 entered as 100 seconds. Based on the performance of the normal physiology group (n=3) who performed the test with their hand, the maximum Index of Function Score (IFS) in this modified SHAP was 63.7 0.5. Participants in the normal physiology group (n=3) performed the SHAP once with their hand. The group with normal physiology who used the prosthesis (n=10) completed the SHAP three times, once with a fixed wrist, once with a flexion wrist, and once with a DTM wrist. The order of these configurations was randomized to remove learning effects in the SHAP (105) Participants with amputations (n=5) only performed the SHAP with the wrist in the DTM condition. Participants were allowed to rest between tasks if needed. 5.2.6 Compensatory M otions Each participant was equipped with a 2 DO F goniometer to measure shoulder ab/adduction and flexion/extension and a torsionmeter to measure upper arm rotation (Motion Labs inc. Cortlandt, NY ). Testing and calibration of these goniometers was described in the previous chapter. The goniometers were zeroed with the shoulder relaxed, elbow at 90 degrees, and forearm parallel to the sagittal plane. 90 degrees of extension was defined with the arm extended forward so the arm was parallel to the ground, 90 degrees of abduction was defined as the arm exten ded to the side with the arm parallel to the ground, and 90 degrees of rotation was defined by having the participant internally rotate their arm to touch their belly with the elbow at a right angle. Motion was recorded in Labview (National Instruments, A ustin, TX) via a NI DAQ 6008. As in the previous study, recording began before the participant began moving and

PAGE 120

101 continued briefly after they had completed the task. Voltage was converted to angle in Labview using calibration data. Data were post processed in Matlab (The MathWorks, Natick, MA) and filtered using a 4 th order zero phase Butterworth filter (94) Although the shoulder is a complex joint, it was expected that measuring the shoulder in this way would measure the most clinically relevant compens atory motions (106) This study faced the same limita tions with regards to other measurement systems as the previous study. While it would have been preferable to measure additional motions, additional goniometers and an appropriate motion capture system were not available for this study. The metric chosen for comparison was the total range of motion. This metric is relevant to injuries related to compensatory motions (107) Range was us ed instead of maximum because it was difficult to tell from the data when exactly the participant began the task. T rials were cropped based on a deviation of five degrees from the initial position but it was unclear whether initial small motions were part of the SHAP motion or readjustments to a new starting position. By comparing the range, all of these motions were accounted for. Analysis of the shoul der motion was compared within G roup 2 (simulated amputations) and G roup 3 (amputations). People with sim ulated amputations have been found to use the same compensatory motions as people with amputations (49,107) Therefore, it was expected that the results from the intact population and the population with amputations would be comparable.

PAGE 121

10 2 5.2.7 NASA TLX In order to measure how dif ficult it was to control the different wrist configurations, participants were asked to complete the NASA TLX. The NASA TLX is a validated, subjective measure of how challenging the participant found a given task. It was originally developed for aviation b ut has since found many other uses. It is easy to administer, reliable, and has a high sensitivity (108) During the NASA TLX participants rated six factors on a scale from 0 to 100. These factors include: Mental Demand, Physical Demand, Temporal Demand, Performance, Effort, and Frustration. Participants were then handed pieces of paper that list e d each combination (109) The circled titles were used to weight the response scores An easy task, such as completing the SHAP with an unimpaired han d has a low score (20 20, n=3). 5.3 Results A total of 18 people were recruited to take part in this study in three categories: Three people (age 24 1 years) with normal physiology performed the test without a prosthesis; ten people (age 28 6 years) with n ormal physiology performed the test using a simulated prosthesis in three configurations: No wrist (noW), Flexion Extension wrist (FE), and Dart 17 years) with trans radial amputations performed the tes t with a custom made socket and the prosthesis in the DTM configuration (AMP). Three participants from Group 2 returned to confirm that the 22

PAGE 122

103 degree and 27 degree wrist produced similar results. The population with amputations included people with congeni tal limb deficiencies (n=3) and with an acquired amputation (n=2). All participants in this group had previous experience with both body powered and myoelectric prosthetics although only one used a myoelectric prosthesis as their primary device. Further in formation on this group is available in Table 5.2 Table 5 2 : Information on participants with amputations. Type of limb deficit Surgical procedure Years with limb def icit Length of residual limb (cm) Use of myo (hrs/day) Use of BP (hr/day) Age Typical prosthesis Acquired unknown 31 14 0 16 65 TRS prehensor Cong enital NA 37 8 0 16 37 BP active open hand Cong enital NA 28 7 3 4 hrs/ weekday, all weekend All day at w ork 28 TRS prehensor sensor speed hand, BeBionic Cong enital NA 30 14 0 0 30 none Acquired myodesis 46 10 0 16 68 TRS prehensor 5.3.1 SHAP R esults The 1 way ANOVA indicated a significant difference between the SHAP scores (p = 4.9 10 7 ). A follow up Tuk e y honest significant difference (HSD) criterion showed that the SHAP IFS for No Wrist (IFS = 31.4 8.0) was significantly higher than both the Flexion Wrist (IFS =

PAGE 123

104 8.7 3.7) and DTM Wrist (IFS = 17.5 7.5). The DTM wrist produced significantly better I FS than the FE wrist. The population with transradial amputations using the DTM wrist (IFS = 9.0 6.8) was statistically equivalent to the DTM wrist and to the FE wrist. Because this study used a single DOF hand that could not emulate specific grip types, the functionality profiles (scores related to specific grip types) were not considered. Figure 5 10 : Index of function scores showing significant differences between the No Wrist condition and all other c onditions and between the DTM wrist and the FE wrist (black). There was significant difference between the FE and DTM wrists (red). The score for the DTM wrist used by people with transradial amputations was not significantly different than the DTM wrist u sed by people with simulated amputations (blue). Error bars represent standard error. 5.3.2 Shoulder R esults Only ab/adduction was found to be a true measure of compensatory motion. Flexion extension motion in the shoulder was strongly influenced by the length of the device used br inging into question its usefu lness as a measurement of compensatory motion. Likewise, rotation appeared to be a factor of where the SHAP task object was placed relative to the

