University of Southern California Dissertations and Theses

Arm choice post-stroke

(USC Thesis Other)

Arm choice post-stroke

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Arm Choice Post-Stroke by Sujin Kim A dissertation presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In partial fulfillment of the requirements for the degree DOCTOR OF PHILOSOPHY BIOKINESIOLOGY August 2016 Doctoral Committee: Associate Professor Nicolas Schweighofer, Chair Professor James Gordon, Professor Carolee Winstein, Associate Professor John Monterosso, Assistant Professor James Finley i DEDICATION To my God ii ACKNOWLEDGMENTS I would like to thank my Heavenly Father for his love, grace, and guidance that he provided me during this long journey. I deeply thank to my academic advisor Dr. Nicolas Schweighofer for his continuous support, encouragement, and immense knowledge. He taught me not only scientific knowledge and research skills, but also a positive attitude toward difficult circumstances that I encountered during my PhD. I also thank Dr. James Gordon for his generous understanding and scientific advices. Whenever I faced challenges, he always provided me good solutions. I am also grateful to have Dr. Carolee Winstein as my committee. She is my role model to be a good physical therapist, scientist, and researcher. She has guided me to see the big picture of my dissertation and helped me specify the goals, hypotheses, and clinical implications. I also sincerely thank Dr. John Monterosso who provided me great insight into psychological aspects of arm choice in patients with stroke, and thank Dr. James Finley, who gave me an opportunity to learn about metabolic effort. In addition, I would like to give my sincere thanks to Dr. Cheol Han, my collaborator. We have worked together since I started my Ph.D program, and he has helped me develop the computational arm choice model for individuals with stroke. I thank all members of Computational Neuro-Rehabilitation Laboratory (CNRL): Dr. Hyeshin Park for her help to modify the BART system, Dr. Amar Bains and Chunji Wang for their iii excellent instruction in MATLAB programming, and Dr. Youngmin Oh for the stimulating discussion. In addition, I thank all the members of Motor Behavior and Neurorehabilitation Laboratory, Neuroplasticity and Imaging Laboratory, and Locomotor Control Laboratory. They were my friends, colleagues, and teachers who have supported me to complete this long journey successfully. I really appreciate their help and I will miss them a lot. I would like to express my special appreciation to father Daeje Choi, Roberto, and my friends in St. Agnes Catholic Church for their spiritual supports. Finally, I would like to give my sincerest gratitude and love to my family. Words cannot express how grateful I am to my parents, two sisters, and their sacrifices that have made for me. iv TABLE OF CONTENTS DEDICATION ................................................................................................................... i ACKNOWLEDGMENTS ................................................................................................ ii TABLE OF CONTENTS ................................................................................................ iv LIST OF FIGURES ........................................................................................................ vii LIST OF TABLES ........................................................................................................... xi ABSTRACT ..................................................................................................................... xii CHAPTER 1: Introduction .............................................................................................. 1 1.1. Statement of the problem ......................................................................................... 1 1.2 Factors influencing arm choice in post-stroke and nondisabled people ................... 3 1.3. Assessment tools for use, non-use, and performance of upper extremity ............... 6 1.4. Overview of dissertation structure ......................................................................... 11 Chapter 2: Quantifying Arm Non-use in Individuals Post-stroke ............................. 13 2.1. Introduction ............................................................................................................ 13 2.2. Method ................................................................................................................... 15 2.2.1. Participants ...................................................................................................... 15 2.2.2. Actual Amount of Use Test (AAUT).............................................................. 17 2.2.3. Bilateral Arm Reaching Test (BART) ............................................................ 17 2.2.4. BART Dependent Measures ........................................................................... 21 2.2.5. Statistical Analyses ......................................................................................... 23 2.3. Results .................................................................................................................... 25 2.3.1. Demographic and stroke characteristic data ................................................... 25 2.3.2. Measuring use, performance, and non-use with the AAUT test ..................... 25 2.3.3. Measuring use, performance, and non-use with BART .................................. 28 2.3.4. BART Test-retest Reliability .......................................................................... 31 2.3.5. External Validity of BART use sBART, performance cBART, and non-use nuBART .......................................................................................................... 33 2.4. Discussion .............................................................................................................. 35 Appendix A ................................................................................................................... 40 v Chapter 3: Objective Assessments of Upper Extremity Performance, Use and, Non-use After Stroke ................................................................................... 42 3.1. Introduction ............................................................................................................ 42 3.2. Method ................................................................................................................... 44 3.2.1. Participants ...................................................................................................... 44 3.2.2. Experimental Setup and Task ......................................................................... 45 3.2.3. Measures ......................................................................................................... 49 Clinical Assessment ................................................................................................... 49 BART dependent measures ....................................................................................... 49 3.2.4. Statistical Analysis .......................................................................................... 51 3.3. Results .................................................................................................................... 51 3.3.1. Patient demographics ...................................................................................... 51 3.3.2. Performance and Use in the no time, medium, and fast conditions ................ 54 3.3.3. Performance and Arm Choice Pattern on the WorkSpace in Fast Condition . 56 3.3.4. Validity of time-based BART for Measuring Use, Performance, and Non-use of the affected arm in the Fast Condition ........................................ 58 3.3.5. Test-Retest Reliability .................................................................................... 60 3.4. Discussion .............................................................................................................. 62 Chapter 4: Habitual versus Adaptive Use of Affected Arm in Right and Left Hemiparesis .................................................................................................. 68 4.1. Introduction ............................................................................................................ 68 4.2. Methods.................................................................................................................. 71 4.2.1. Participants ...................................................................................................... 71 4.2.2. Experimental setup.......................................................................................... 73 4.2.3. Experimental Task .......................................................................................... 73 4.2.4. Measurements for Arm Choice, Movement Duration, Effort Estimation, and Success ..................................................................................................... 77 4.2.5. Analysis for Arm Choice, Movement Duration, Effort Estimation, and Success ............................................................................................................ 80 4.2.6. Arm Choice Model ......................................................................................... 81 4.2.6. Effect of Failure on Arm Choice .................................................................... 84 4.3. Results .................................................................................................................... 85 4.3.1. Participant demographics and clinical data..................................................... 85 4.3.2. Arm Choice ..................................................................................................... 85 4.3.3. Movement Duration ........................................................................................ 88 4.3.4. Effort ............................................................................................................... 91 4.3.4. Success Rates .................................................................................................. 92 4.3.5. Difference between Arms in Movement Duration, Effort, and Success Rate 93 4.3.6. Predicting Arm Choice: Effort and Success Differentially Influence Arm Choice ............................................................................................................. 95 4.3.7. Model accuracy and simulation of choice data for the models ....................... 98 4.3.8. Habitual vs. Adaptive arm choice strategies between RH and LH ................. 99 vi 4.4. Discussion ............................................................................................................ 106 Chapter 5. Summary and future study ....................................................................... 112 References ...................................................................................................................... 116 vii LIST OF FIGURES Figure 2.1. Measuring arm use with the Bilateral Arm Reaching Test: the home position is identified by the green circle and a target by the white circle. For each trial, participants were instructed to reach to the target with their choice of hand using the index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. ......................................................................................................... 19 Figure 2.2. Computing nonuse with BART in 1 session over the 2D reaching work space for right-affected poststroke participant, ID2, in session 3: A. Constrained use (performance) probability for ID2. B. Average spontaneous use probability for right-handed nondisabled participants (normative hand use). C. Constrained use (performance) probability after masking with normative hand use of panel (B) poststroke. D. Spontaneous use probability. E. Nonuse probability for ID2. Color coding: red = 100% use of the paretic arm (right arm for healthy controls), blue = 0% use of the paretic arm. The indifference line, indicated by the thick black line, corresponds to the 50% decision boundary. Note that the nonuse probability map in (E) is solely for illustrative purposes; it was obtained by subtracting, for each target, the probability of successful reach with the affected arm in the spontaneous condition from the probability of successful reach with the affected arm in the constraint condition. ......................................................... 24 Figure 2.3. Arm use and nonuse in participants poststroke as estimated from AAUT QOM. The total height of each bar is cAAUT, the score in the constrained use condition. Because nuAAUT = cAAUT − sAAUT, cAAUT decomposes into sAAUT (gray), performance in the spontaneous use condition, and nuAAUT (black), estimated arm nonuse.Abbreviations: AAUT, Actual Amount of Use Test; QOM, quality of movement; c, constrained; nu, nonuse; s, spontaneous. ... 27 Figure 2.4. Examples of use and nonuse with BART in 1 session over the 2D reaching work space for 3 right-affected participants poststroke. Each row represents a different participant: a participant with little nonuse (ID10; session 2, FM 63, coordination subscale of the FM-UE [FM_cor] 5, sBART = 0.64, cBART = 0.57, nuBART = 0.113), a participant with large nonuse, albeit mild impairment (ID1; session 2, FM 57, FM_cor 4, sBART = 0.10, cBART = 0.54, nuBART = 0.433), and a participant with moderate nonuse (ID3; session 2, FM 49, FM_cor 3, sBART = 0.18, cBART = 0.340, nuBART = 0.239). Maps from left to right for each row: spontaneous use, performance, performance after masking with normative hand use data, and nonuse. As in Figure 2, nonuse maps are shown for illustrative purpose only. Abbreviations: BART, Bilateral Arm Reaching Test; c, constrained; nu, nonuse; s, spontaneous. ................................................. 30 viii Figure 2.5. External validity of nuBART shown by plotting nuAAUT as a function of nuBART for 15 participants poststroke in the validity study (correlation between nuBART and nuAAUT,r = 0.683, P = .005, Spearman).Abbreviations: BART, Bilateral Arm Reaching Test; AAUT, Actual Amount of Use Test; c, constrained; nu, nonuse; s, spontaneous. .............................................................. 34 Figure 3.1. The time-based Bilateral Arm Reaching Test (BART), target location, and protocol. A. The participant seats on the chair with trunk restraint belt. The green circle and the white circle identify the home position and target. For each trial, participants were instructed to reach to the target using either right or left index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. B. The time-based BART workspace with 35 targets and a home target surrounded by a square. Movement duration constraints for each targets are different as a function of target distance and angle. Color shows the time constraint for each target in the medium and the fast conditions. C. The time-based BART2 protocol. No time constraint, Medium, and Fast conditions and each condition has spontaneous use and constraint use sessions. A reminder session is tested before the spontaneous use session for both medium and fast conditions. .................................................................... 48 Figure 3.2. Use and Performance as a function of condition in participants poststroke. A. Performance measurement in the constraint use session for each condition. Each marker represents each individual with stroke. A number of targets successfully reached by the more-affected arm decreases in the fast condition compared to those in the no time constraint and the medium condition (p<0.001). B. Use measurement in the spontaneous use session for each condition. A number of targets successfully reached by the more-affected arm is significantly greater than 0 in the no time constraint (p<0.001). The fast condition shows lower use of the more-affected arm than the no time constraint and the medium condition (p<0.001). ............................................. 55 Figure 3.3. Examples of performance, use, and nonuse in the fast condition over the 2D reaching work space for 2 right-affected participants poststroke and averaged data for age-matched nondisabled participants. Top row represents averaged data for age-matched nondisabled participants, while middle and bottom row represent a participant with nonuse (ID11; session 2, FM 53) and a participant with no nonuse, albeit mild impairment (ID 5; session 2, FM 52). A. Use of the more-affected arm in participants poststroke (middle and bottom) and use of the right arm for right-handed nondisabled participants in the spontaneous use session (normative hand use). Black circles represent the targets that participants successfully reached using their more-affected arm (or right arm for nondisabled). B. Performance measured in the constrained use session. C. Performance after masking with normative hand use of nondisabled participants (panel A, top). D. Nonuse. Black circles represent non-use of the more-affected arm. ................................................................... 57 ix Figure 3.4. Validation of BART2 system. The correlation between sBART fast and sAAUT QOM is strong (A), the correlation between cBART fast and WMFT time – all items is strong (B). The correlation between cBART fast and WMFT time – Hand related items (C) is stronger than the correlation between cBART fast and WMFT time – Arm related items (D). .................................. 59 Figure 3.5. Test-Retest Reliability for cBART fast (A), sBART fast (B), and nuBART fast (C). X-axis is tests conducted every 2 weeks. Y-axis is a number of targets successfully reached by the more- affected arm. Both cBART fast and sBART fast show excellent test-retested reliability and nuBART fast shows acceptable test-retest reliability. ..................................... 61 Figure 4.1. The time-based Bilateral Arm Reaching Test (BART), target location, and protocol. A. The participant seats on the chair with trunk restraint belt. The green circle and the white circle identify the home position and target. For each trial, participants were instructed to reach to the target using either right or left index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. B. The time-based BART workspace with 35 targets and a home target surrounded by a square. Movement duration constraints for each targets are different as a function of target distance and angle. Color shows the time constraint for each target in fast condition. C. The time-based BART2 protocol. No time constraint, Medium, and Fast conditions and each condition has spontaneous use (SU) and constraint use (CU) sessions. A reminder session is before SU for both medium and fast conditions. .............. 