PAGE 124

105 participant despite attempts to ensure the objects were pl aced directly in front of the prosthesis. The data from participants 3 and 10 under the DTM condition and participant 5 under FEw condition did not record properly and were removed from the calculation. A wide variety of compensatory strategies were observ ed in the participants. Large shoulder motions were common, both in abduction and adduction. Compensatory motions often included shoulder lift which were not measured by the goniometers. Participants also often compensated by bending the trunk instead of m oving the shoulder. A selection of observed compensatory motions are shown in Figure 5.11 Figure 5 11 : Compensatory motions employed by participants during the S HAP. A wide variety of compensatory motions were used, including: large abduction (left), shoulder lift (middle), and torso bending (right). test was used to compare the shoulder ROM between the no wrist condition and the two active wrist co nditions. Intra subject ROM were compared by subtracting the mean ROM from each subject. This was done because of the high variance In most tasks, there was no significant difference between the no wrist and active w rist (p>0.0 5 ). However there was a strong bi modal distribution of compensatory motions. That is, in about half of the cases, the DTM wrist

PAGE 125

106 reduced ROM from the FE and no wrist conditions and in the other half the FE wrist had lower ROM. This may be a n in dication of different compensatory strategies used by the participants. Overall, there appeared to be a trend toward lower ROM in the functional wrists. Results of this analysis can be seen in Figure 5.12 Figure 5 12 : Comparison of shoulder ab/adduction ROM during the SHAP tasks. A test indicated significance (p>0.05) in only a few cases. However, a clear bi modal response is evident in most of the SHAP tasks indicating different compensatory strategies. Error bars represent standard error. 5.3.3 NASA TLX R esults A 1 way ANOVA on the NASA TLX scores indicated significant difference in the scores (p = 0.02 4). However, a Tukey HSD analysis indicated that only the No Wrist and AMP (DTM) conditions were significantly different. This indicates that adding a powered wrist to the prosthesis did not increase the difficulty using the prosthesis.

PAGE 126

107 Figure 5 13 : Comparison o f NASA TLX scores. Only the No Wrist and AMP (DTM) conditions were significantly different. Higher score indicates greater difficulty. 5.4 Discussion The DTM wrist showed important improvement s over no wrist and FE wrist. In the primary measurement, SHAP score the DTM wrist performed better than the FE wrist. T his experiment did not detect any significant overall reduction in shoulder ab/adduction compensatory motions because of the DTM wrist. However, about half of the participants reduced compensatory motion s from the FE to the DTM wrist and nearly all participants had lower compensatory motions when using a functional wrist compared with a fixed wrist The NASA TLX score indicate d that using a DTM wrist is no more challenging than using a device without a wr ist.

PAGE 127

108 5.4.1 SHAP SHAP IFS are related to time to completion, so it was expected that the No Wrist condition, when participants did not need to spend time maneuvering the wrist, would have the highest score. A comparison of the two conditions with wrist devices, FE and DTM, indicated that the DTM wrist does improve the IFS over a traditional FE device This is likely because it allowed for more direct angle of attack to grasp objects. The better angle of attack allowed participants to achieve the proper grip angle with less wrist movement thus reducing the time needed to complete the task. Subjectively, this was observed to be the case but these metrics were not specifically recorded. Based on the PCA in the last chapter, the DTM wrist was expected to recover 45% of the functionality of an anatomical wrist. The IFS achieved using the DTM wrist (17.5) was only 28.1% of the IFS achieved by people performing the limited SHAP with their hand (64.5). It is likely that mechanical limitations (e.g. slow speed to close) a ccounted for this additional loss of function. In comparison, the FE wrist score (8.7) is only 13.6% of the fully functional score. That is, by this metric, the DTM wrist more than doubled the functionality of the FE wrist. It is also likely that some of this reduction from the expected score was a result of delays in the control system. Farrell and Weir identified the optimal controller delay to be between 100 and 125 ms (110) While the prosthetic hand and wrist used in this study typically perfo made this much longer and was more common with larger objects. This was not usually a problem, but occasionally produced delays between 500 1000 ms between muscle

PAGE 128

109 activation and when the participant saw motion. Participants were instructed that they could minimize the overshoot effect by relaxing once they had gripped the object. While a few participants voiced frustration with this at the beginning (one participant described it as ping on a computer with a delay between when you hit the key and when the letter and later reported that they had no issues with this system There was also a small delay between when the participant began contracting to move the set point and when the device began to move. This was due to the fact that the difference between the set point and the current motor position needed to be great enough to produce a PWM that could overcome the mechanical resis tance of the system before mo tion began This delay was typically around 100 ms. Similarly, the time required to hold a co contraction was tailored to each participant and was between 200 and 700 ms. This increased the time to completion in each task and subsequently, the SHAP score. Despite these delays it was believed that this system was appropriate because it resolved the issue of backdrivability and improved grip strength. 5.4.2 Shoulder The goniometers measured few significant, overall differences in sho ulder motion between the different wrist configurations. However, the DTM wrist did reduce compensatory motions for about half of the participants in most of the SHAP tasks. The bi modal nature of the results indicated that a wide range of compensatory mot ions were employed by participants. Because t he goniometers were placed across the glenohumeral