77 Figure 4.2. The time-based Bilateral Arm Reaching Test (BART), target location, and protocol. A. The participant seats on the chair with trunk restraint belt. The green circle and the white circle identify the home position and target. For each trial, participants were instructed to reach to the target using either right or left index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. B. The time-based BART workspace with 35 targets and a home target surrounded by a square. Movement duration constraints for each targets are different as a function of target distance and angle. Color shows the time constraint for each target in fast condition. C. The time-based BART2 protocol. No time constraint, Medium, and Fast conditions and each condition has spontaneous use (SU) and constraint use (CU) sessions. A reminder session is before SU for both medium and fast conditions. .............. 87 x Figure 4.3. Choice, Movement Duration, Effort, and Success. Each column shows averaged choice, MD, effort and success for both the right and the left arm in the RH (A), the LH (B), and the Control (C) groups. Each box shows choice, MD, effort, and success change for the right and the left arms as a function of condition. ..................................................................................................... 90 Figure 4.4. The z-transformed differences between arms in movement duration, effort, and success rate for each target in the fast condition. Each target is representing the averaged z-transformed value of difference between arms for all participants in each group for Effort (A), MD(C), and Success rate (D). The more-affected arm choice (for the RH and the LH groups) and the right arm choice (for the Control group) are shown in (A). ..................................... 94 Figure 4.5. Comparison of the fixed-effect coefficients for best-fit models for the Stroke and the Control groups. A. The computational model for the Stroke group shows that effort and movement duration determine the more-affected arm choice for both the RH and the LH groups. Success additionally determines the more-affected arm choice for the LH group. The fixed-effect coefficients, that are statistically significantly different from zero, are indicated by * (p<0.001). The fixed-effect coefficient for success in the LH group is significantly greater than in the RH group, as noted by ** (p<0.01). B. The computational model for the Control group shows that effort and movement duration determine the right arm choice. The fixed-effect coefficients for effort and movement duration are significantly different from zero, are indicated by * (p<0.001). Note that sings for effort and movement duration are negative whereas sign for success is positive. ............................. 97 Figure 4.6. Actual choice and model predicted choice. Actual choice is measured from the experiment and model predicted choice is estimated from best-fit model. A. Actual choice in the RH (dark gray) and the LH (white) groups. B. Best-fit model estimates the more-affected arm choice well. C. Best-fit model without success estimates the more-affected arm choice less accurately for the LH group compared to best-fit model, especially in the no time and the medium conditions. D. Actual choice in the Control group. E. Best-fit model estimates the right arm choice well. E. Success added to best-fit model does not change model predicted choice compared to the best-fit model. ..... 100 Figure 4.7. Habitual and adaptive use of affected arm in the RH and the LH groups. Dark gray is the RH group and white is the LH group. The LH group shows higher switch rate than the RH group, while they show similar stay rate. ..... 101 xi LIST OF TABLES Table 2.1. Demographic data of the stroke group .................................................... 25 Table 3.1. Demographic information of participants. Abbreviations: SE, standard error; UEFM, Fugl-Meyer; UEFM-cor, coordination subscale of the UEFM– Upper Extremity ............................................................................................... 53 Table 4.1. Summary of participant information ......................................................... 102 Table 4.2. Choice, Movement Duration, Effort, and Success for the right and the left arms across conditions in the RH, the LH, and the Control groups. *right vs left; †paretic arm in stroke group vs intact arm in control; § right paretic vs left paretic; ..................................................................................................... 103 Table 4.3. Mixed effect logistic regression model fits for Stroke group ................... 104 Table 4.4. Mixed effect logistic regression model fits for the Control group. Note: *p<0.05, **p<0.01 ***p<0.001 ..................................................................... 105 xii ABSTRACT Most individuals with upper extremity disability resulting from a stroke face difficulties effectively using their more-affected arm and hand in daily activities. Spontaneous use of the more- affected arm in daily life is a meaningful indicator of motor recovery for stroke rehabilitation. However, the spontaneous use of the more-affected arm varies among patients. Some patients with stroke continue to use the more-affected arm after treatment, while others avoid using their more-affected arm in the real world even though they have a residual capacity to use it. Therefore, it is important to understand the mechanisms underlying arm choice in individuals with post-stroke as a basis for effective rehabilitation strategies. This dissertation has two aims. The first is to develop a simple, objective, and replicable tool to assess movements in upper extremities in individuals with stroke. The Bilateral Arm Reaching Test (BART) system has been proposed to assess spontaneous use, non-use, and performance of upper extremities. The BART shows excellent test-retest reliability and strong external validity with clinical assessments for non-use of the more- affected arm in individuals with stroke. In addition, we modified the BART system (time- based BART) by adding additional movement duration constraints and evaluated performance and use of the more-affected arm. The fast condition in the time-based BART system shows excellent test-retest reliability and strong validity with Actual Amount of Use Test (AAUT) and Wolf Motor Function Test (WMFT) for spontaneous use and performance assessments. xiii The second aim of this dissertation is to understand the mechanisms that determine arm choice for both patients with stroke and age-matched nondisabled people by clarifying individual factors related to arm choice during reaching tasks. Lesion side (Right hemiparesis vs. Left hemiparesis) leads to different arm choice patterns and mechanisms. Computational modeling of arm choice revealed that patients with right- hemiparesis took effort and movement duration into account when choosing the more-affected arm, while patients with left-hemiparesis additionally took success into account. The results of this dissertation may ultimately further the development of rehabilitation strategies customized for each individual with stroke. 1 CHAPTER 1: Introduction 1.1. Statement of the problem Individuals post-stroke have difficulties in effectively using their affected arms and hands for normal daily activities after stroke. Recent research has shown that therapies delivered in the sub-acute to chronic phase can increase spontaneous use and function of the affected limb (Michaelsen and Levin, 2004; Taub, Uswatte and Elbert, 2002; Winstein et al., 2004; Wolf, McJunkin, et al., 2006). However, long-term change in arm and hand use in the months following therapy is variable among patients. In some patients, use continues to improve in the months following therapy, while for other patients, improvement in use is short-lasting (Hidaka et al., 2012; Schweighofer et al., 2009; Wolf et al., 2010). Furthermore, some patients do not use the affected limb after being discharged from the hospital or after therapy sessions, although they have relatively high function and show increases in use during treatment sessions. This poor transfer of gains made in the clinical setting to daily life is frequently observed (G. Uswatte and Taub, 2005) and is called ‘non-use’, in reference to the discrepancy between motor ability and the actual use of an extremity in real-world situations (Andrews and Stewart, 1979; Sterr, Freivogel and Schmalohr, 2002). Little is known however about the reasons that patients with stroke avoid using their affected limb and how we as clinicians can lead them to continuously use their affected limb even after therapy has ceased. 2 To understand the underlying mechanisms of recovery and to assess the effects of therapy, the appropriate measurement tool is needed. Most studies used clinical tests, including the Fugl-Meyer (FM) for measuring motor impairment; the Wolf Motor Function Test (WMFT), the French Arm Test (FAT), the Action Research Arm Test (ARAT), the Box and Block (BB) and the Nine Hole Peg Test (NHP) for measuring function; the Motor Activity Log (MAL), the Actual Amount of Use Test (AAUT) for measuring participation (Fugl-Meyer et al., 1975; Mathiowetz et al., 1985; G. Uswatte, Taub, et al., 2006). These clinical tests provide general information regarding the functional status of the affected limb, such as how fast patients with stroke move, how many times, and how well they use their affected limb. However, some of these clinical measurements are subjective because they rely on either participants’ self-evaluation or testers’ evaluation, or they are less time efficient because they require a relatively longer administration time and a formalized training program to certify the evaluators. Furthermore, they do not specify how motor control changes over time post-stroke; nor do they provide conclusive evidence as to whether improvements shown after therapy are “true recovery” (i.e. “restitution”) or rather “compensation” (i.e. “substitution”) with alternative movement strategies. Therefore, more objective outcome measures are necessary in order to understand the mechanisms underlying stroke recovery and therapies. The aims of the current dissertation are 1) to develop simple, objective, and replicable tools to measure spontaneous use, non-use, and function of the affected limb, and 2) to understand the mechanisms that determine arm choice among patients with stroke by clarifying individual factors related to hand selection during reaching tasks. Understanding arm choice mechanisms and using assessments that are more precise may 3 ultimately provide insight when we establish rehabilitation strategies customized for each individual with stroke. The next section of this general introduction explores the factors that possibly influence arm choice in patients with stroke and in nondisabled people. Subsequently, current tools to assess use, non-use, performance of upper extremity for stroke rehabilitation are reviewed. This general introduction ends with the outline of the current dissertation. 1.2 Factors influencing arm choice in post-stroke and nondisabled people Arm choice is influence by several factors such as lesion characteristics, pain, sensory motor impairment, rewarding or punishing experience of using the affected arm after stroke, and motivation. Several models describe factors influencing arm choice both in patients with stroke and in nondisabled people. For post-stroke, Taub proposed that patients with stroke show a decrease in use of their affected limb due to impaired basic motor control, which consequently requires more effortful movement for successful use of the affected arm. In addition, repeated attempts to use the affected arm often lead to failures and these failures constitute punishments that suppress arm use. On the contrary, using less-affected arm is easier and more successful to complete the tasks than using the affected arm, therefore, patients positively reinforce this compensational behavior (Taub et al., 1994; Taub et al., 1993; Taub, Uswatte and Elbert, 2002). While Taub and his colleague argued that patients with stroke “learn" not to use the affected limb as a result of previous unsuccessful experience, Sunderland & Tuke (2005) insisted that decreased spontaneous 4 use is a results of the increases in effort and attention, and the decrease in willingness to the use affected arm. Specifically, using the affected arm requires more effort and attention than using the unaffected arm, patients with stroke might thus not be willing to use the affected arm, not just that they “learn” non-use of the affected arm (Sunderland and Tuke, 2005). Repetitive task-oriented interventions such as constraint induced movement therapy (CIMT) promote spontaneous arm use throughout the functional recovery of the affected arm. One recent study proved the existence of the ‘functional threshold’ for spontaneous use and suggested that the trainings leading patients to exceed this functional threshold are effective for spontaneous use. If function of the affected arm after therapy is above this threshold, patients with stroke keep using the affected arm, otherwise they choose the less-affected arm when function is below the threshold (Schweighofer et al., 2009). Another study proposes that the expected success performed with the affected arm influences arm choice in post-stroke. Using simulation, Han et al. (2008) revealed that one arm is chosen based on a comparison of the ‘values’ (expected success) of using the affected arm or the less-affected arm (Han, Arbib and Schweighofer, 2008). In addition to the research that focuses on the relationship between arm choice and physical factors related to body impairment and functionality, some studies take into account the effects of psychometric properties including perception, expectation, perception, and emotion on arm choice. Self-efficacy, which reflects how patients with stroke perceive their capability to use the affected arm, has been proposed as another possible factors in arm choice. It has recently been shown that patients with high self- efficacy in using the affected arm show a higher frequency of selecting the affected arm 5 whereas patients with low self-efficacy do not (Chen et al., 2013). The patients’ self- confidence level can be changed throughout positive verbal feedback after task performance, or improvement on function which allows patients to have successful experience; thus leading to increased confidence in using the affected arm (Hidaka et al., 2012). Several studies with nondisabled participants have also been conducted to further understand hand choice mechanisms. In healthy people, handedness greatly affects hand choice, but we also see impressive flexibility in arm choice in our everyday behavior (Schenker et al., 2006). For instance, the location of an object to be grasped, the type of tasks, and comfort levels of dominant and non-dominant hand change arm choice patterns. Expected reward (note that expected reward is equivalent to “success” of the task in goal-directed upper extremity movement) also influences arm choice. Stoloff et al. (2011) clearly showed that nondisabled participants used their non-dominant arm more in situations in which a higher reward was assigned to the non-dominant hand and a decreased reward was assigned to the dominant arm for the target reaching task (Stoloff et al., 2011). In line with this study, Schweighofer et al. found that biomechanical effort accompanied by reward also plays a key role in arm choice (Schweighofer et al., 2015). While the effects of those factors such as handedness, target size and location, comfort level, expected reward, biomechanical effort, and neural substrates for decision making on the arm choice are continuously investigated in healthy people, arm choice mechanisms in patients with stroke are still unclear. Therefore, identifying the factors influencing arm choice seems a necessary first step toward understanding the general principle underlying stroke recovery. 6 1.3. Assessment tools for use, non-use, and performance of upper extremity Assessments of rehabilitation outcomes require measuring the extent to which the affected arm is spontaneously used in the real world, as opposed to simply evaluating motor ability elicited by, for example, a performance test in the laboratory. This is because there are frequently large disparities between motor ability and the motor behaviors actually performed in everyday life (Lang et al., 2013; Taub et al., 1994; G. Uswatte and Taub, 2005). As motor ability measured in the laboratory might not correspond with what patients actually do outside of this environment, several assessments capturing spontaneous use of the affected arm in the real world have been developed. There are three widely used measurements in both clinical practice and research trials for spontaneous use of the affected arm in individuals with post-stroke: the Motor Activity Log (MAL), the Actual Amount of Use Test (AAUT), and the accelerometer. Use assessments The MAL consists of a semi-structured interview that assesses a patient’s use of their paretic arm and hand during activities of daily living (ADL) (Taub et al., 1993; G. Uswatte, Taub, et al., 2006). The patients with stroke are asked to evaluate ‘how frequently’ (Amount of use; AOU) and how well (Quality of use; QOU) the affected arm participated in thirty common and important ADLs including brushing teeth, buttoning a shirt or blouse, and eating with a fork or spoon during the past several days. For the AOU score, patients are asked to indicate the frequency with which they use their affected arm for the activity, 7 with possible scores ranging from 0 (never use the affected arm for this activity) to 5 (always use the affected arm for this activity). For the QOM score, patients are asked to indicate how well they use the affected arm to perform the activity, with possible scores ranging from 0 (inability to use the affected arm for this activity) to 5 (ability to use the affected arm for this activity is the same as before the stroke) (Taub, Uswatte and Pidikiti, 1999; G. Uswatte, Taub, et al., 2006; van der Lee et al., 2004). However, the MAL relies on subjective participants’ ratings and recall, as a result, the scores can be inaccurate if participants have difficulty remembering the past, specifically in relation to the somewhat automatic actions on the survey (e.g., turning on a light switch). Furthermore, the general positive attitude of patients at the end of a treatment may bias the ratings of amount and quality in order to please the tester or themselves. Thus, use of the MAL as a primary outcome measure in clinical trials is not recommended (van der Lee et al., 2004). The AAUT is a covert performance-based assessment with 17 daily activities, such as pulling out a chair, placing a photo in an album, and filling out a form. Participants are covertly videotaped while performing a standardized scenario of 17 tasks. Trained, masked evaluators watch the videotape to assess the participants’ spontaneous use of their affected arm, both in terms of how much (on a 2-point scale of amount of use; AOU) and how well (on a 5-point scale of quality of use; QOM) the affected arm is used (Taub et al., 2006). However, the AAUT cannot be repeated or completed if participants with stroke realize what the AAUT test is or if they are in this testing, because participants are biased, resulting in unnatural hand choice patterns. In addition, low resolution of scoring on the AOU scale is insensitive in detecting the smallest amount of changes without inherent variation or noise in the measure itself (minimal detectable change; MDC) in arm choice before and 8 after therapy (Chen et al., 2012; Han et al., 2013). An accelerometer provides an accurate, reliable, and stable measure of the duration of arm movement, a parameter that is a valid index of limb use in the real world (G. Uswatte et al., 2000). The unit of measurement is an activity count and is relatively sensitive to movements performed at frequencies between 0.1 Hz and 3.6 Hz. Uswatte and colleagues used the accelerometer to measure the effect of CIMT in stroke patients. They calculated the ratio of the accelerometer recordings of the more-affected arm to the less-affected arm recording (G. Uswatte, Foo, et al., 2005). However, the ratio of accelerometer recordings that they used omitted information about separated usage times of the affected and the unaffected limbs. Furthermore, purposeful and non-purposeful use (e.g. arm use for opening door vs. arm swing during walking) was not dissociated well, thus new data processing techniques for accelerometer data are required (Michielsen et al., 2012). To our knowledge, the MAL, the AAUT, and the accelerometer are the only instruments in clinical trials that capture changes in spontaneous arm use in patients with stroke. As discussed above, these assessments have several limitations; thus, there is considerable need to develop valid and reliable tools to capture spontaneous arm use. Non-use assessments Non-use refers to the discrepancy between what patients with stroke can do and what they actually do. Non-use was discovered by Andrew and Steward (1979) when they observed patients with stroke both at clinic and at home. They found that patients with 9 stroke used their affected arm less for the activities of daily livings in the home situation than in the clinic and suggested that the gain during rehabilitation in clinic is not fully transferred to the real-world (Andrew and Stewart, 1979). So far, there is no direct way to measure objectively the non-use shown in post- stroke. A study conducted by Sterr et al. (2002) indirectly measured the non-use of the affected arm in patients with brain damage (Sterr, Freivogel and Schmalohr, 2002). They measured AAUT twice; the first measurement was for a free condition in which participants chose either arm to perform the tasks, while the second was for a forced condition in which participants used only the affected arm. Non-use was estimated by the difference between residual movement capability measured in the forced condition and the spontaneous use measured in the free condition. Although the existence of non-use in upper extremity was proved, the participant population recruited was not patients with stroke and the AAUT was not capable of repetition. Thus, new tools to assess non-use in post-stroke is needed. 10 Performance assessment To measure performance of the affected limb, the Wolf Motor Function Test (WMFT), the Arm Motor Ability Test (AMAT), the Jebsen Hand Function Test (JHFT), and the Box and Block Test (BB) are commonly used. The WMFT quantifies the ability to use a UE via timed single- or multiple-joint motions and functional tasks (Wolf et al., 1989). The test consists of 17 items, 2 of which involve strength measures (lifting and handgrip) and 15 of which involve timed performance on various tasks. Performance time (up to a maximum of 120 s), quality of motor function (on a 6-point Functional Ability Scale, FAS), and muscle strength are assessed. The WMFT has been established as a reliable and valid tool for assessing stroke patients with mild to moderate UE impairment (Morris et al., 2001; Wolf et al., 2001). However, the WMFT requires relatively long administration time (e.g., around 30~40 minutes), additional time for scoring FAS, and a certification process to train evaluators, wondering the WMFT less efficient in measuring time and effort (Lang et al., 2013). The AMAT measures the ability of patients to perform 13 instrumental activities of daily living with their affected arm (Kopp et al., 1997). All test items are practical, multiple-step ADL tasks such as cutting with a knife and fork, opening a jar, tying a shoelace, and putting a t-shirt. Performance time and functional ability are measured for each item (Kopp et al., 1997). The Jebsen Hand Function Test is specialized for measuring uni-manual hand functions required for activities of daily living (ADLs). It includes 7 sub-tests such as writing, turning a cart, and stacking checkers and instructs participants to perform the tasks 11 uni-manually. The less-affected hand is assessed first, followed by the more- affected hand, and the total time to complete task is measured (Spaulding et al., 1988). The Box and Block test also measures hand and arm ability by instructing participants to move as many blocks as possible, one at a time, from one compartment to another for a period of 60 seconds. Many of current clinical tests for performance are time-based. Unfortunately, the measured time is affected by the items uncompleted in the given time limit (e.g. 120 sec for WMFT) or by the measurement error occurred when evaluators press the button on stopwatch; thus, time score is less accurate. In addition, the ordinal scale used for quality of movement or functional ability in WMFT, AMAT, and the Jebsen hand function test is subjective and is not sufficiently sensitive to capture further improvement for patients with mild impairments (van Kordelaar et al., 2012). Furthermore, if patients with stroke have no residual capacity to open and close their hands, BB or the Jebsen hand function test requiring dexterous hand manipulation cannot not be tested. Therefore, quick and objective assessments to measure performance of the affected arm in patients with stroke are required. 1.4. Overview of dissertation structure The aim of the current dissertation is to understand arm choice in individuals with post-stroke. For this purpose, I developed a tool that allows us to assess use, non-use, and performance of upper extremity objectively and repeatedly. In chapter 2, I explain how non-use of the affected arm in post-stroke is measured in the laboratory setting. In chapter 3, a simple, quick, and objective assessment for use and performance of the affected arm 12 is introduced. In chapter 4, factors influencing arm choice in post-stroke are addressed and the different affected arm choice patterns between RHD and LHD and the factor dissociating RHD and LHD are discussed. In chapter 5, the effect of newly designed rehabilitation intervention, called ASAP, on arm choice is reported. Chapter 6 provides a summary of the dissertation, clinical implication, and limitations with suggestions for future direction. 13 Chapter 2: Quantifying Arm Non-use in Individuals Post-stroke 2.1. Introduction Most individuals with upper extremity disability resulting from a stroke face difficulties to effectively use their paretic arm and hand in daily activities, resulting in significantly reduced quality of life.(Dobkin, 2005; Duncan et al., 2003; Mayo et al., 2002) Such “non-use” has been defined as the difference between what the individual can do when constrained to use the paretic arm and what the individual does when given a free choice to use either arm.(Andrew and Stewart, 1979) Non-use in individuals with hemiparetic stroke (or with other predominantly unilateral motor neurological disorders) can arise from a number of factors such as pain, limited range of motion, as well as higher effort and attention required for successful use of the impaired hand.(Sunderland and Tuke, 2005) Non-use has furthermore been hypothesized to have a learned component.(Sterr et al., 2002; Taub et al., 1994; Taub and Uswatte, 2003, 2006) According to this “learned non-use” hypothesis, non-use would develop either after unsuccessful repeated attempts to use the affected arm and hand, or after negative consequences resulting from paretic limb use (such as spilling hot coffee or dropping a retrieved object). 14 Despite the high clinical and scientific significance of non-use, little work has been conducted to directly quantify non-use in individuals post-stroke. A seminal study by Sterr et al.(Sterr, Freivogel and Schmalohr, 2002) estimated non-use in a relatively heterogeneous group of brain injured adolescents with the Quality of Movement subscale of the Motor Activity Log test (MAL QOM)(G. Uswatte, Taub, et al., 2006) and the Amount of Movement subscale of the Actual Amount of Use Test (AAUT AOU)(Sterr, Freivogel and Schmalohr, 2002; G. Uswatte, Taub, et al., 2005). Non-use was estimated for each test by the difference between the actual test score and the score obtained when the participants actually perform the tests with the affected hand. Although the Sterr et al. study demonstrated the feasibility of measuring non-use in hemi-paretic individuals with the MAL QOM and the AAUT AOU, neither instrument fulfills the five criteria for an ideal measurement tool in neurological rehabilitation (Wade, 1992): simplicity, objectiveness, test-retest reliability, external validity, and sensitivity. First, neither MAL nor AAUT are simple to administer, and both require adequate training of the tester by experienced therapists. Second, the MAL relies on subjective participants’ ratings of the amount of use and quality of movement of their more affected arm in functional daily activities outside the laboratory. Third, to preserve validity, the modified AAUT cannot be administered repeatedly, and as such, it may lack good test-retest reliability. Fourth, the AAUT AOU has large variability and is relatively insensitive to treatment effect because of the low resolution of scoring.(Chen et al., 2012) In light of these limitations, and in order to advance both theoretical and practical knowledge about the recovery of arm use after stroke, there is therefore a considerable need to develop tools that capture purposeful arm use and non-use with objective activity monitoring.(Chen et al., 2012) 15 The goal of the present study is threefold. First, we investigate the existence of non- use in participants with chronic stroke, as measured from the AAUT QOM. Second, we propose a novel, simple, and quick to administer laboratory-based measurement tool, the Bilateral Arm Reaching Test (BART) to quantify paretic arm use and non-use objectively and repeatedly. Third, we investigate whether measurement of arm non-use with BART in participants with chronic stroke is reliable in test-retest, and whether it exhibits external validity when compared to non-use as estimated from the AAUT QOM. 2.2. Method 2.2.1. Participants Twenty-four participants with chronic stroke (18 males, 6 females) were enrolled in this study. The inclusion criteria were: (1) Mini-Mental State Examination score > 25/30; (2) at least 6 months post-stroke; (3) no pain in the paretic arm and hand; (4) right-hand dominant pre-stroke; (5) the ability to reach the farthest target displayed at 30 cm anterior to the midline trunk (the target is the farthest target in front of the body midline presented by BART). For the validity study, an additional criterion was (6) no visual neglect -- all of the lines in both left and right workspaces were crossed out in Albert’s test(Fullerton, McSherry and Stout, 1986). Pre-stroke hand dominance was self-reported. The upper extremity score of the Fugl-Meyer test (FM-UE)(Fugl-Meyer et al., 1975) was administered to all participants post-stroke by four different testers, all physical therapists with more than two year of clinical experience. In addition, ten non-disabled (4 16 males, 6 females) right-handed (according to Edinburgh Handedness Inventory)(Oldfield, 1971) age-matched participants were recruited (age of participants post-stroke: 62.25 ± 2.64 (SE: Standard error), age of non-disabled participants: 58.10 ± 3.76 (SE), t-test: p=0.390). Note that participants in this group were not recruited to serve as controls for comparison of use and non-use with post-stroke participants, but instead to define the average hand use of right-handed age-matched non-disabled participants. The average measure of use in non-disabled participants was utilized in BART as a baseline to compute non-use in participants post-stroke. Thus, by definition, participants had zero non-use if they behaved like the average non-disabled participant. Using the average behavior is reasonable because there is little between subject variability in hand use between non- disabled participants (see below and Appendix, Figure A). To test BART test-retest reliability, 19 participants with stroke performed three test sessions at least four days apart. To assess BART validity, 15 participants with stroke performed two test sessions at least four days apart. Validity was assessed with BART data from the second test session. Note that five participants were only enrolled to assess validity and as such performed only two test sessions. Ten participants participated in both reliability and validity, and as such performed three sessions. The study was approved by the Institutional Review Board of the University of Southern California, and all participants read and signed a written informed consent form prior to study enrollment. 17 2.2.2. Actual Amount of Use Test (AAUT) The AAUT was administered to the 15 participants post-stroke enrolled in the validity study first in the spontaneous use condition (sAAUT) and in the constrained use condition (cAAUT), as in Sterr et al.(Sterr et al., 2002). We analyzed only the first 14 items of the original 17 items AAUT, because these items are related to arm use while last three items are general activities such as gesturing and posture throughout testing. The AAUT test was administered by three experimenters and rated by one evaluator, a physical therapist with more than two year of clinical experience. We used the AAUT QOM instead of the AAUT AOU to test the external validity of BART for two reasons. First, the cAAUT AOU scores in Sterr et al. were close to maximum for most participants. Second, we recently showed greater variability and insensitivity to treatment effect for the AAUT AOU compared to the AAUT QOM (Chen et al., 2012). The AAUT QOM scores in the spontaneous use condition, sAAUT, and in the constrained use condition, cAAUT, were calculated and expressed as average scores, from zero to 5. Non-use was computed by nuAAUT = cAAUT – sAAUT. 2.2.3. Bilateral Arm Reaching Test (BART) Apparatus: Because target distance and location largely influence arm choice in pointing movements (Oldfield, 1971), BART displayed one of 100 targets at each trial in pseudorandom order on a 2D hemi-workspace. Targets were white disks of 2 cm in 18 diameter projected on the table from an overhead projector with a preset target presentation schedule. Targets were displayed on 6 arcs, all ranging between 10° and 170° (0 degree is on axis is parallel to the body, to the right). Arcs radii ranged from 10 to 30 cm, with equal 4 cm distance between arcs. The targets were placed every 10 degrees along the arcs. The leftmost and rightmost targets in the upper corners were not presented, making a total of 100 targets (6 arcs *17 position per arcs minus 2 targets). Two Mini-Bird model 500 (5mm) magnetic sensors (Ascension Technology Corporation) were positioned on the tip of the index finger of each hand to measure finger motion and arm choice (sampling 100 Hz). Testing procedure: Participants were seated with a seat belt to limit upper body movement (Roby-Brami et al., 2003). The position of the chair was adjusted to ensure that participants could comfortably reach to the end of the workspace without bending their trunks. The participants were instructed to place both index fingers on the home position (a green disk of 2 cm in diameter), as shown in Figure 2.1. 19 Figure 2.1. Measuring arm use with the Bilateral Arm Reaching Test: the home position is identified by the green circle and a target by the white circle. For each trial, participants were instructed to reach to the target with their choice of hand using the index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. 20 In both spontaneous and constrained use conditions, after a target was presented, participants were instructed to reach the target as quickly and as accurately as possible, to remain in contact with the target until it disappeared, and to return to the home position. After the start position was maintained for 1 sec, the home target disappeared and a target appeared at one of 100 locations. The inter-trial interval was 3 sec. A successful trial was defined when participants reached the target within 1.2 seconds. A pleasant sound was played following successful trials; an unpleasant sound was played following unsuccessful trials. Dragging the arm along the table surface was instructed to be inappropriate and was discouraged whenever observed by the experimenter. In the spontaneous use condition, participants were instructed to reach each successively displayed target with the index finger of their preferred hand. Targets appeared twice at each position in a pseudorandom order, resulting in 200 trials. Participants were reminded that there was no right or wrong answer in their arm choice of arm, but were instructed to maximize the number of successful trials. In the constrained use condition, participants post-stroke were instructed to reach each target with the index finger of the paretic arm. Targets appeared once at each position in a pseudorandom order, resulting in 100 trials. Note that non-disabled participants only performed the spontaneous condition, as we noticed in our beta version that all non-disabled individuals could reach all targets in the hemi-workspace within 1.2 sec (and thus obtain a maximum score in the constrained condition). 