PAGE 129

110 joint in the medial plane to capture shoulder ab/adduction and flexion/extension they did not capture all of the complexities of the compensatory motion s In addition to the complex nature of the individual compensatory motions, there were a variety of compensatory motions available to participants and participants did not use the same strategies. For example, p articipants could achieve similar results by abduc ting the shoulder or by bending the torso. These additional motions were not captured by this test. Th e variability in compensatory motions has been previously identified (95) Further complicating this measure, was the fact that some participants said that the DTM wrist provided a better sta rting position and so they did not activate the wrist. After the test was completed, a ttempts were made to determine whether the participant used shoulder or torso motion to compensate, but this proved impossible to do without corroborating data such as vi deo or motion capture Subjectively, it was clear that participants were using a wide range of compensatory motions that often did not involve shoulder ab/adduction. For example, if compensating would have required a very large abduction, participants ofte n employed shoulder lift or trunk bend instead of abduction. These motions would not have been recorded by the goniometers. Similarly, from 0 to 90 degrees, ab/adduction is prima rily in the gleno humeral joint, but beyond 90 degrees, further abduction occu rs because of scapular thoracic motion (shoulder lift) These additional motions complicate the analysis of the shoulder compensatory motion. However, ab/adduction has been shown to be one of the best indicators of compensatory motion (49,111) Therefore, the combination of improvements in half of the

PAGE 130

111 su bjects and the observation that other compensatory motions were employed while using the active wrists is encouraging. 5.4.3 NASA TLX The similarity in the NASA TLX scores confirmed the secondary hypothesis that the DTM wrist would be no more difficult to contr ol than not having a wrist. Indeed, this study showed that adding a single DOF wrist, whether it uses the DTM or traditional FE, to a prehensor does not increase the difficulty. While the NASA TLX is a validated test, as a subjective measure it has some l imitations. Experimenter bias could affect the results. That is, if the experimenter was positive and talkative with the participant throughout the test, that could lead to the participant indicating lower levels of frustration and mental demand. Conversel y, if the experimenter became stressed during the test along with the participant, the participant would likely rate the test as more frustrating and demanding. Because of this, effort was made to engage all participants equally and avoid influencing the N ASA TLX score. 5.4.4 Corrected A ngle The prosthesis was developed based on the angle determined in section 4.5.2 Therefore, the wrist prosthesis wa s not designed exactly at the DTM angle determined in Section 4.5.3 However, 27 degrees falls within the tolerance determined by the minimum clinically measurable angle in the wrist (98) This angle is also within the range of DTM angl es determined by other studi es This means that, while the DTM angle used in the

PAGE 131

112 prosthesis may not be as precise as it could have been, the results presented here are clinically relevant and within the expected tollerance To confirm that the 27 degree DTM wrist produced similar re sults to a 22 degree DTM wrist, three participants from Group 2 were invited to return to the lab to repeat the SHAP using the 22 degree wrist. SHAP IFS from this follow up test were statistically similar to the results from the 27 degree DTM wrist (Studen test, p>0.90) This indicates that the tolerance was appropriately selected and that the results presented above are accurate. 5.5 Conclusion This study confirmed the hypothesis that a prosthetic wrist that incorporates the DTM would improve functionali ty over a traditional FE wrist. It also confirmed the secondary hypothesis that the DTM wrist would not be more difficult to control than a prosthesis without a wrist. Although the DTM wrist did not reduce overall compensatory shoulder motions, it did sugg est that improvements were made for some participants Based on these findings, there are several next steps that could be taken to improve the understanding of the benefits of including the DTM in the prosthetic wrist. It would be beneficial to compare th e DTM wrist to a wrist rotator. Rotation is usually included in a prosthesis after prehension but there is doubt as to whether this is the most beneficial DOF. Another study that compares DTM and rotation would help settle this debate. Improving the hand a nd the control system wo uld make these differences more clear.

PAGE 132

113 With regards to measuring compensatory motion, a future study should compare FE and DTM wrists motion capture and video to better record the motions. In particular, video recording would allow for the correlation of recorded shoulder motion with visible compensatory strategies and with the duration of the SHAP motions. Motion capture or additional goniometers could record trunk motion and scapular motion to get a more complete measure of the co mpensatory motions (49) This study found that the DTM wrist provides a benefit over the FE wrist and should be used instead of the FE wrist when a wrist device is included. Additiona lly, orienting the wrist in the DTM angle instead of anatomical FE in passive wrists may also provide a benefit. While the DTM wrist did not reduce overall compensatory motions, it did improve functionality and its use in a 2 DOF prosthesis should be stron gly considered over a traditional FE wrist.

PAGE 133

114 C HAPTER VI C ONCLUSIONS AND F UTURE W ORK 6.1 Conclusions The goal of this research was to identify and resolve specific limitations in current prosthetic wrists. The questions specifically addressed are summarized as: Are ADLs that typically utilize an optimal wrist motion (the DTM) more challenging to perform with a prosthesis than non DTM tasks? Can the optimal wrist motion be identified in standardized ADL tasks? Does a prosthetic wrist that incorporates the DTM perform better than a traditional FE wrist while still being easy to use? These three questions were answered by three successive studies that built upon each other to produce a prosthetic device that achieved this goal. 6.1.1 Survey An online survey asked peop le with transradial amputations about their feelings toward performing different tasks with their prosthetic device The results showed that people with ULAs were less satisfied performing tasks that typically use the DTM. It also showed that participants were motivated to learn to use a new device that utilized this motion. Follow up interviews confirmed these sentiments. The results of this survey motivated a study to identify this optimal wrist angle.

PAGE 134

115 6.1.2 DTM angle In this study, able bodied participants pe rformed a series of standardized motions (SHAP) while the ir wrist angles were measured with electrogoniometers. These standardized motions included manipulating abstract objects and performing simulated ADLs. Principal component analysis was performed on t he wrist angles and revealed that FE and RU are coupled in a plurality of these motions. It also identified the ideal angle of rotation for this motion from the anatomical FERU plane to be 22 degrees offset from the anatomical FERU plane These motions wer e similar to the DTM angles identified in previous studies but this was the first time they had been measured with prosthetic design specifically in mind. This study also suggested that a two DOF wrist incorporating both the DTM and rotation could provide about three fourths of the functionality of the anatomical wrist. These results informed the design of a new prosthetic wrist that incorporated the DTM. 6.1.3 DTM P rosthetic W rist A new prosthetic wrist was designed and built to address the limitations reported in the first study by using the DTM angle found in the second study. People with simulated and genuine ULAs performed the SHAP with the new device. The DTM wrist was compared to a traditional FE wrist and to a device without a wrist. The DTM wrist out perf ormed the FE wrist on the SHAP and was no more difficult to use than the device without a wrist. The DTM wrist did reduce overall compensatory motions, though it did reduce compensatory motions in about half of the participants. The bimodal response sugges t that participants