21 2.2.4. BART Dependent Measures Overview: By definition, non-use is computed by subtracting what post-stroke participants do from what they can do. In BART, direct subtraction of successful movements made by the paretic in the spontaneous condition from those made in the constrained condition would not work, however, for the following reason. Non-disabled participants can reach targets in both left and right workspaces with their dominant right hand within 1.2 sec with no failures. However, when given a choice, they use their right hand for the right workspace and their left hand for the left workspace. Computation of non-use by simple subtraction of the “spontaneous reach area” from the “constrained area” would yield large non-use for non-disabled participants, which is not appropriate. We therefore assume that the average non-disabled participant has zero non-use. To compute non-use, we collected non-disabled participant choice behavior in the spontaneous BART condition, and use the average behavior as a mask to the performance data of the participant post-stroke. In this way, when the use measure of the paretic arm in the spontaneous condition is equal to the “masked” performance, there is zero non-use. Any use in the spontaneous condition that is less than masked performance signifies non-use. Spontaneous use (sBART): To compute arm use in participants post-stroke, we quantified the successful trials made by the affected arm (for age-matched controls, we quantified the successful trials made by the dominant arm). For each target, we computed the probability of successful reach with the paretic arm ) rm| target affectedA handChoice (success, P free . A second order logistic regression model interpolated and smoothed these probabilities to obtain a probabilistic use map over 22 the 2D workspace. The logistic regression model has an extended input feature space, [x 2 ,y 2 ,xy,x,y,1], where x and y are the coordinates of 2D workspace. The choice was established based on our data showing that the indifference line (i.e., line of 50% choice probability) can be very curved in participants post-stroke (e.g., Figure 2d); a simple first order logistic regression model cannot account for such curvature. sBART was then computed by integrating the volume beneath the probability surface given by the logistic regression model of use, using 1000 by 1000 grid over the workspace. Performance after masking (cBART): As discussed above, because the workspace covered in the constrained use condition can be equal to the whole workspace in participants post-stroke with mild impairments, we cannot directly subtract sBART from the integrated performance over the whole hemi-workspace in the constrained use condition. We therefore masked the successful reach area in the constrained use condition with the average arm use of the non-disabled participants from the spontaneous use condition. Specifically, we first defined the probability of reaching each target successfully with the affected arm (see Figure 2.2 for steps of the computation): _ (, | ) perf const free healthy PsuccesshandChoice affectedArmtarget P(success| target) P (handChoice affectedArm| target) (1) where target) (success| P const was computed from a second order logistic regression model (as above) with the successful trial data in the constrained use condition (Figure 2a), and _ free healthy P (handChoice affectedArm| target) was computed from a logistic regression model with the successful trial data from the spontaneous condition for non-disabled controls (see Figure 2b). ) | , ( target m affectedAr handChoice success P perf was thus a 23 probability of successful reach in the constrained condition, masked by the spontaneous choice data in non-disabled age-matched controls (Figure 2c). cBART was then computed by integrating the volume beneath the probability surface given by: ) | , ( target m affectedAr handChoice success P perf . Non-Use (nuBART): Finally, nuBART was then computed by subtracting sBART from cBART, and taking possible negative values to 0, in line with the definition of non- use and similarly to non-use as estimated with the AAUT. Thus, sBART cBART nuBART (2) where the function x returns x if x>0 and 0 otherwise. 2.2.5. Statistical Analyses The number of trials per session in the spontaneous use condition (200) was determined from a sensitivity analysis of the non-disabled participants’ choice data: We computed sBART for increasing number of trials from 50 to 400, and observed that sBART converged at between 150 trials and 200 trials for all participants (data not shown).For each variable, sBART, cBART, and nuBART, we performed test-retest reliability, and external validity computations. Reliability was assessed with intraclass correlation coefficients (ICCs) for three sessions and with Pearson or Spearman correlations for session-to-session measurements. External validity was tested with Pearson or Spearman correlations. External validity for sBART, cBART, and nuAAUT were tested by correlations with 24 sAAUT, cAAUT, and nuAAUT respectively. In addition, we tested for possible correlations between nuBART and months since stroke and age, and tested the effect of the side of paresis. Figure 2. 2. Computing nonuse with BART in 1 session over the 2D reaching work space for right-affected poststroke participant, ID2, in session 3: A. Constrained use (performance) probability for ID2. B. Average spontaneous use probability for right- handed nondisabled participants (normative hand use). C. Constrained use (performance) probability after masking with normative hand use of panel (B) poststroke. D. Spontaneous use probability. E. Nonuse probability for ID2. Color coding: red = 100% use of the paretic arm (right arm for healthy controls), blue = 0% use of the paretic arm. The indifference line, indicated by the thick black line, corresponds to the 50% decision boundary. Note that the nonuse probability map in (E) is solely for illustrative purposes; it was obtained by subtracting, for each target, the probability of successful reach with the affected arm in the spontaneous condition from the probability of successful reach with the affected arm in the constraint condition. 25 For all analyses, the level of statistical significance was set at P < 0.05. When the data were not normal (as tested with Shapiro-Wilk test), non-parametric statistics were used (as indicated in results). Data analyses were performed using the SPSS 13.0 and MATLAB 7.5. All average results are reported as average ± standard errors (SE). 2.3. Results 2.3.1. Demographic and stroke characteristic data Table 2.1 summarizes the demographic data for the 24 participants post-stroke. There was no difference in age between groups (62.25 ± 2.64 post-stroke, 58.10 ± 3.76 controls P = .390). In the stroke group, there was no difference between affected sides (P = .186). Time from stroke onset was 79.46 ± 12.19 months (range 11 to 275 months). The FM-UE was 49.21 ± 2.18 (range 22 to 63). The 15 participants in the validity study had a score of 0 on Albert’s test, indicating no neglect. Table 2. 1. Demographic data of the stroke group Mean SE Minimum Maximum Age(yrs) 62.25 2.64 35 83 Time from Onset (month) 79.46 12.19 11 275 FM-UE Motor Score (66 max) 49.21 2.18 22 63 FM-Cor (6 max) 4.21 0.24 2 6 Gender 18 Male, 6 Female Affected Hand 13 Right , 11 Left Reliability/Validity 19 Reliability, 15 Validity * FM=Fugl-Meyer, FM-Cor=coordination subscale of the FM-UE, SE=Standard Error 26 2.3.2. Measuring use, performance, and non-use with the AAUT test For those 15 post-stroke participants enrolled for the validity study, the sAAUT QOM was 1.50 ± 0.32, and cAAUT QOM was 3.17 ± 0.28. Non-use, nuAAUT was normally distributed (P = .209), and significantly greater than zero (1.67 ± 0.18 P < .0001, one sample t-test). Thus, all participants did exhibit some degree of non-use overall, although the range of non-use observed was large (range: 0.18 - 2.64; Figure 2.3). The cAAUT correlated with FM-UE (r = 0.758, P = .001). nuAAUT did not correlate with the FM-UE, as expected (r = -0.136, P = .629) – see Appendix, Figure B. In addition, nuAAUT did not correlate with stroke duration (P > .5). There was no difference in nuAAUT between left and right hand paresis (P > .1, Mann–Whitney U test). There was a trend for a positive correlation between nuAAUT and age (r = 0.473 and P = .075, Pearson). 27 Figure 2. 3. Arm use and nonuse in participants poststroke as estimated from AAUT QOM. The total height of each bar is cAAUT, the score in the constrained use condition. Because nuAAUT = cAAUT − sAAUT, cAAUT decomposes into sAAUT (gray), performance in the spontaneous use condition, and nuAAUT (black), estimated arm nonuse.Abbreviations: AAUT, Actual Amount of Use Test; QOM, quality of movement; c, constrained; nu, nonuse; s, spontaneous. 28 2.3.3. Measuring use, performance, and non-use with BART Non-Disabled Participants: All non-disabled participants successfully reached all targets within the 1.2-second time constraint (Figure 2.2B). The mean indifference line was slightly shifted leftward of the midline. Consequently, mean sBART of non-disabled participants was greater than 0.5, with sBART = 0.60 ± 0.10, indicating a 10% handedness bias on average. As discussed above, out results highly depend on this “average non-disabled” use behavior, because non-use is computed with these average use data. We therefore verified with a bootstrap analysis (1000 samples) that the indifference line varied only for a few degrees on the right and left of the average indifference line, suggesting that arm choice showed little variability among the right-handed non-disabled age-matched participants (Appendix, Figure A). Participants Post-Stroke: Figure 2.4 shows examples of use and non-use probability maps (non-use maps are shown for illustrative purpose only, since non-use was computed by subtracting sBART from cBART, in line with the definition and with non-use as estimated with the AAUT) for three participants: a participant with little non-use (ID10); a participant with large non-use, albeit mild impairment (ID1); and a participant with moderate non-use and impairment (ID3). In participants with low sBART values, the indifference line was not often straight, and could form an island instead of a straight boundary on the 2D map (see for example, Figure 2.4, ID3, spontaneous use). 29 For those 24 post-stroke participants enrolled for the study overall, sBART was 0.28 ± 0.04, and cBART was 0.41 ± 0.03. The nuBART was significantly greater than zero (0.17 ± 0.04, P < 0.001, One Sample Wilcoxon Signed Ranks Test), although not normally distributed (P <0. 01 Shapiro-Wilk test). There were however large differences in use, performance, and non-use across participants post-stroke. Eight participants out of 24, 2 with left- and 6 with right- hemiplegia, exhibited no arm use, with sBART = 0. Three of these participants showed maximal performance cBART – these three participants thus exhibited maximal non-use with BART. The other 16 participants use their arm to some extent, as sBART was positive (0.42 ± 0.03, P < .001, Mann-Whitney). Eight of these participants, all with left hemiplegia, exhibited non-disabled-like patterns of use (sBART 0.4 for left hemiplegia, 0.6 for right hemiplegia). 30 Figure 2. 4. Examples of use and nonuse with BART in 1 session over the 2D reaching work space for 3 right-affected participants poststroke. Each row represents a different participant: a participant with little nonuse (ID10; session 2, FM 63, coordination subscale of the FM-UE [FM_cor] 5, sBART = 0.64, cBART = 0.57, nuBART = 0.113), a participant with large nonuse, albeit mild impairment (ID1; session 2, FM 57, FM_cor 4, sBART = 0.10, cBART = 0.54, nuBART = 0.433), and a participant with moderate nonuse (ID3; session 2, FM 49, FM_cor 3, sBART = 0.18, cBART = 0.340, nuBART = 0.239). Maps from left to right for each row: spontaneous use, performance, performance after masking with normative hand use data, and nonuse. As in Figure 2, nonuse maps are shown for illustrative purpose only. Abbreviations: BART, Bilateral Arm Reaching Test; c, constrained; nu, nonuse; s, spontaneous. 31 Thirteen participants out of 24, 6 with left- and 7 with right-hemiplegia, had non- disabled-like performance, as measured by cBART. Among the other eleven participants, cBART was 0.26 ± 0.03 for the 5 left hemiplegia participants and 0.40 ± 0.08 for the 6 right hemiplegia participants (note that because of the masking procedure, average maximum cBART for left hemiplegia was 0.4 and for right hemiplegia was 0.6, see above). We verified that these eleven participants with lower performance had a lower coordination subscale of the FM-UE (FM_cor) (median FM_cor of 3, range 2-5) than the 13 participants with non-disabled-like performance (median FM_cor of 5, range 4-6, P < .05, Mann- Whitney). Eleven participants out of 24, 9 with left-, and 2 with right-hemiplegia showed no non-use, with nuBART = 0. The thirteen participants with positive ‘non-use’ had a median nuBART = 0.235 (different from 0, P = .0001, Mann-Whitney). Most participants with right-hemiplegia exhibited at least some degree of non-use (2 out of 13), in contrast to participants with left-hemiplegia for whom only 2 out of 11 exhibited greater than zero non-use. Finally, there were no significant correlations between nuBART with months since stroke (P > .4), and age (P > .5). 2.3.4. BART Test-retest Reliability For the 19 participants post-stroke who participated in the reliability study, sBART was 0.25 ± 0.04 for the first session, 0.28 ± 0.05 for the second session and 0.29 ± 0.04 for the third session. sBART, cBART, and nuBART had good test-retest reliability across the 32 three sessions: sBART intraclass correlation coefficient (ICC) was 0.840, (P < .0001), cBART ICC was 0.807, (P < .0001), and nuBART ICC was 0.786, (P < .0001; note however that these ICC results should be considered with caution because sBART and nuBART were not distributed normally). The correlations of sBART, cBART, and nuBART between sessions 1 and 2 were good (sBART: r = 0.668, P = .002, Spearman correlation; cBART: r = 0.803, P < .0001, Pearson correlation; nuBART: r = 0.711, P = .001, Spearman correlation), and increased between session 2 and 3 (sBART: r = 0.854, P < 0.0001, Spearman correlation; cBART: r = 0.883, P < .0001, Pearson correlation; nuBART: r = 0.950, P < .0001, Spearman correlation). This demonstrates excellent reliability of BART between session 2 and 3, but lower reliability between session 1 and 2. As a result, we considered the first session as a familiarization session. In the validity study, we therefore only analyzed the data from session 2. Unlike non-disabled participants who could always reach to all the targets within 1.2 sec, reach success in the sBART condition was lower for participants with stroke, although success rates improved somewhat over the sessions (first session success rate = 87.26 ± 2.64%, second session 91.97 ± 1.96% and third session, 91.02 ± 2.19%. repeated ANOVA, P < .037). Success rates were greater in session 2 than in session 1 (paired t-test P = .012), but not between session 2 and 3 (paired t-test P > 0.5). 33 2.3.5. External Validity of BART use sBART, performance cBART, and non- use nuBART Here, we validate sBART, cBART, and nuBART with equivalent AAUT QOM measures for 15 participants who participated in the validity study, using BART measurements from the second session. Correlation between sBART and sAAUT QOM was good (r = 0.679, P = .005, Spearman). Correlation between cBART and cAAUT QOM just reached significance (cAAUT QOM r = 0.515, P = .05, Spearman). Finally, correlation between nuBART and nuAAUT was good (r = 0.683, P = .005, Spearman). Figure 2.5 shows nuAAUT as a function of nuBART. Note that nuBART was zero for six participants and very close to zero (0.008) for another participant. 34 Figure 2.5. External validity of nuBART shown by plotting nuAAUT as a function of nuBART for 15 participants poststroke in the validity study (correlation between nuBART and nuAAUT,r = 0.683, P = .005, Spearman).Abbreviations: BART, Bilateral Arm Reaching Test; AAUT, Actual Amount of Use Test; c, constrained; nu, nonuse; s, spontaneous. 35 2.4. Discussion The present study makes three important and novel contributions. First, we have, to our knowledge, for the first time directly quantified paretic arm non-use in individuals with chronic stroke. For this purpose, we subtracted average use obtained in the spontaneous use condition of the AAUT QOM from average performance obtained in the constrained use condition of the AAUT QOM. Second, we have developed a novel laboratory-based measurement tool, the Bilateral Arm Reaching Test (BART), to measure arm non-use objectively. BART is simple to administrate, requires minimal instructions to the participant and personnel training, and is objective. Although we have developed and studied BART especially for individuals post-stroke, we believe that BART can reliably quantify arm use and performance in participants with other (lateralized) neurological conditions such as Parkinson’s disease, Cerebral Palsy, focal dystonia, or even non- neurological conditions such as scapular pain. Third, we showed that non-use measured with BART has excellent test-retest reliability and good external validity with non-use measured with the AAUT QOM. We found no correlation between either measure of non-use, nuAAUT and nuBART, with time since stroke and age. However, we found with BART that non-use depends on the arm affected. Most participants with affected right (dominant) arm exhibited non-use; in contrast, most left arm participants did not exhibit any non-use. This last result is in line with previously reported differences in use between affected left and right hand(Rinehart et al., 2009). It also indirectly support the learned non-use hypothesis, because our (right-handed) left hemiplegic participants presumably have few opportunities to experience negative consequences following use of their paretic arm in daily activities. 36 This result needs to be independently reproduced however because we found no such difference between arms with the nuAAUT. As a measure of arm use, BART has several advantages compared to existing instruments such as the MAL and AAUT. First, it provides an objective and quantitative measure of voluntary paretic arm use in daily life post stroke. In contrast, MAL scoring is based on participants’ recall and is subjective. Second, BART exhibits excellent reliability for repeated measures, making it ideal for examining the effectiveness of rehabilitation on non-use. In contrast, the AAUT, is best used only once, or at most only infrequently at it was in the EXCITE trial (Winstein et al., 2003), because the spontaneous use condition must be covert. Third, BART is simple and timesaving. Unlike the MAL, which takes at least an hour to administer, a single BART session takes less than 15 minutes. There are nonetheless several limitations of the present study. First, the amount of non-use quantified with the AAUT may be due at least in part to a learning effect in AAUT since the constrained use condition is given soon after the spontaneous use condition. We indeed found a small improvement on the tasks that the participants could perform (score > 0) in the cAAUT condition (mean score 3.61 ± 0.11 on these tasks) compared to the sAAUT condition (3.26 ± 0.13 on these tasks; Paired t-test: P < 0.0001). Thus, a small learning effect cannot be excluded, however, this value (0.35 in average) is small compared to the average nuAAUT = 1.67 in our participants. Second, because of the difficulty in recruiting a large group of left-hand dominant participants before stroke, we only developed BART for participants who were right hand dominant before stroke. Third, we did not control for compensatory arm movements such as excessive shoulder elevation and abduction during reaching(Steenbergen et al., 2000), and we observed that several 37 participants were using such compensatory strategies. Preventing such movements may further reduce arm use. Fourth, an important limitation of our current BART system is the time allowed to complete the reaching movements, which was set to a single value of 1.2 sec. In our beta version of BART, we noticed that without a time limit in the constrained condition, participants post-stroke reached almost all targets all the times. We thus set a time limit of 1.2 sec. Data from the beta version revealed that median movements time was 845ms with lower quartile 738ms, and upper quartile 1008ms. We therefore hypothesized that 1.2 second could discriminate performance of post-stroke individuals with mild to moderate impairments from non-disabled controls. However, with this 1.2 sec time limit, 13 out of 24 participants post-stroke still showed control-like performance as measured by cBART. This suggests that 1.2 sec is too long for most of these 13 participants. In contrast, eight participants chose not to use the paretic arm at all in the second session with BART, as measured by sBART. This suggests that 1.2 sec is too short for at least of sub-group of our participants. Thus, in future work, we will need to parameterize BART to detect non- use across a large proportion of patients with variable initial characteristics. Fifth, the pointing task in BART mimics only one aspect of upper extremity use (e.g. reaching), but does not include other actions that might be part of daily use, such as stabilizing, supporting, grasping, tapping, etc. This may be the reason why we found a somewhat lower than expected correlation between the AAUT and BART in the spontaneous condition. In future work, improved systems to automatically assess arm and hand use could, for instance, present tools that allow for grasping at different spatial locations. Task-based rehabilitation robot, i.e.(Choi et al., 2009; Choi et al., 2011), could be modified for this purpose. Finally, while administering BART requires very little expertise from the tester 38 (unlike the MAL and the AAUT), the hardware (mini-bird magnetic sensors and projector mounted above table) and software (Matlab for data analysis) needed in the current implementation makes it a research tool only. Cheaper versions will need to be developed for use in the clinic. In addition, the current method to compute use and non-use is somewhat complicated, notably with the 2 nd order logistic model fit. While we could have used a simpler method without model fitting and smoothing, and only based on counts at each target, the proposed method leads the way for the spatial assessment and treatment of non-use. In the current manuscript, to validate BART, we computed an overall non-use value. However, inspection of the maps of figures 2.2 and 2.4 show that we can compute target-by-target, performance, use, and non-use values. Thus, as an example of a strategy to reduce non-use, reaching movements with the hemiparetic arm to targets in areas of greater non-use could be rewarded to a greater extent than reaching to other targets. Besides the Sterr et al. study(Sterr et al., 2002) discussed above, we are aware of only two other studies related to the present work. First, Johnson et al.(Johnson et al., 2011) evaluated non-use in five participants post-stroke by measuring errors during a steering task in affected, unaffected, and bimanual arms conditions. Second, Brown (Brown, 2011) measured arm use (but not non-use) in 13 participants with stroke with a device similar to our BART device with 5 actual reaching tasks involving proximal and/or distal hand movements. In addition to its usefulness as a measurement tool to capture use and non-use of the paretic limb after stroke, BART may be useful to elaborate our understanding of motor 39 control and decision making underlying reaching choice, and notably the factors that affect changes in these choices, including changes in performance due to motor therapy. We have previously proposed and have begun to test an “optimal therapy threshold” hypothesis,(Han, Arbib and Schweighofer, 2008; Hidaka Y. et al., 2012; Schweighofer et al., 2009) according to which a minimum amount of therapy is needed to reach a minimum arm use level such that use and performance continue to improve following therapy. BART can be helpful in determining such a threshold with repeated measurements of use and non- use before, during, and after therapy. Acknowledgements We thank Dr. Young Geun Choi for help with computer programming, Dr. James Gordon for his inputs during development of the task, and Neerav Parikh for help with data collection. This work was in part supported by NIH grants P20 RR020700-01 and R03 HD050591-02 and R01 HD065438-01A2. CEH is in part supported by the WCU program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (R32-10142). RO is in part supported by the Funding Program for Next Generation World-Leading Researchers, Japan and SRPBS, MEXT, Japan. 40 Appendix A Figure A1. Variability in average indifference lines for non-disabled participants. We obtained a distribution of indifference lines with 1000 bootstrapping samples (draws with replacements) of the non-disabled participants. The original indifference line is shown in solid black and the lower-quartile (25% percentile of decision lines) and upper quartile (75% percentile of decision lines) are shown in dashed black. Yellow color represents higher probabilities of indifference lines. 41 Figure A2. Relation between residual recovery (FM) and the magnitude of non-use in 15 participants with stroke in the validity study. Correlation was not significant between FM and nuAAUT ( r = -0.136, P = 0.629, Pearson), and between FM and nuBART (r = -0.192, P = 0.494, Pearson). 42 Chapter 3: Objective Assessments of Upper Extremity Performance, Use and, Non-use After Stroke 3.1. Introduction Proper assessment of performance and use of more-affected arms is important in stroke rehabilitation to evaluate the patient’s current medical status and to establish rehabilitation goals leading to recovery (Chen et al., 2012; Lang et al., 2013; G. Uswatte, Taub, et al., 2006; van Kordelaar et al., 2012). Given that performance does not necessarily reflect how much patients with stroke use the more-affected arm (i.e. non-use, which is a difference between what the individual can do and what the individual does using the more- affected arm)(Han et al., 2013; Sterr, Freivogel and Schmalohr, 2002; Stewart and Cramer, 2013), independent assessments for performance, use, and non-use are necessary for stroke rehabilitation. There are numerous outcome measures commonly used in clinical trials for performance, use, and non-use assessment. The Wolf Motor Function Test (WMFT)(Wolf et al., 2001), a measure of performance, the Motor Activity Log (MAL), the measure of use (G. Uswatte, Taub, et al., 2006; G. Uswatte, Taub, et al., 2005), and the Amount Arm Use Test (AAUT), a measure of use and non-use (Sterr, Freivogel and Schmalohr, 2002; G. Uswatte, Taub, Morris, Barman, et al., 2006), have often been used in stroke clinical trials. Although these tests are low-cost and relatively easy to use, they require long 43 administration times (e.g., around 30~40 minutes for evaluating both arms in the WMFT) and need thorough evaluator training via specialized standardization processes to prevent inter-rater variability or measurement errors (Duff et al., 2015). In addition, they suffer from the evaluators’ and patients’ bias during scoring (G. Uswatte, Taub, et al., 2006; G. Uswatte, Taub, et al., 2005), and some tests (i.e. the AAUT) are not repeatable once participants become aware of this test (Sterr, Freivogel and Schmalohr, 2002; G. Uswatte, Taub, Morris, Barman, et al., 2006). Thus, these clinical tests are need to be carefully use for the longitudinal clinical trials. To address these concerns, we recently developed a novel, simple, and objective assessment tool, the Bilateral Arm Reaching Test (BART), to measure use, performance, and nonuse of the more-affected arm in patients with stroke (Han et al., 2013). This computer-based reaching system is easy to operate and capable of repeatedly assessing upper extremity movements in a relatively short time (around 15 -20 minutes). By asking participants with stroke to reach the targets in 1.2 sec, using either the more-affected or the less-affected arm (spontaneous use condition) or using the more-affected arm only (constraint use condition), we measured performance and use, and computed non-use of the more-affected arm. The BART system showed a moderate correlation with the AAUT in assessments for use, performance, and non-use and good test-retest reliability. However, the 1.2 sec time limit was too long for some patients so that they used their more-affected arm similar to non-disabled people (ceiling effect) or too short for some patients so that they never used their affected arm (floor effect). Therefore, these results together suggest that we need a much shorter movement duration constraint than 1.2 sec to distinguish between participants with mild, moderate, and severe impairments as well as a no time constraint to prevent patients with stroke from exhibiting ‘zero-use’ of the affected arm. 44 3.2. Method 3.2.1. Participants Twenty-two right-handed hemiparetic stroke participants (age 60.14 ± 2.94 years; mean ± SE; 5 Female) and seven right-handed age-matched nondisabled participants (54.71 ± 5.0 years; mean± SE) were recruited. Stroke participants were involved in a phase 1 randomized controlled trial (NCT 01749358) investigating the dose effect of the Accelerated Skill Acquisition Program (ASAP) (Winstein et al., 2016)), a structured, task- oriented upper extremity motor training program, on performance and use of upper extremity in post-stroke. This study analyzed only pre-training (i.e., baseline 1, 2 and pre- training 1) data. Inclusion criteria for the participants with stroke were (1) ischemic or intraparenchymal hemorrhagic stroke without intraventricular extension with confirmatory neuroimaging more than 180 days (6 months) after onset; (2) Age ≥ 21 and no upper limit; (3) impaired upper extremity motor function indicated by the Fugl-Meyer motor and coordination score no less than 19 out of 66 on the total motor score (Fugl-Meyer et al., 1975); (4) no UE neglect as determined by Albert Test; (5) Mini-Mental State Examination (MMSE) score > 24/30; (6) no previous or current UE musculoskeletal injury or conditions that limited use; (7) eligibility for MRI scan. Participants who could not reach the farthest targets (30 cm away from the front edge of the table), or who utilized the BART assessments as motor training (i.e., using the affected arm all the time), or who were left handed before stroke were excluded for further analysis. All subjects were tested to be right handed with the Edinburgh Handedness 45 Inventory. Among 22 participants with stroke, 12 patients with stroke had left-hemisphere damage, and 10 participants with stroke had right-hemisphere damage. Participants with stroke completed three test sessions with a two-week interval between sessions and age- matched disabled completed two test sessions. This study obtained approval from the Human Research and Review Committee of the University of Southern California and informed consent from each participant. 3.2.2. Experimental Setup and Task The time-based BART system consisted of a computer, an over-head projector illuminating virtual targets on the table, two Mini-Bird magnetic sensors (Ascension Technology Corporation), and a belt to prevent compensational trunk movement. The time- based BART assessment included a series of free or forced reaching movements with or without movement duration constraints. Participants completed whole tests in 30-40 minutes depending on their impairment levels. Participants sat on a chair in front of a desk with the home position indicated by a taped black triangle (Figure 3.1A). A home target was shown on the tip of the triangle on the table from the overhead projector and participants were asked to place their index fingers on the home target. A virtual target (white disk, 2 cm in diameter) appeared at one of the pre-defined 35 target locations (5 distances from 10 to 30 cm and 7 angles from 30 to 150 degrees with respect to the front edge of the table). Hand positions were sampled using magnetic sensors attached underneath the right and left index fingers (sampling 100 Hz). The time-based BART system workspace with the target locations is shown in Figure 3.1B. 46 The time-based BART task protocol included three movement duration constraints: No time, Medium, and Fast conditions (Figure 3.1C). In the no time constraint condition, the targets did not disappear until participants reached them, whereas in the medium and fast conditions, the targets disappeared around 1 sec or 0.5 sec after movement onset. This movement duration constraint varied across the targets such that farther targets had longer movement duration than the closer targets. Specifically, when participants used the right arm, the target on the 150 degree angle at the 30 cm distance had a longer movement duration constraint than the target on the 50 degree angle at the 10 cm distance. If participants used the left arm, this time constraint became the opposite. Figure 3.1B shows the movement duration constraint for each target, ranging between 350ms and 580ms for the fast condition. Movement time constraints for the medium condition were 500ms longer than the fast condition; thus, movement constraint ranged from 850ms to 1080ms. These movement duration constraints were estimated from previous reaching data in non-disabled subjects (see Park et al. 2015) (Park et al., 2015). Each condition consisted of the spontaneous use and the constraint use sessions (Figure 3.1C). In the spontaneous use session, participants were asked to choose one arm over the other to reach the targets. If they successfully reached the targets within given time constraints, they heard a pleasant sound and received 1 point. If they failed, they heard an unpleasant sound and received 0 point. Participants were instructed to maximize the number of points by either moving rapidly or switching to the other hand. Each target appeared twice at the same location, resulting in 70 trials. The constrained use session was configured the same as the spontaneous use session, except there was no arm choice. Participants were asked to place their affected arm only in the home target while the less- 47 affected arm was in the lap. Each target appeared once so there were total 35 trials in the constraint use session. In both spontaneous use and constraint use sessions, participants were instructed to reach the target as quickly and as accurately as possible and to stop on the target until the target disappeared, and then return to the home position. The next target appeared at 1 of 35 locations in pseudo-random order after the start position was maintained for 1 s. To avoid an effect of constraint use session on hand choice in spontaneous use session, spontaneous use session was conducted before constraint use session in each condition. For the medium and fast speed condition, a reminder session (similar to the spontaneous use session, but with fewer trials 35), and auditory and visual feedback about success or failure were provided after each trial) was given before spontaneous use session (Figure 3.1C). The test encompassed of 385 trials, including 105 trials in each condition and 35 additional trials for the reminder sessions in the medium and fast conditions. The order for the condition was fixed from the no time constraint to the medium and the fast conditions. All the participants had a short familiarization period on the time-based BART system on their first visit. 48 Figure 3.1. The time-based Bilateral Arm Reaching Test (BART), target location, and protocol. A. The participant seats on the chair with trunk restraint belt. The green circle and the white circle identify the home position and target. For each trial, participants were instructed to reach to the target using either right or left index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. B. The time-based BART workspace with 35 targets and a home target surrounded by a square. Movement duration constraints for each targets are different as a function of target distance and angle. Color shows the time constraint for each target in the medium and the fast conditions. C. The time-based BART2 protocol. No time constraint, Medium, and Fast conditions and each condition has spontaneous use and constraint use sessions. A reminder session is tested before the spontaneous use session for both medium and fast conditions. 