PAGE 135

116 were using a wide range of compensatory motions in addition to shoulder abduction including shoulder lift and torso bend This study showed that a prosthetic wrist that incorporates the DTM has great promise for improving functionality and should be favored over a FE wrist in 2 DOF prosthetic systems. 6.2 Future work The results of this thesis are promising but there is certainly more work that could be done and improvements to be made. A few ideas for extensions of these studies and new st udies based on this work are presented here. 6.2.1 Rotator vs DTM One of the interesting secondary findings of the wrist measurement studies was that the majority of the wrist motion could be accounted for by the DTM and rotation. The survey results also indicat ed that more people were interested in wrist FE and RU than rotation and other studies have called into question whether rotation is really the most useful wrist DOF. A new study could compare the DTM wrist to a commercial wrist rotator and to a 2 DOF wris t that includes both the DTM and rotation. This study should follow the protocol described in section 5.2 and use the same EMG control system and hardware as the study described in this thesis. A rotator would need to be develo ped or purchased that used a servo motor like those used in this design that could be controlled with the penny boards. The LabView code would work the same as with the DTM wrist, although it would have to be expanded if the researcher was interested in co ntrolling three DOF at once. This

PAGE 136

117 study provide further insight into how the DTM could be used in a prosthesis and may change the prioritization of DOF in commercial wrists. 6.2.2 Correction of the DTM A ngles After the DTM wrist prosthesis was designed and test ed it was discovered that the angle of 27 degrees was based on an incorrect assumption. Although this was within tolerance and a small follow up study indicated no difference between the two angles a full follow up study could be done to confirm these res ults The proper angle can be checked by redoing the last study with the adapter plate set at 22 degrees and the wedge remade to meet the new angles. The same protocol as described in section 5.2 can be used. The major challeng e to completing this study will be recruiting a new set of nave participants with upper limb amputations. 6.2.3 Assessment of C ompensatory M otions The measurements of the shoulder compensatory motions were based on a review of other literature on compensatory motions and limited by the number of goniometers available for this study. This meant that measuring the shoulder was the least robust part of the experiments described here. While the approach used here provided interesting insight into the potential bene fits of the DTM wrist in reducing compensatory motions, the complexity of the shoulder means that these measurements did not tell the whole story. Future work should include video recording and wiring the SHAP timer to LabView to record a clear start and stop condition. This would allow for two improvements in assessing

PAGE 137

118 the shoulder motions. First, it would allow the researcher to identify the rest, reach, grasp, and final rest components of the SHAP tasks as in the second experiment. Second, it would prov ide a better way to time normalize the motion. The start and stop conditions could be adjusted depending on the tactics of the participants. For example, many participants found it easier to contract their muscles and position the prosthesis before reachin g to the object. Video would allow the researcher to recognize this technique and move the start time to reflect the actual start of motion. These features would allow the researcher to create an ensemble average of the compensatory motions. Borrowing too ls from gait analysis, the researcher could then calculate the root mean squared difference between compensatory motions of people using the prosthesis and people performing the SHAP with their intact limb. This validated tool gives better insight into ho w well each type of device allowed the participant to In particular, it would allow for both an overall assessment of improvement, as described here, and specific insight into how the different devices change how compensatory motions are performed. Attempts were made in this dissertation to perform this type of analysis, but without video providing these key references, it was not possible. One of the chief limitations in measuring compensatory motion in this thesis was the l imited number of DOF that could be measured. M otion capture could potentially identify both shoulder lift and trunk bend and could be very useful for determining whether the DTM wrist reduces compensatory motions that may lead to overuse injuries and socia l difficulties (107,112) In addition, recordi ng EMG from muscles in the shoulder and trunk

PAGE 138

119 may give insight into how much work is being done by the participant under each wrist condition. Individual, c ompensatory motions have been studied using a ccelerometers (107) and motion capture (49) To date, no overall measure of compensatory motions has been described. A ngles are not vectors, so they cannot b e simply added together to produce a Therefore, any future study accessing the benefits of a new prosthesis will need to choose the priority of each joint motion. Previous studies have identified sh oulder abduction and trunk bend as the most important but as other motions such as elbow FE and shoulder lift should also be considered This future study could follow the protocol described in section 5.2 with the addition of video and, potentially, mo tion capture. While the Optotrack system described in section 4.1.1 was ineffective for measuring the wrist, it may prove effective here if it can be made to work with the existing labview code. 6.2.4 Combination of W rist and H and P osition in P ostural C ontroller Finally, there has been much work done by Santello where PCA is used to reduce the dimensionality of the hand positions (97) This work could potential ly be expanded to include the wrist angles. Combining wrist and hand position with PCA could lead to a postural controller that would allow for simple control of the hand and wrist as one system. This concept was presented at the Rocky Mountain American So ciety of Biomechanics in 2016 (113) Though it was not used in this thesis, the hand position data necessary to perform this test was recorded using a Cyber Glove II system w hile testing for the optimal

PAGE 139

120 wrist angle in specific aim 2. Performing PCA or other analysis on the combined hand and wrist data may reveal a novel way to control these two systems together. 6.3 Final T houghts People with upper limb amputations face a number of social and economic challenges stemming from a lack of functionality in their prostheses. Current prosthetic devices have come a long way in the past century, but are still not enabling users to meet these challenges. Basing new designs on input from u sers and focusing on improving functionality while keeping the device simple to use has the greatest potential to solve these issues. From the beginning, this work took a human centered design approach to address one of the major limitations of prostheti c devices: lack of functionality due to limited control inputs. Users of upper limb prosthetics were involved in forming the initial design criteria, consulted throughout the design process, and tested the final device They provided valuable insight that ultimately produced a device that alleviate s some of the challenges they face in work and social life. The final product of this dissertation is a prosthetic wrist that improves functionality while not being difficult to use. It was the result of several i terations of wrist design based on the desires of users and a study of the anatomical wrist. The DTM wrist is a mechanical solution to what is, essentially, a control problem. It provides the major component of the three DOF of the wrist while only requiri ng the user to control a single additional DOF. This dissertation showed that the DTM wrist can improve functionality by reducing the dimensionality of the wrist specifically in tasks that people with ULAs find challenging. It

PAGE 140

121 has the potential to improv have historically been difficult for people with ULAs to make it easier to perform ADLs, and to improve social interactions. Future work in this vein will continue to improve the prosthesis an d lead to even better solutions to the challenges faced by people with ULAs.

PAGE 141

122 R EFERENCES 1. Ziegler Graham K, MacKenzie EJ, Ephraim PL, Travison TG, Brookmeyer R. Estimating the prevalence of limb loss in th e United States: 2005 to 2050. Arch Phys Med Rehabil. 2008 Mar;89(3):422 9. 2. Raichle KA, Hanley MA, Molton I, Kadel NJ, Campbell K, Phelps E, et al. Prosthesis use in persons with lower and upper limb amputation. J Rehabil Res Dev. 2008;45(7):961 972. 3. A, Doyle A, Desmond D. Environmental barriers, activity limitations and participation restrictions experienced by people with major limb amputation. Prosthet Orthot Int. 2011 Sep;35(3):278 84. 4. Risher P, Amorosi S. The 19 98 N.O.D./Harris Survey of Americans with Disabilities. 1998. 5. U.S. Census Bureau; SELECTED ECONOMIC CHARACTERISTICS FOR THE CIVILIAN NONINSTITUTIONALIZED POPULATION BY DISABILITY STATUS 2009 2011 American Community Survey 3 Year Estimates; generated b y Matthew Davidson; using American FactFinder; ; (28 February 2013). 7. Ali M, Schur L, Blanck P. What types of jobs do people with disabilities want? J Occup Rehabil. 2011 Jun;21(2):199 210. 8. Jang CH, Yang HS, Yang HE, Lee SY, Kwon JW, Yun BD, et al. A survey on activities of daily living and occupations of upper extremity amputees. Ann Rehabil Med. 2011 Dec;35(6):907 21. 9. Sears HH, Iversen E, Archer S, Linder J, MacDonald JR. Evaluation studies of new electric terminal devices. In: Proceedings of the 2005 MyoElectric Controls/P owered Prosthetics Symposium. Fredericton, New Brunswick, Canada; 2005. p. 1 6. 10. Fougner A, Stavdahl Kyberd PJ, Losier YG, Parker PA, Member S. Control of Upper A Review. IEEE Tra ns NEURAL Syst Rehabil Eng. 2012;20(5):663 77. 11. Weir R. Design of Artificial Arms and Hands for Prosthetic Applications. In: Kutz M, editor. Standard Handbook of Biomedical Engineering & Design. New York, NY: McGraw Hill; 2003. p. 32.1 32.61. 12. Ba jaj NM, Spiers AJ, Dollar AM. State of the art in prosthetic wrists: Commercial and research devices. IEEE Int Conf Rehabil Robot. 2015;2015 Septe:331 8. 13. Biddiss E, Beaton D, Chau T. Consumer design priorities for upper limb prosthetics. Disabil Reha bil Assist Technol. 2007;2(6):346 57.

PAGE 142

123 14. Weir RF ff., Sensinger J. The Design of Artificial Arms and Hands for Prosthetic Applications. In: Kutz M, editor. Biomedical Engineering & Design Handbook. New York, NY: McGraw Hill; 2009. p. 537 98. 15. Segil J, Weir R, Reamon D. Design Of A Myoelectric Controller For A Multi Dof Prosthetic Hand Based On Principal Component Analysis. In: Proceedings of the 2011 MyoElectric Controls/Powered Prosthetics Symposium. Fredericton, New Brunswick, Canada; 2011. p. 3 6 16. Segil JL, Huddle S, Weir RF. Functional Assessment of a Myoelectric Postural Controller and Multi Functional Prosthetic Hand by Persons with Trans Radial Limb Loss. IEEE Trans Neural Syst Rehabil Eng. 2016; 17. Oskoei MA, Hu H. Myoelectric contro l systems A survey. Biomed Signal Process Control. 2007 Oct;2(4):275 94. 18. Troyk PR, DeMichele G a, Kerns D a, Weir RF. IMES: an implantable myoelectric sensor. In: Annual International Conference of the IEEE Engineering in Medicine and Biology Society Lyon, France; 2007. p. 1730 3. 19. Palmer A, Murphy D, Glisson R. Functional wrist motion: A biomechanical study. J Hand Surg Am. 1985;10A(1):39 46. 20. Crisco JJ, Heard WMR, Rich RR, Paller DJ, Wolfe SW. The mechanical axes of the wrist are oriented obliquely to the anatomical axes. J Bone Joint Surg Am. 2011 Jan 19;93(2):169 77. 21. Calfee RP, Leventhal EL, Wilkerson J, Moore DC, Akelman E, Crisco JJ. Simulated Radioscapholunate Fusion Alters Carpal Kinematics While Preserving Dart Motio n. J Hand Surg Am. 2008;33(4):503 10. 22. Mcfarland L V, Winkler SLH, Otr L, Heinemann AW, Jones M, Esquenazi A. Unilateral upper limb loss: Satisfaction and prosthetic device use in veterans and servicemembers from Vietnam and OIF/OEF conflicts. J Rehab il Res Dev. 2010;47(4):299 316. 23. With Decreased Wrist Motion. J Hand Surg Am. 2008;33A:485e1 485e9. 24. Heinemann AW, Gershon R, Fisher WP. Development and Appl ication of the Orthotics and Prosthetics User Survey: Applications and Opportunities for Health Care Quality Improvement. 2006;(6). 25. Hudak PL, Amadio PC, Bombardier C. Development of an Upper Extremity Outcome Measure: The DASH (Disabilities of the Ar m, Shoulder, and Head). 1996;29:602 8. 26. Orr CM, Leventhal EL, Chivers SF, Marzke MW, Wolfe SW, Crisco JJ. Studying primate carpal kinematics in three dimensions using a computed tomography based markerless registration method. Anat Rec. 2010;293(Janua ry):692 701.