49 3.2.3. Measures Clinical Assessment We assessed upper-extremity use and non-use in patients with stroke with the Actual Amount of Use Test (AAUT) using the quality of movement scale (QOM) since this did not rely on participants’ recall as the Motor Activity Log (MAL) does (Taub, Crago and Uswatt, 1998; Taub et al., 2006). In addition, a time score in the Wolf Motor Function Test (WMFT time) was assessed for the performance (Wolf et al., 2001; Wolf et al., 1989). To compute non-use, we conducted the AAUT twice. First was spontaneous AAUT, which we observed participants’ natural arm use when they performed the AAUT items, while in the second session, constraint AAUT, we asked the participants to perform the same AAUT tasks using their more-affected arm and hand only. The averaged AAUT QOM scores in the spontaneous use condition (sAAUT) and in the constrained use condition (cAAUT) reflect use and performance of the more-affected arm and non-use was computed as nuAAUT = cAAUT – sAAUT. A main evaluator in DOSE study administered the AAUT and WMFT time and another evaluator blinded from the participants’ information rated the AAUT scores (sAAUT QOM and cAAUT QOM). These two evaluators in this study passed the standardization process training for the AAUT and WMFT. BART dependent measures We directly measured performance in the constraint use session and use in the spontaneous use sessions for each condition. In the spontaneous use session, we first included successful trials only and then counted the number of targets reached by the more- affected arm for participants with stroke or by the right arm for nondisabled participants in both medium (sBART medium) and fast condition (sBART fast). The no time constraint 50 did not have time limits so we counted the number of targets reached by the more-affected arm or right arm (sBART no time). In the constraint use session, because there was no arm choice, we counted the number of targets that participants successfully reached within the time limit using their more-affected arm (or the right arm for nondisabled participants) in the medium (cBART medium) or in the fast condition (cBART fast). The no time constraint did not have movement duration constraints so all participants obtained 35 max count (cBART no time is always 35). All age-matched nondisabled participants were able to reach all targets successfully using the right arm in the constraint use session for both medium and fast conditions, except for a few unsuccessful trials (9 trials out of 490 trials); thus their count of successful reaching was always close to the max score (on average 34.36 out of 35 targets). We computed non-use by subtracting what participants with stroke do in the spontaneous use session from what they can do in the constraint use session for each condition. We first computed average nondisabled participant hand use in the spontaneous session (Figure 3.3A, top) and used this normative hand use as a mask to the performance data in the constraint use session in the participant with stroke (Figure 3.3C, middle and bottom). We then counted a number of targets after this masking process to know “maximal use of the more-affected arm (max Use).” Next, in the spontaneous use session, we selectively looked at the first 35 reaching trials among 70 reaching trials, since each target appeared once for the first 35 trials and appeared again for the last 35 trials. Given the first 35 targets, we counted the number of targets successfully reached by the more-affected arm (sBART first)(Figure 3.3A, middle and bottom) and computed non-use by subtracting sBART first from “maximal use of the more-affected arm.” We repeated same process for 51 the last 35 targets and finally computed non-use as nuBART = nuBART1 + nuBART2, where nuBART 1 = max Use – sBART first and nuBART2 = max Use – sBART last. 3.2.4. Statistical Analysis In the first analyses, we studied the effect of condition on performance and use. We used mixed effect linear regressions, with condition (the no time, the medium, and the fast) as fixed factors and participants as a random factor. External validities for sBART no time, sBART medium, and sBART fast were tested by correlation with sAAUT QOM, and the validations for cBART no time, cBART medium, and cBART fast were tested by correlation with WMFT time. Because of skewness in WMFT time, a logarithmic transformation was used for the WMFT time (log WMFT time) (Wolf, Winstein, et al., 2006). External validities for nuBART no time, nuBART medium, and nuBART fast were tested by correlation with nuAAUT. Reliability for sBART, cBART, and nuBART in each condition was assessed with intraclass correlation coefficients (ICCs) for 3 visits (visit 1,2 and 3). α values were set at P = .05, and statistics were run using customized code in R and MATLAB. All average results are reported as average ± SEs. 3.3. Results 3.3.1. Patient demographics Demographic information and stroke-specific characteristics of 22 participants with 52 hemiparesis are provided in Table 3.1. Among 22 participants with stroke (60.14 ± 2.94 years; mean± SE), 12 participants had hemiparesis on their right side and 10 participants had hemiparesis on the left side of the body. All of the participants with stroke were chronic and the time interval since stroke varied from 0.47 to 14.38 years (2.75 ± 0.76 years; mean± SE) and they had moderate to mild impairment based on UEFM scores, ranging between 19 and 58 (43.41 ± 2. 29 out of 66 max score; mean± SE). In the control group, there were 7 non-disabled age matched participants including 2 males and 5 females between 44 and 82 years of age (54.71 ± 5.0 years; mean± SE).There was no significant difference between stroke and control groups with respect to age (Mann-Whitney Test, P = 0.304). 53 Subject Age (years) Gender Stroke Duration (years) Dominant Hand More- affected side FM S1 46 Female 1.41 Right Left 57 S2 62 Male 1.39 Right Right 52 S3 63 Male 5.69 Right Right 53 S4 45 Male 1.21 Right Right 38 S5 56 Female 2.1 Right Right 58 S6 69 Male 0.64 Right Right 53 S7 69 Male 14.38 Right Left 40 S8 35 Female 0.53 Right Left 47 S9 72 Male 4.06 Right Left 51 S10 42 Male 0.51 Right Left 28 S11 40 Male 0.55 Right Right 49 S12 60 Female 0.47 Right Left 36 S13 58 Female 0.47 Right Left 35 S14 61 Male 0.98 Right Right 34 S15 47 Male 1.94 Right Left 28 S16 71 Male 9.9 Right Left 39 S17 76 Male 4 Right Right 19 S18 78 Male 1.19 Right Right 50 S19 46 Male 2.08 Right Right 43 S20 78 Male 1.76 Right Right 37 S21 80 Male 4.62 Right Right 53 S22 69 Male 0.66 Right Left 55 Mean± SE 60.14±2.94 17 Males/ 5 Females 2.75±0.76 22 Right 12 Right/ 10 Left 43.41±2.29 Table 3. 1. Demographic information of participants. Abbreviations: SE, standard error; UEFM, Fugl-Meyer; UEFM-cor, coordination subscale of the UEFM–Upper Extremity 54 3.3.2. Performance and Use in the no time, medium, and fast conditions Figure 3.2 shows performance and use as a function of condition in participants with stroke. In the constraint use session, a number of targets successfully reached by the more-affected arm significantly decreased in the fast condition (23.28 ± 1.20; Mean ± SD ) compared to the no time constraint (35 ± 0.00) and the medium condition (32.90 ± 1.20)(mixed-effect linear regression, P < 0.001)(Figure 3.2A). In the spontaneous use session, use of the more-affected arm in the no time constraint was statistically significantly greater than zero mixed-effects linear regression analysis, P <0.001)(Figure 3.2B). The amount of targets successfully reached by the more-affected arm decreased in the fast condition (18.95 ± 2.95) compared to no time constraint (30.68 ± 1.84) and the medium condition (27.50± 1.95)( mixed-effects linear regression analysis, P <0.001). In contrast, the age-matched nondisabled participants showed consistent performance and use of the right arm across the conditions (mixed-effects linear regression analyses, P =0.11 and P = 0.38 for performance and use assessments respectively). 55 Figure 3.2. Use and Performance as a function of condition in participants poststroke. A. Performance measurement in the constraint use session for each condition. Each marker represents each individual with stroke. A number of targets successfully reached by the more-affected arm decreases in the fast condition compared to those in the no time constraint and the medium condition (p<0.001). B. Use measurement in the spontaneous use session for each condition. A number of targets successfully reached by the more- affected arm is significantly greater than 0 in the no time constraint (p<0.001). The fast condition shows lower use of the more-affected arm than the no time constraint and the medium condition (p<0.001). 56 3.3.3. Performance and Arm Choice Pattern on the WorkSpace in Fast Condition Figure 3.3 shows performance in the constraint use session and use in the spontaneous use session in the fast condition for both control and stroke groups. Age- matched nondisabled participants showed perfect performance of the right arm in the constraint use session in the fast condition (Figure 3.3C, top). In addition, these participants used their right arm slightly more than their left arm in the spontaneous session in the fast condition. Specifically, on average, 57% of total targets including center targets were reached with the right arm while 43% of total targets were reached with the left arm in age- matched nondisabled participants (Figure 3.3A, top). Non-use was not shown in nondisabled participants (Figure 3.3D, top). In contrast, use of the more-affected arm in the spontaneous use session in the fast condition varied among participants. Some used their more-affected arms close to maximal use in the constraint use session (Figure 3, bottom), but others used their more-affected arm less than maximal use in the constraint use session, resulting in non-use (Figure 3.3, middle). 57 Figure 3.3. Examples of performance, use, and nonuse in the fast condition over the 2D reaching work space for 2 right-affected participants poststroke and averaged data for age-matched nondisabled participants. Top row represents averaged data for age- matched nondisabled participants, while middle and bottom row represent a participant with nonuse (ID11; session 2, FM 53) and a participant with no nonuse, albeit mild impairment (ID 5; session 2, FM 52). A. Use of the more-affected arm in participants poststroke (middle and bottom) and use of the right arm for right-handed nondisabled participants in the spontaneous use session (normative hand use). Black circles represent the targets that participants successfully reached using their more-affected arm (or right arm for nondisabled). B. Performance measured in the constrained use session. C. Performance after masking with normative hand use of nondisabled participants (panel A, top). D. Nonuse. Black circles represent non-use of the more-affected arm. 58 3.3.4. Validity of time-based BART for Measuring Use, Performance, and Non-use of the affected arm in the Fast Condition For patients with stroke, we validated performance, as measured in the constraint use session, with a log WMFT time. cBART fast showed a strong correlation with log WMFT time (r=-.838, P <.001, Spearman)(Figure 3.4A). cBART medium showed a moderate correlation with log WMFT time (r = -.538, P < .01, Spearman), while cBART no time did not have any correlation with log WMFT time because all participants were able to reach all targets when there was no time constraint. Use measured in the spontaneous use session was correlated with AAUT QOM. The correlation between sBART fast and AAUT QOM was strong (r= .829, P <.001, Spearman)(Figure 3.4B). In contrast, sBART medium showed a moderate correlation with AAUT QOM (r = .538, P = .009, Spearman), but sBART no time did not show a significant correlation with AAUT QOM (r = 0.363, P = .096, Spearman). Non-use computed in the fast condition (nuBART fast) showed moderate correlation with non-use computed by AAUT (nuAAUT QOM) (r= .548, P =008, Pearson)(Figure 3.4C). In contrast, neither nuBART medium nor nuBART no time showed the correlations with nuAAUT QOM (r = .407, P = .06, Pearson, between nuBART medium and nuAAUT QOM, r= .129, P = .569, Pearson, between nuBART no time and nuAAUT QOM). 59 Figure 3.4. Validation of BART2 system. The correlation between sBART fast and sAAUT QOM is strong (A), the correlation between cBART fast and WMFT time – all items is strong (B). The correlation between cBART fast and WMFT time – Hand related items (C) is stronger than the correlation between cBART fast and WMFT time – Arm related items (D). 60 3.3.5. Test-Retest Reliability The stroke group showed excellent test-retest reliability in sBART fast (ICC = 0.960, P < .0001) and cBART fast (ICC = 0.962, P < .0001) and showed acceptable reliability in nuBART fast (ICC=0.739 P < .0001) (Figure 3.5). sBART fast was 17.22 ± 2.95 (the number of targets that participants successfully reached out of total 70 targets; Mean ± SE) for the first session, 17.23 ± 2.95 for the second session, and 19.00 ± 2.82 for the third session. Similarly, cBART fast was stable across the sessions; 17.61 ± 2.40 for the first session, 18.87 ± 2.28 for the second session, and 19.74 ± 2.10 for the third session (Figure 3.5A and 5B). nuBART fast was 0.59 ± 0.82 for the first session, 1.39 ± 0.76 for the second session, and 1.25 ± 0.67 for the third session (Figure 3.5C). In addition, the control group showed excellent test-retest reliability in sBART fast (ICC = 0.954, P < .001) with scores for the successful use of the right arm of 38.86 ± 2.08 for the first session and 38.43 ±1.21 for the second session. Nondisabled participants were able to reach all the targets successfully across all conditions except for a few failures; thus, cBART fast was close to 35 in all non-disabled participants. 61 Figure 3.5. Test-Retest Reliability for cBART fast (A), sBART fast (B), and nuBART fast (C). X-axis is tests conducted every 2 weeks. Y-axis is a number of targets successfully reached by the more- affected arm. Both cBART fast and sBART fast show excellent test-retested reliability and nuBART fast shows acceptable test-retest reliability. 62 3.4. Discussion This study objectively quantified performance, use, and non-use of the upper extremity during the target-reaching task using the time-based BART system in both participants with stroke and age-matched nondisabled participants. We assessed use of the more-affected arm during the spontaneous use session in which the participants were asked to use either the right or the left arm. We also measured performance during the constraint use session in which the participants were asked to use the right arm only (for age-matched nondisabled) or the more-affected arm only (for participants with stroke). We then computed non-use by subtracting use in the spontaneous session from performance in the constraint use session after masking process with arm use in non-disabled participants. Our results demonstrated that people with mild to moderate hemiparesis showed decreases in use and performance of the more-affected arm when they encountered a more challenging task environment such as the fast condition. We found the strong correlations between clinical tests and our time-based BART system, especially in the fast condition for both use and performance assessments. In addition, the correlation between nuAAUT and nuBART fast was moderate. Test-retest reliability was excellent in both sBART fast and cBART fast and was acceptable in nuBART fast, suggesting that the fast condition in the time-based BART system was sufficient to be an alternative to clinical tests to objectively and repeatedly measure performance, use, and non-use in upper extremities in patients with stroke. 63 As a measure of performance, use, and non-use of the more-affected arm, the time- based BART has several advantages compared with existing clinical tests such as the WMFT, the MAL, and the AAUT. First, the time-based BART is an objective, accurate, and quantitative measurement. In contrast, clinical assessments are subjective due to reliance on participants’ recall and on subjective judgements of the evaluators and are less accurate due to measurement errors that evaluators make when pressing the stopwatch during the WMFT. Second, our current system exhibits excellent reliability for repeated measures at least for 1 month of baseline tests in our experiment. Unlike the AAUT, which is unsuitable for repetitive assessments, the time-based BART could possibly be used for repeated assessments in longitudinal studies to measure the changes before and after interventions. Third, the time-based BART saves administration time. The time-based BART is easy to operate and requires a short test time of less than 40 minutes for all performance, use, and non-use assessments. Furthermore, it requires only a simple data analytic process (i.e., counting the number of targets that are successfully reached with the affected arm) so it can save evaluators’ time. However, our system is a simple reaching task, which does not cover other components of upper extremity movements, such grasping or stabilizing required for the activities of daily living. Nevertheless, the time-based BART system, especially the fast condition, not medium condition or no time constraint, shows a strong correlation with both sAAUT QOM (r= .829, P <.001) and log WMFT time (r=-.838, P <.001). In addition, nuBART fast shows a moderate correlation with nuAAUT (r=.545, P = .008), thus it can indirectly measure beyond the simple pointing movements. 64 We believe that the time constraint in our system plays an important role to measure for performance, use, and non-use. During performance assessment in our previous study using a 1.2 sec time constraint, we found that 13 out of 24 participants with stroke showed control-like performance, as measured in the constraint use session. We found similar results in current study that, in constraint use session in medium condition with around 1 sec time constraint, 13 out of 22 participants with stroke showed control-like performance. However, this ceiling effect disappeared in the fast condition as none of our participants with stroke achieved a perfect score. Furthermore, cBART fast showed a strong correlation with log WMFT and demonstrated excellent test-retest reliability (ICC for cBART fast was 0.962), suggesting that cBART fast differentially measures performance on a diversity of patient populations. During use assessments, our previous study showed that 8 out of 24 participants with stroke never used their more-affected arm at all in spontaneous use session with 1.2 sec time constraint. In contrast, in current study, none of our participants with stroke showed ‘zero-use’ of the more-affected arm in the spontaneous use session under the no time constraint and only 2 out of 22 participants did not use in the fast condition. We decreased this ‘zero-use’ by gradually introducing the challenging task conditions (i.e., no time constraint to fast condition) and showed that sBART fast captured spontaneous use of the more-affected arm well, as supported by the strong correlation between sBART fast and AAUT QOM (r= 0.829, p<0.0001). In addition, sBART fast shows excellent test- retest reliability, supporting applicability of out time-based BART system for use assessments. 