PAGE 143

124 27. Moritomo H, Apergis EP, Herzberg G, Werner FW, Wolfe SW, Garcia elias M. 2007 Called Dart Throwing Motion of the Wrist. J Hand Surg Am. 2007;32A(9):1447 53. 28. Light CM, Chappell PH, Kyberd PJ. Establishing a standardized clinical assessment tool of pathologic and prosthetic hand function: Normative data, reliability, and validity. Arch Phys Med Rehabil. 2002 Jun;83(6):776 83. 29. Moore KL, Agur AM, Dalle y AF. Essential Clinical Anatomy. 4th ed. Moore KL, editor. Philadelphia, PA: Lippincot Williams & Wilkins; 2011. 30. Snyder PJ, Whitaker H a. Neurologic heuristics and artistic whimsy: the cerebral cartography of Wilder Penfield. J Hist Neurosci. 2013;2 2(3):277 91. 31. Charles SK, Hogan N. Dynamics of wrist rotations. J Biomech. Elsevier; 2011 Feb 24;44(4):614 21. 32. of function. New York, NY: Macmillan; 1950. 248 p. 33. Young RW. Evolution of the human hand: the role of throwing and clubbing. J Anat. 2003 Jan;202(1):165 74. 34. Penfield W, Boldry E. Somatic Motor and Sensory Representation in the Cerebral Cortex of Man as Studied by Electrical Stimulation. Bra in. 1937;60(4):389 443. 35. Montagnani F, Controzzi M, Cipriani C. Is it finger or wrist dexterity that is missing in current hand prostheses? 2015;21(c). 36. Sears HH, Iversen EK, Archer S. Evaluation of a new flexion wrist integrated with electric ha nd. In: Proceedings of the 2002 MyoElectric Controls/Powered Prosthetics Symposium. Fredericton, New Brunswick, Canada; 2002. 37. Kaufmann R a, Pfaeffle HJ, Blankenhorn BD, Stabile K, Robertson D, Goitz R. Kinematics of the midcarpal and radiocarpal join t in flexion and extension: an in vitro study. J Hand Surg Am. 2006 Sep;31(7):1142 8. 38. Metcalf CD, Notley S V, Chappell PH, Phappell H, Burridge JH, Yule VT. Validation and application of a computational model for wrist and hand movements using surfac e markers. IEEE Trans Biomed Eng. 2008 Mar;55(3):1199 210. 39. Rohde R, Joseph C, Wolfe S. The Advantage of Throwing the First Stone: How Understanding the Evolutionary Demands of Homo sapiens Is Helping Us Understand Carpal Motion. J Am Acad Orthop Surg 2010;18(1):51 8. 40. Wiesendanger M. Manual dexterity and the making of tools an introduction from an evolutionary perspective. Exp Brain Res. 1999;128:1 5. 41. Crisco JJ, Coburn J, Moore D, Akelman E, Weiss A P, Wolfe S. In Vivo Radiocarpal Kinema 40. 42. Bugden B. A proposed method of goniometric measurement of the dart throwers motion. J Hand Ther. Hanley & Belfus; 2013;26(1):77 9; quiz 80.

PAGE 144

125 43. Dillingham T, Pezzin L, Macke nzie E. Limb Amputation and Limb Deficiency. South Med J. 2002;95(8). 44. Ouellette EA. Surgical, Prosthetic, and Rehabilitation Principles. In: Bowker H, Michael J, editors. Atlas of Limb Prosthetics. 2nd ed. Rosemont, IL: American Academy of Orthopedic Surgeons; 2002. 45. Ovadia S, Askari M. Upper Extremity Amputations and Prosthetics. Semin Plast Surg. 2015 Feb 4;29(1):055 61. 46. Kuiken T. Consideration of nerve muscle grafts to improve the control of artificial arms. Technol Disabil. 2003;15(2):1 05 11. 47. Aaron RK, Herr HM, Ciombor DM, Hochberg LR, Donoghue JP, Briant CL, et al. Horizons in Prosthesis Development for the Restoration of Limb Function. J Am Acad Orthop Surg. 2006;14(10):198 204. 48. Zinck A, Kyberd PJ, Hill W, Bush G, Stavdahl Biden E, et al. A study of the use of compensation motions when using prosthetic wrists. In: Proceedings of the 2008 MyoElectric Controls/Powered Prosthetics Symposium. New Brunswick, Canada; 2008. 49. Carey SL, Jason Highsmith M, Maitland ME, Dubey R V. Compensatory movements of transradial prosthesis users during common tasks. Clin Biomech. Elsevier Ltd; 2008;23(9):1128 35. 50. Kyberd PJ, Lemaire ED, Scheme E, MacPhail C, Goudreau L, Bush G, et al. Two degree of freedom powered prosthetic wrist. J Rehabil Res Dev. 2011;48(6):609. 51. Kidd PS, McCoy C, Steenbergen L. Repetitive Strain Injuries in Youth. J Am Acad Nurse Pract. 2000 Oct 1;12(10):413 26. 52. Gambrell C. Overuse syndrome and the unilateral upper limb amputee: consequences and prevent ion. JPO J Prosthetics Orthot. 2008;20(3):126 32. 53. Wright V. Prosthetic Outcome Measures for Use With Upper Limb Amputees: A Systematic Review of the Peer Reviewed Literature, 1970 to 2009. In: JPO Journal of Prosthetics and Orthotics. 2009. p. P3 63. 54. Smith DG, Michael JB, Bowker J, editors. Atlas of amputations and limb deficiencies: surgical, prosthetic, and rehabilitation principles. 3rd ed. Rosemont, IL: American Ascociation of Orthopaedic Surgeons; 2004. 55. Childress DS. Historical aspect s of powered limb prosthetics. Clin Prosthetics Orthot. 1985;(9):2 13. 56. Wright TW, Hagen AD, Wood MB. Prosthetic usage in major upper extremity amputations. J Hand Surg Am. 1995;20(4):619 22. 57. RSLSteeper. RSL Steeper cosmetic hand [Internet]. [ci ted 2016 Jan 1]. Available from: http://rslsteeper.com/products/prosthetics/products/upper_limb/passive/hands_cos metic

PAGE 145

126 58. Ritchie S, Wiggins S, Sanford A. Perceptions of cosmesis and function in adults with upper limb prostheses: a systematic literature review. Prosthet Orthot Int. 2011;35(4):332 41. 59. e NABLE project [Internet]. [cited 2016 Jan 1]. Available from: http://enablingthefuture.org/ 60. Radocy B. Upper Limb Prosthetics for Sports and Recreation. In: Care of the Combat Amputee, Textbooks o f Military Medicine. 1st ed. Walter Reed Army Medical Center Borden Institute; 2009. p. 641 68. 61. Gwynne G. Manual of Upper Extremity Prosthetics. 2nd ed. Los Angeles: University of Southern California; 1958. 33 69 p. 62. Alley RD, Sears HH. Powered Upper Limb Prosthetics in Adults. In: Powered Upper Limb Prostheses. Berlin, Heidelberg: Springer Berlin Heidelberg; 2004. p. 117 45. 63. Kyberd PJ. The influence of control format and hand design in single axis myoelectric hands: assessment of functiona lity of prosthetic hands using the Southampton Hand Assessment Procedure. Prosthet Orthot Int. 2011 Sep;35(3):285 93. 64. Kyberd PJ. The influence of passive wrist joints on the functionality of prosthetic hands. Prosthet Orthot Int. 2011;36(1):33 8. 65 Sears HH, Iversen E, Archer S, Jacobs T. Wrist Innovations To Improve Function of Electric Terminal Devices. In: Proceedings of the 2008 MyoElectric Controls/Powered Prosthetics Symposium. Fredericton, New Brunswick, Canada; 2008. p. 13 6. 66. Weir R, Mitchell M, Clark S, Puchhammer G, Kelley K, Haslinger M, et al. New multifunctional prosthetic arm and hand systems. In: Proceedings of the 29th annual international conference of the IEEE EMBS. Cite Internationale, Lyon, France; 2007. p. 4359 60. 67. Atkins DJ, Heard DCY, Donovan WH. Epidemiologic Overview of lndividuals with Upper Limb Loss and Their Reported Research Priorities. J Prosthetics Orthot. 1996;8(1):2. 68. Landry JS. Optimal Fixed wrist alignment for below elbow, powered, prosthetic hand s. University of New Brunswick; 1993. 69. LeBlanc MA. Innovation and improvement of body powered arm prostheses: A first step. Clin Prosthet Orthot. 1985;9(1):13 6. 70. Belter JT, Segil JL, Dollar AM, Weir RF. Mechanical design and performance specific ations of anthropomorphic prosthetic hands: a review. J Rehabil Res Dev. 2013 Aug;50(5):599 618. 71. Waldera KE, Heckathorne CW, Parker M, Fatone S. Assessing the prosthetic needs of farmers and ranchers with amputations. Disabil Rehabil Assist Technol. 2013 May;8(3):204 12.

PAGE 146

127 72. Office of Special Education and Rehabilitative Services. National Institute on Disability and Rehabilitation Research Disability and Rehabilitation Research Projects and Centers Program Disability Rehabilitation Research Project s (DRRPs), Rehabilitation. Fed Regist. 2008;73:6132 6146. 73. Kestner S. Defining the Relationship between Prosthetic Wrist Function and Its Use in Performing Work Tasks and Activities of Daily Living. JPO J Prosthetics Orthot. 2006;18(3):80 6. 74. Bid diss E a, Chau TT. Upper limb prosthesis use and abandonment: a survey of the last 25 years. Prosthet Orthot Int. 2007;31(3):236 57. 75. Jones and Bartlett; 2011. 76. Sc heme E, Englehart K. Electromyogram pattern recognition for control of powered upper limb prostheses: State of the art and challenges for clinical use. J Rehabil Res Dev. 2011;48(6):643. 77. Miller LA, Swanson S, Otr L. Summary and Recommendations of the of the Science Conference on Upper Limb Prosthetic Outcome Measures. In: Journal of Prosthetics & Orthotics. 2009. p. 83 9. 78. Rubin J, Chisnell D. Handbook of Usability Testing: Howto Plan, Design, and Conduct Effective Tests. 2nd ed. indianapolis, IN: Wiley Publishing, Inc.; 2008. 384 p. 79. Burger H, Franchignoni F, Heinemann AW, Kotnik S, Giordano A. Validation of the o rthotics and prosthetics user survey upper extremity functional status module in people with unilateral upper limb amputation. J Rehabil Med. 2008;40(5):393 9. 80. Lavrakas PJ, editor. Encyclopedia of survey research methods. 1st editio. Thousand oaks, C A: SAGE publications Inc.; 2008. 81. Brown SM, Mcbride G, Collingridge DS, Butler JM, Kuttler KG, Hirshberg EL, et al. Validation of the Intermountain patient perception of quality ( PPQ ) survey among trospective validation study. BMC Health Serv Res. Biomed Central; 2015;15(155):1 7. 82. Garca de Ybenes Prous MAJ, Rodrguez Salvans F, Carmona Ortells L. [Validation of questionnaires]. Reumatol Clin. Elsevier; 2009 Jan 1;5(4):171 7. 83. Stewart D W, Shamdanani PN, Rook DW. Focus Groups: Theory and Practice. 2nd ed. Thousand oaks, CA: SAGE publications Inc.; 2007. 188 p. 84. Lieber RL. Statistical significance and statistical power in hypothesis testing. J Orthop Res. 1990;8(2):304 9. 85. Davis Iii RB, 'unpuu S, Tyburski D, Gage JR. A gait analysis data collection and reduction technique. Hum Mov Sci. 1991;10(5):575 87. 86. Wang PT, King CE, Do AH, Nenadic Z. A durable, low cost electrogoniometer for dynamic measurement of joint trajectories. M ed Eng Phys. 2011 Jun;33(5):546 52.