65 Our findings that use of affected arm is flexible and participants switch from affected arm to unaffected arm if affected arm showed low task successes, is in line with previous research that arm choices change with respect to the task environments in young healthy individuals and individuals with stroke (Habagishi et al., 2014; Schweighofer et al., 2015). Stoloff et al. found that people were less likely to use right arm when right arm was less successful for the reaching task compared to left arm by modulating success rate for right arm using different target sizes (Stoloff et al., 2011). Emily et al also found that the individuals with stroke used affected arm less when the tasks became more challenging from simple task requiring reaching only to complex task requiring reaching and grasping (Emily, 2011). There findings and our results together suggest that we need a challenging environment to accurately measure performance and use of upper extremity in patients with stroke. Though we showed the excellent test-retest reliability and external validity of our system, several limitations remain. First, the participants with stroke that we recruited had mild to moderate impairments and they could voluntarily move their more-affected arms to reach all targets on the time-based BART workspace. If participants had difficulty moving their upper arm so could not reach the targets, these participants were not eligible for our testing. On the other hand, participants with very mild impairment were able to obtain the highest score on cBART fast so that our system may not be able to detect the improvements after intervention for these participants clearly. Second, because of the difficulty in recruiting prestroke left-hand dominant participants, we tested only the time- based BART for prestroke right-hand dominant individuals. Third, the order of the condition was fixed; thus, all the participants followed the same order (starting with the no 66 time constraint and finishing with the fast condition). Positive or negative reinforcements after successful or failed reaching during the first two conditions might have influenced arm use in the fast condition. Specifically, if participants experienced frequent failure in the medium condition, they would hesitate to use the affected arm in the fast condition or vice versa. Though we knew the potential effect of the sequence of the conditions, we kept using a fixed order to prevent ‘zero-use’ of the affected arm in spontaneous use sessions, as we observed in our previous study. There are several possible indications as to how the time-based BART could be used in the clinics. First, as shown earlier, our system can be used to assess performance, use, and non-use for each treatment session. The time-based BART records the all data and this recorded history would provide information whether individuals poststroke show improvements after intervention. Second, the time-based BART can provide the locational insight that where patients have difficulty using their affected arms and where they actually use affected on the workspace. This locational information can be used as a basis for setting the rehabilitation strategies. For instance, if a patient with stroke has the ability to reach all the targets in the constraint use session of the fast condition, but chooses the affected arm for few targets in the spontaneous use session of the fast condition, therapists can encourage the patient to select his or her affected arm for the targets the patient previously avoided using the affected arm. With many repetitions, we believe that it is possible to develop habitual patterns of the affected arm use in the spontaneous sessions and hope that it could be carried out in the real world. Third, the time-based BART system can also be utilized for treatment purposes. Recently, we conducted two days of intensive reach training using our computer based reaching system after modifying the target locations and time 67 constraints on the system. Participants with chronic stroke showed a 22.8% decrease in movement duration and a 23% increase in box and block test at retention test (one day after intensive training) compared to baseline test, and these improvements lasted up to 1 month after training (Park et al., 2015). Last, BART may be useful to help us elucidate underlying mechanisms of arm choice and identify the factors influencing use of the more-affected arm in patients with stroke. Recently our group found that effort and success influence arm choice in young healthy adults (Schweighofer et al., 2015), and we expect that these factors differentially affect arm choice in patients with stroke. 68 Chapter 4: Habitual versus Adaptive Use of Affected Arm in Right and Left Hemiparesis 4.1. Introduction Spontaneous use of the more-affected arm in individuals with stroke is a crucial indicator of recovery and effectiveness of treatment intervention in stroke rehabilitation (Barker and Brauer, 2005; Chen et al., 2012). Interestingly, the spontaneous use of the more-affected arm varies among individuals with stroke. Some individuals keep using their more-affected arm after treatment, while others avoid using more-affected arm in the real world, even though they have a residual capacity to use it (Hidaka et al., 2012; Stewart and Cramer, 2013). Therefore, it is important to understand the mechanisms underlying arm choice in individuals with stroke. Several factors influence arm choice in individuals with stroke. Haaland et al. (2012) found that the side affected by stroke influenced the more-affected arm use in right- handed individuals post-stroke (Haaland et al., 2012). When performing instrumental activities in daily life, individuals with right hemiparesis (RH: right-dominant arm is more- affected) used their more-affected right-dominant arm to a greater extent than individuals with left hemiparesis (LH: left-nondominant arm is more-affected) used their more- affected left-nondominant arm (Haaland et al., 2012). Another study found that individuals 69 with LH used their less-affected dominant-right arm to a greater extent than individuals with RH used their less-affected nondominant-left arm during instrumental activities in daily life (Rinehart et al., 2009). Although these findings together suggest that the side of stroke could be one of the factors influencing arm choice, mechanisms underlying discrete arm choices between individuals with RH and individuals with LH have not been extensively investigated. We recently developed a novel model to understand the factors underlying arm choice in nondisabled participants (Schweighofer et al., 2015). The arm choice model in our study was based on a recent framework in which the probability of arm choice is based on comparison of “action values,” for the right and the left arm, in which values are the weighted sum of all expected cost and expected rewards (Han, Arbib and Schweighofer, 2008). Similarly to the cost function in models of optimal control (Todorov and Jordan 2002), values were defined as a combination of biomechanical effort, which is computed from the integral of the motor command (Cos, Belanger and Cisek, 2011; Guigon, Baraduc and Desmurget, 2007) and rewards, reflecting how well participants control these arms (Stoloff et al., 2011). Using this arm choice model, we showed that expected reward (i.e., expected success in hitting the target), expected motor cost (i.e., biomechanical effort), and a general handedness accurately predicted arm choice in nondisabled young adult during rapid reaching (Schweighofer et al., 2015). Here, we reasoned that movement duration is likely to be an important additional predictor of arm choice in individuals with stroke. Because individuals with stroke often show slow movement when they use their more-affected arm (Cirstea and Levin, 2000; Cirstea, Ptito and Levin, 2003), movement duration can differ significantly between the 70 more-affected and the less-affected arms in individuals with stroke. Although individuals with stroke can use their more-affected arm successfully to complete the task, using this arm with longer movement duration increases a cost; thus, they may not select the more- affected arm to use. In addition, we assumed that success of task (i.e., reward) strongly influences arm choice in individuals with stroke than it does in nondisabled people, because the former often experience unsuccessful trials (i.e., they may drop a cup or spill water) using their more-affected arm. This negative experience may also result in decreased use of the more-affected arm (Taub, Uswatte and Pidikiti, 1999). The aim of this study is to understand the mechanisms underlying arm choice in both individuals with right hemiparesis and individuals with left hemiparesis by identifying factors influencing arm choice. For this, we updated our previous model for arm choice in non-disabled young adults to individuals with right- and left- hemiparesis. To account for the difference between arms in movement duration, we added movement duration, a cost for time, as in Hoff’s experiment, as an additional term in the action value (Hoff, 1992). We hypothesize that in a simple reaching task with high task success rates, arm choice for each target depends on the between-arm difference in expected motor effort and expected movement duration and on an overall handedness bias for both individuals with stroke and age-matched nondisabled people. We hypothesize further that in a simple reaching task with low task success rates, choice additionally depends on the between-arm difference in expected task success. To test these hypotheses, we established a reaching task that instructed participants to use the more-affected arm only (constraint use session), or to use either the more-affected or the less-affected arm (spontaneous use session) to reach the targets with three different movement duration constraint conditions. No time constraint 71 led to perfect task success rates all the time, whereas medium and fast condition led to relatively low task success rates. We recorded the actual arm choice in the spontaneous use sessions, measured movement duration and success rate, and estimated effort for each target for each arm in the constraint use sessions. Action values were computed as the weighted sum of expected effort, expected movement duration, and expected success. We adopted the softmax choice model in reinforcement learning theory and rewrote it to a logistic regression model and added random slope and intercept to take account into different characteristics in participants. 4.2. Methods 4.2.1. Participants Stroke participants took part in the DOSE clinical trial (NCT 01749358). DOSE is a phase 1 randomized controlled trial investigating the dose effect of the Accelerated Skill Acquisition Program (ASAP) (Winstein et al., 2016), a structured, task-oriented upper extremity motor training program, on performance and use of upper extremities in individuals with post-stroke. This study analyzed baseline data. Twelve participants with right hemiparesis (RH) (62.83 ± 13.95 years; mean± SD), ten participants with left hemiparesis (LH) (56.90 ± 13.53 years; mean± SD), and seven age-matched nondisabled participants (Control) (54.71 ± 13.22 years; mean± SD) participated in this study. Inclusion criteria for the participants with stroke were (1) ischemic or intraparenchymal hemorrhagic stroke without intraventricular extension with confirmatory neuroimaging more than 180 days (6 months) after onset; (2) Age ≥ 21 and no upper limit; (3) impaired upper extremity 72 motor function indicated by a Fugl-Meyer motor and coordination score for the upper extremity (UEFM) no less than 19 out of 66 on the total motor score (Fugl-Meyer et al., 1975); (4) no visual neglect as determined by Albert Test (Fullerton, McSherry and Stout, 1986) ; (5) Mini-Mental State Examination (MMSE) score > 24/30 (Folstein, Folstein and McHugh, 1975); (6) no previous or current UE musculoskeletal injury or conditions that limited use; (7) eligibility for MRI scan. Exclusion criteria were patients with stroke: (1) who could not reach the farthest targets (30 cm away from the front edge of the table); (2) who utilized the BART assessments as motor training (i.e., using the affected arm all the time); and (3) who were left handed before stroke as tested by the Edinburgh Handedness Inventory (Oldfield, 1971). Forty-five participants were enrolled in the DOSE. However, twenty-three participants were excluded; because (1) we made changes to a target-reaching task (time- based Bilateral Arm Reaching Test (time-based BART), see Method), so first ten participants were tested with a different version of the time-based BART; (2) six participants could not reach the farthest targets; (3) three participants used their more- affected arm all the time to reach the targets; (4) three participants were left handed; and (5) one participant dropped the study. Therefore, data from twenty-two participants with stroke were analyzed for the study. This study obtained approval from the Human Research and Review Committee of the University of Southern California and each participant signed an informed consent. A physical therapist recruited and screened the participants with inclusion and exclusion criteria, and the other physical therapist blinded from the participants’ information administered the time-based BART. These two physical therapists in this study had more than two years of clinical experience. 73 4.2.2. Experimental setup The experimental set-up is shown in Figure 4.1. Participants sat in a wooden chair with a restraining belt to minimize compensatory trunk movements during reaching. At each trial, a target ( white disk in 3 cm of diameter), was projected on the table from an overhead projector. This target appeared at one of the pre-defined 35 target locations, which were spaced in 7 angles, between 30° and 150° with in 30-degree angles and in 5 distances, between 10 to 30 cm with equal 5-cm distances. A home position was indicated by a black triangle taped on the table and a home target was projected on the tip of the triangle. The chair was positioned so that when either hand was on the home target, the corresponding arm posture was given by a 30° shoulder flexion and 60° elbow flexion, as depicted in Figure 4.1A. Two Mini-Bird magnetic sensors (Ascension Technology Corporation) were placed underneath the right and the left index fingers to measure hand trajectory and arm choice (sampling 100 Hz). 4.2.3. Experimental Task All participants performed a target-reaching task using the time-based Bilateral Arm Reaching Test. The time-based BART includes three conditions, no time constraint, medium, and fast conditions, and two sessions in each condition, spontaneous use and constraint use sessions. In the spontaneous use session (SU), the participants were asked to reach target using either their right or the left arm. In contrast, in the constrained use session (CU), the participants were asked to use the more-affected arm only (for stroke groups) or the right dominant arm only (for control group). In the no time constraint condition, the 74 targets did not disappear until the participants reached them, whereas in the medium and the fast conditions, the targets disappeared around 1 sec or 0.5 sec after movement onset. Movement duration constraint in these two conditions varied across the targets, such that farther targets had longer movement duration than the closer targets. These movement duration constraints were estimated from previous reaching data in non-disabled subjects (see Park et al. 2015)(Park et al., 2015). Figure 4.1B shows the movement duration constraint for each target, ranging between 350ms and 580ms for the fast condition. Movement time constraints for the medium condition were 500ms longer than the fast condition; thus, movement constraint ranged from 850ms to 1080ms. Please see Kim et al. (2015) for more detailed information about the experimental set-up and task. At each trial, one of the 35 targets light up. Participants were instructed to move their index finger from the home position to the target as rapidly and accurately as possible and to stop on the target for 0.5 s to prevent over-shooting. After the target disappeared, participants were asked to move their finger back to the home target for the next trial. After the finger stayed within the home target 1s, the home target disappeared, and a new target appeared at 1 of 35 locations. Pleasant auditory feedback was provided following success, i.e., when participants reached the targets in the given time; unpleasant sound was provided following failure to reach the target in the given time. In addition, for motivational purposes, subjects received 1 point for success or 0 point for failure at the end of each trial. During the spontaneous use sessions in the medium and the fast conditions, participants were instructed to minimize unsuccessful trials either by moving fast or by switching to the other arm for the next trial after they missed the target. For the constraint use sessions, participants were instructed to minimize unsuccessful trials by moving fast. There was no failure in the no time constraint condition, so participants always obtained 1 point and a 75 pleasant sound after each trial. In the constraint use session, each target appeared once so there were in total 35 trials. In contrast, in the spontaneous use session, each target appeared twice at the same location, resulting in 70 trials. The time-based BART had two different testing schedules across three testing days (Figure 4.1C and 4.1D). At Day 1, participants had three conditions and each condition consisted of three sessions: two constraint use sessions, one for the less-affected arm and the other for the more-affected arm, and one spontaneous use session (Figure 4.1C). Because of the need to record full movement trajectory to estimate movement time and effort, the targets at Day 1 did not disappear until participants reached the targets. Instead, the participants received auditory and visual feedback informing them of successful or unsuccessful reaching. At Day 2 and 3, participants also had three speed conditions, but each condition consisted of one spontaneous use session and one constraint use session (Figure 4.1D). Targets disappeared after pre-defined time limits in the medium and the fast conditions and the participants received auditory feedback only after each trial. In addition, the spontaneous session was always presented before the constraint use session to prevent bias of arm use. Reminder sessions, similar to the spontaneous use session but with fewer trials (35 trial) and auditory and visual feedback about success or failure, were included to remind the participants of time limits in the medium and the fast conditions. The order of conditions was fixed from the no time constraint, to the medium, and to the fast condition across the testing days (Figures 4.1C and 4.1D). Participants with stroke performed the time-based BART at Day1, 2, and 3 with at least a 2-week interval between testing days, whereas control participants performed at Day 1 and Day 2. 76 77 Figure 4. 1. The time-based Bilateral Arm Reaching Test (BART), target location, and protocol. A. The participant seats on the chair with trunk restraint belt. The green circle and the white circle identify the home position and target. For each trial, participants were instructed to reach to the target using either right or left index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. B. The time-based BART workspace with 35 targets and a home target surrounded by a square. Movement duration constraints for each targets are different as a function of target distance and angle. Color shows the time constraint for each target in fast condition. C. The time-based BART2 protocol. No time constraint, Medium, and Fast conditions and each condition has spontaneous use (SU) and constraint use (CU) sessions. A reminder session is before SU for both medium and fast conditions. 