PAGE 147

128 87. Overview of the SHAP [Internet]. 2016. Available from: http://www.shap.ecs.soton.ac.uk/about.php 88. Adams J, Hodges K, Kujawa J, Metcalf C, Harth A. Test retest Reliability of the Southampton Hand Assessment Proc edure. Int J Rehabil Res. 2009;32:32 3. 89. Overview of the SHAP. 2016. 90. Demsar U, Harris P, Brunsdon C, Fotheringham a. S, McLoone S. Principal Component Analysis on Spatial Data: An Overview. Ann Assoc Am Geogr. 2012;(July). 91. Daffertshofer A Lamoth CJC, Meijer OG, Beek PJ. PCA in studying coordination and variability: A tutorial. Clin Biomech. 2004;19(4):415 28. 92. Shlens J. A tutorial on principal component analysis: derivation, discussion and singular value decomposition. Online Note ht tpwww snl salk edushlenspubnotespca pdf. 2003;2:1 16. 93. Boas ML. Mathematical Methods in the Physical Sciences. Hoboken, NJ: John Wiley & Sons, Inc.; 1983. 743 p. 94. Murgia A, Kyberd PJ, Chappell PH, Light CM. Marker placement to describe the wrist movements during activities of daily living in cyclical tasks. Clin Biomech (Bristol, Avon). 2004 Mar;19(3):248 54. 95. Major MJ, Stine RL, Heckathorne CW, Fatone S, Gard S a. Comparison of range of motion and variability in upper body movements between transradial prosthesis users and able bodied controls when executing goal oriented tasks. J Neuroeng Rehabil. 2014;11(1):132. 96. Winter DA. Biomechanics and Motor Control of Human Movement. 4th ed. Waterloo, Ontario, Canada: John Wiley & Sons, Inc.; 200 9. 307 p. 97. Santello M, Flanders M, Soechting JF. Postural Hand Synergies for Tool Use. J Neurosci. 1998;18(23):10105 15. 98. Tajali SB, MacDermid JC, Grewal R, Young C. Reliability and Validity of Electro Goniometric Range of Motion Measurements in Patients with Hand and Wrist Limitations. Open Orthop J. 2016;10(519):190 205. 99. Horger MM. The Reliability of Goniometric Measurements of Active and Passive Wrist Motions. Am J Occupalional Ther. 1990;44(4):342 8. 100. Weir RF, Grahn E. A Myoelectri cally Controlled Prosthetic Hand for Transmetacarpal Amputations. In: Proceedings of the 2002 MyoElectric Controls/Powered Prosthetics Symposium. New Brunswick, Canada; 2002. p. 21 4. 101. Montagnani F, Controzzi M, Cipriani C, Schultz AE, Kuiken TA, Osk oei MA, et al. Independent Long Fingers are not Essential for a Grasping Hand. Sci Rep. Nature Publishing Group; 2016;6(October):35545. 102. Magee DJ. Orthopedic Physical Assessment. 5th ed. St. Lous, MO: Saunders; 2007. 1152 p.

PAGE 148

129 103. Peizer E, Wright D W, Mason C, Pirrello C. Guidelines for standards for externally powered hands. Bull Prosthet Res. 1969;10 12(Fall):118 55. 104. Davidson M. Principal components of the wrist motion in the SHAP. In: Rocky Mountain American Society of Biomechanics. Estes P ark, CO; 2015. 105. Vasluian E, Bongers RM, Reinders messelink HA, Burgerhof JGM, Dijkstra PU, Sluis CK Van Der. Original Report of Learning Effects of Repetitive Administration of the Southamption Hand Assessment Procedure in Novice Prosthetic Users. J Rehabil Med. 2014;46(8):788 97. 106. stlie K, Lesj IM, Franklin RJ, Garfelt B, stlie K, Lesj IM, et al. Disability and ills and the actual use of prostheses in activities of daily life Prosthesis use in adult acquired major uppe. Disabil Rehabil Assist Technol. 2012;3107. 107. Mell AG, Childress BL, Hughes RE. The effect of wearing a wrist splint on shoulder kinematics d uring object manipulation. Arch Phys Med Rehabil. 2005;86(8):1661 4. 108. Hart SG. NASA TASK LOAD INDEX ( NASA TLX ); 20 YEARS LATER. Proc Hum Factors Ergon Soc Annu Meet. Moffett Field, CA; 2006;50(9):904 8. 109. NASA Task Load Index (TLX) v1.0 Paper and Pencil Package. Moffett Field, CA: Human Performance Research Group NASA Ames Research Center; 110. Farrell TR, Weir RF. The Optimal Controller Delay for Myoelectric Prostheses. IEEE Trans NEURAL Syst Rehabil Eng. 2007;15(1):111 8. 111. Zinck AL. Investigation of Compensatory Movements in Prosthesis Users and the Design of a Novel Wrist. Vol. MR74324. 2009. 316 p. 112. Bertels T, Schmalz T, Ludwigs E. Objectifying the Functional Advantages of Prosthetic Wris t Flexion. J prosthetics Orthot. 2009;21(2):74 8. 113. Davidson M. Toward a postural controller for the wrist. In: Rocky Mountain American Society of Biomechanics2. Estes Park, CO; 2016.