78 4.2.4. Measurements for Arm Choice, Movement Duration, Effort Estimation, and Success For the participants with stroke, arm choice for each target in each condition was measured in the spontaneous use session at Day 3 (Figure 4.1D). Movement duration and effort estimation for both the more-and the less-affected arms for each target in each condition were measured and estimated in the constraint use session at Day 1 (Figure 4.1C). Success in use of the more-affected arm for each target was measured in the constraint use session in both the medium and the fast condition at Day 1,2, and 3. Success in use of the less-affected arm was assessed only at Day 1. For the participants in the Control group, arm choice was measured at Day 2 (Figure 4.1D). Movement duration and effort estimation for both the right and left arms for each target in each condition were measured and estimated in the constraint use session at Day 1 (Figure 4.1C). Success in use of the right arm for each target was measured in the constraint use session in both the medium and the fast condition at Day 1,2 while success in use of the left arm was assessed only at Day 1 (Figure 4.1C and 4.1D). Movement duration was measured based on the tangential velocity profile (Figure 4.1E). We first detected peak tangential velocity. Movement start was defined as the first minimum of 5% of the peak velocity by searching backwards from the time of peak velocity to the start of movement. Movement end was identified by first determining the time that the sensor attached to the index finger was inside of the target, and then searching forwards from the time of peak velocity to find the first minimum 5% of the peak after the sensor was inside of the target. 79 To estimate effort required to move the arm to each target, we performed an inverse kinematics and an inverse dynamics transformation with two planar 2DOF arm model including a shoulder and an elbow joint. Finger positions (x,y) were transformed into joint angles (shoulder and elbow joint angles) using inverse kinematics (Figure 4.1F). Then, the shoulder and elbow torques were estimated using inverse dynamics (Figure 4.1G). All parameters for inverse dynamics were taken from previous research by van Beers (van Beers, Haggard and Wolpert, 2004). We computed “absolute effort” defined by summing thee absolute torques at both joints from movement start to movement end. We also computed “squared effort,” defined by summing the square of torques at both joints from movement start to movement end, which is often used in optimal control models and was used in previous arm choice research (Guigon, Baraduc and Desmurget, 2007; Park et al., 2015; Todorov and Jordan, 2002). We developed a model with both types of estimated effort and found that absolute effort provided a better fit for our arm choice model (based in BIC values, see below), and therefore used absolute effort instead of squared effort in all models below. Success rate was calculated based on whether participants successfully reached the target in the given time (i.e., before the target disappeared) across three testing days (Day 1,2, and 3). For instance, if a participant successfully reached a target on two testing days, but missed it one time, the success rate was 66%. Note that we assumed that all participants with stroke successfully used their unaffected arm throughout the entire reaching tasks. This assumption is based on the following. First success rates of the right arm in the participants in the Control group, as measured in the constraint use sessions at Day 1 and 2, were 100%, 99.78% and 96.04% in the no time, the medium, and the fast condition 80 respectively. Second, in day 1, movement durations of the less-affected arm in the medium and the fast conditions were below the time constraints for 99.05% and 87.67% of the movements. 4.2.5. Analysis for Arm Choice, Movement Duration, Effort Estimation, and Success In the first analyses, we studied the difference between the right and the left arms in movement duration, effort estimation, and success across three conditions in each group. We used mixed effect linear regressions, with arm (right and left) and condition (the no time, the medium, and the fast) as fixed factors and participants as a random factor. Arm choice was studied with mixed effect linear regressions with condition (the no time, the medium, and the fast) as the fixed factors and participants as the random factor. We also compared the participants in the RH group to the participants in the LH group. Arm choice, movement duration, effort estimation, and success of the more-affected arms were analyzed using mixed effect linear regressions, with the hemiparesis side (right hemiparesis or left hemiparesis) and condition (the no time, the medium, and the fast) as the fixed factors and participants as the random factor. In addition, we compared the stroke groups to the control group. Arm choice, movement duration, effort estimation, and success of the participants in the RH group (hemiparetic arm: right) were compared with those of the control group that performed the task with their right arm. Arm choice, movement duration, effort estimation, and success of the participants in the LH group (hemiparetic arm: left) were compared with those of 81 the control group that performed the task with their left arm. We used mixed effect linear regressions with group (hemiparetic or healthy control group) and condition (the no time, the medium, and the fast) as the fixed factors and participants as the random factor. Post hoc analyses were performed using Tukey-test, which corrects for multiple comparisons. Statistical significance levels were set to 0.05. All statistical analyses were carried out using R statistical software. 4.2.6. Arm Choice Model We studied the predictors of arm choice. For this, we developed an arm choice model similar to that developed by Schweighofer et al. for arm choice in healthy subjects. Briefly, arm choice for each target depends on a comparison of “action values,” defined as total expected reward and cost for the left and right arm movements to a target (Park et al., 2015). The probability to select either left or right arm for each target is then computed by a softmax function of the difference in values for each arm. See Han et al. 2008 and Schweighofer et al. for rationale and details (Han, Arbib and Schweighofer, 2008; Park et al., 2015). In the study conducted by Schweighofer et al., arm choice was well modeled by action value composed of expected cost and expected success, together with overall patient-specific handedness bias for the dominant arm. Here, we reasoned that movement duration would be largely different between the two arms, especially in the participants with hemiparesis. Therefore, we added a third term in the value function, a cost for time. In addition, to take into account the largely different participants’ characteristics in our 82 post-stroke population, we used random slope and intercept in mixed effect logistic regression so that each participant had his or her own slope and intercept (see below). We developed the two arm choice models: one is for the participants in Stroke groups (the RH and the LH groups) and the other is for the participants in the Control group. The formats of the two arm choice models are exactly same, but the arm choice model for the stroke groups aims to elucidate which factors are related to the more-affected arm choice in the participants in the RH group and the participants in the LH group. In contrast, the arm choice model for the Control group aims to elucidate which factors are related to the right arm choice in the participants in the Control group. Therefore, in the arm choice model for the Stroke groups, the probability to choose the more-affected arm, for each target i and each participant j is defined by following equation: P(ij,MA)= 1 1+ e −(a j (<Effort ij,MA > − < Effort ij,LA >)+b j (< MD ij,MA > − < MD ij,LA > )+ c j (<Success ij,MA > − <Success ij,LA >)+d j ) where, a j , b j , c j , and d j are the parameters for the differences between the more-affected (MA) and the less-affected (LA) arms in expected effort, expected movement duration, expected success, and general hand preference bias for subject j. The parameters for each participant j (a j , b j , c j , and d j ) consist of the mean of all participants (fixed effect:a μ , b μ , c μ , and d μ ), and how much each participant (random effect) deviates from the mean of all participants. The brackets < > indicate expected values of effort, movement duration, and success for the reaching movement to target i in subject j, and measured and/or estimated based on previous movements to this target at Day 1. The intercept parameter d j can be interpreted as a hand preference bias for each subject j, and it is a target-and movement-independent. This constant intercept will bias choice to all targets when it is 83 different from 0. Taken together, in our arm choice model, if the probability is greater than 0.5, the more-affected arm was chosen, while the less-affected arm was chosen when the probability was less than 0.5. The probability of choosing the less-affected arm is simply P(ij,LA) =1- P(ij,MA) In addition, in arm choice model for Control group, the probability to choose the right arm, for each target i and each participant j is defined by following equation: P(ij,right)= 1 1+ e −(a j (<Effort ij,right − Effort ij,left >)+b j (<MD ij,right − MD ij,left >)+ c j (< Success ij,right − Success ij,left >)+d j ) , where, a j , b j , c j , and d j are the parameters for the differences between the right and the left arms in expected effort, expected movement duration, expected success, and general bias for subjects j. Our arm choice model estimates these parameters. The probability of choosing the left arm is simply P(ij,left) =1- P(ij,right) We developed and selected possible choice models using a forward-selection approach. We started with the simplest (base) model with random intercepts. Next, we developed the models with one predictor: the effort model (difference between right and left arm in effort), the success model (difference between right and left arm in success rate), the movement duration model (difference between right and left arm in movement duration). Then we extended the model by including two predictors among effort, success, and movement and then developed the full model including three predictors. In addition, 84 for the comparison between the participants in the RH group and the participants in the LH, we included additional term, group (RH group and LH group) as a factor. For the model selection, we first minimized the Bayesian Information Criterion (BIC). If the lowest BICs were similar for several models, we compared the models by conducting a Likelihood Ratio Test (LRT) using the R ANOVA function. Once we found the best-fit model, fixed- effect parameters a μ , b μ , c μ , and d μ of the best-fit model were further analyzed. 4.2.6. Effect of Failure on Arm Choice Our model analysis revealed that LH participants showed significant sensitivity to expected success (in other words, expected failures, see Results), and that, in contrast, RH participants seemed to consider expected success less when choosing their more-affected arm. To verify the actual impact of such sensitivity to expected failures, we studied the proportion of movements with the more-affected arm for which the participants changed arm choice because of the expected failure (switch to the less-affected arm) vs the proportion of movements for which the participants used the same arm (stay with the more- affected arm). We first considered the targets for which participants used their more- affected arm to reach in the SU under no time constraint. Next, among these targets, we computed how many targets were unsuccessfully reached with the more-affected arm in the CU in the fast condition. Finally, among these targets of unsuccessful reaching with the more-affected arm in the CU in the fast condition, we studied arm choice in the SU in the fast condition. A student t-test was used to find the group difference (RH and LH) in the different responses to failure (stay and switch). 85 4.3. Results 4.3.1. Participant demographics and clinical data Table 1 summarizes demographics and clinical data. The RH, the LH, and the Control groups did not significantly differ in age (p=.387, Kruskal-Wallis Test). There was no difference between the LH and the RH groups in upper extremity motor impairment (LH: 41.60 ± 3.31, RH: 44.92 ± 3.21, Mann-Whitney U, p = 0.456) or in time post-stroke (LH: 3.43 ± 1.53, RH: 2.18 ± 1.67, Mann-Whitney U, p = 0.497). The FM scores of the patients in this study ranged from 19 to 58, indicating moderate to mild motor impairment. 4.3.2. Arm Choice Figure 4.2 and 4.3 show arm choices in the no time, the medium, and the fast conditions for the RH, the LH, and the Control groups. The participants in the RH group showed a significant decrease in more-affected right arm choice compared to right arm choice of the participants in the Control group (mixed-effects linear regression analysis, P < 0.01)(Table 4.2). The participants in the RH group used their less-affected left arm more than the more-affected right arm (mixed-effects linear regression analysis, p < 0.001). Interestingly, this arm choice pattern did not change across the conditions. There was no an effect of condition on right arm choice (mixed-effects linear regression analysis, p =1)(Figure 4.2 right and Figure 4.3A). The participants in the LH group showed no statistical significant decrease in more- affected left arm choice compared to left arm choice of the participants in the Control group 86 (mixed-effects linear regression analysis, P = 0.20)(Table 4.2). However, arm choice pattern changed as a function of condition in the participants in the LH group (mixed- effects linear regression analysis, p <0.001). In the no time constraint, the more-affected left arm choice in the participants in the LH group was as similar to the left arm choice in the participants in the Control group. However, unlike arm choice among the participants in the RH group, the more-affected left arm choice decreased as the movement duration constraint became short in the fast condition. Post-hoc analysis revealed that the more- affected arm use in the fast condition was lower than both the medium and the no time constraint conditions (p<0.001)(Figure 4.2 left and Figure 4.3C). Participants in RH and participants in LH showed different more-affected arm choice with respect to condition. Post hoc analysis revealed that there was no difference between the more-affected arm choice in the RH group and in the LH group for the no time and the medium conditions (p=0.47 for no time and p=0.21 for medium). However, in the fast condition, the more-affected left arm choice in LH group was 34% lower than the more-affected right arm choice in RH group (p < 0.01)(Table 2). The participants in the Control group, the right arm choice was significantly higher than the left arm choice (mixed-effects linear regression analysis, p<0.001), but there was no effect of condition on arm choice (mixed-effects linear regression analysis, p = 0.38) (Figure 4.2 middle and Figure 4.3C). 87 Figure 4.2. The time-based Bilateral Arm Reaching Test (BART), target location, and protocol. A. The participant seats on the chair with trunk restraint belt. The green circle and the white circle identify the home position and target. For each trial, participants were instructed to reach to the target using either right or left index finger as quickly and accurately as possible. Magnetic sensors were attached to the tips of index fingers to record choice of hand and performance. B. The time-based BART workspace with 35 targets and a home target surrounded by a square. Movement duration constraints for each targets are different as a function of target distance and angle. Color shows the time constraint for each target in fast condition. C. The time-based BART2 protocol. No time constraint, Medium, and Fast conditions and each condition has spontaneous use (SU) and constraint use (CU) sessions. A reminder session is before SU for both medium and fast conditions. 88 4.3.3. Movement Duration Figure 4.3 shows movement durations in the no time, the medium, and the fast conditions for the RH, the LH, and the Control groups. Although our instruction was always same across the conditions- “move as rapidly and accurately as possible,” our participants appeared to move more rapidly in the fast condition than in the no time constraint, as shown in a previous study (Dejong and Lang, 2012). The participants in the RH group showed longer movement duration with the more- affected right arm compared to the less-affected left arm (mixed-effects linear regression analysis, p < 0.001)(Table 4.2). There was an effect of condition on movement duration and post-hoc analysis revealed that for both arms, the fast and the medium conditions had shorter movement duration than the no time constraint (p<0.001), but movement duration was not significantly different between the medium and the fast condition (p=0.96)(Figure 4.3). Similarly, the participants in the LH group showed longer movement duration with the more-affected left arm compared to the less-affected left arm (mixed-effects linear regression analysis, p < 0.001) (Table 4.2). There was an effect of condition on movement duration and post-hoc analysis revealed that both arms showed a decrease in movement duration from the no time constraint to the medium (p<0.001), and from the medium to the fast condition (p<0.001)(Figure 4.3). Overall, movement durations of the more-affected arm were not significantly different between the participants in RH group and the participants in the LH group (mixed- effects linear regression analysis, p > 0.1)(Table 4.2). However, the effect of condition on 89 movement duration was different between groups. Post-hoc analysis revealed that the participants in the LH group showed longer movement duration than the participants in the RH in the no time condition (p<0.05), while the two groups did not show different movement durations in the medium and the fast conditions (p>0.2 and p>0.4, for the medium and the fast conditions respectively). Among participants in the Control group, overall movement duration for the right arm was slightly faster than movement duration for the left arm (mixed-effects linear regression analysis, p < 0.001)(Table 4.2). Both arms showed decreases in movement duration as a function of condition (mixed-effects linear regression analysis, p < 0.001). Post-hoc analysis revealed that movement duration for both arms decreased from the no time constraint to the medium condition (p<0.001), and from the medium to the fast condition (p<0.001)(Figure 4.3, Table 4.2). 90 Figure 4.3. Choice, Movement Duration, Effort, and Success. Each column shows averaged choice, MD, effort and success for both the right and the left arm in the RH (A), the LH (B), and the Control (C) groups. Each box shows choice, MD, effort, and success change for the right and the left arms as a function of condition. 91 4.3.4. Effort Figure 4.3 shows efforts for both arms in the no time, the medium, and the fast conditions for the RH, the LH, and the Control groups. Among participants in the RH group, effort in the more-affected right arm was slightly higher than effort in the less- affected left arm (mixed-effects linear regression analysis, p < 0.05)(Table 4.2). There was an effect of condition on effort and post-hoc analysis revealed that both arms showed significant increases in effort from the no time to the medium (p<0.01), and to the fast (p<0.001), but effort in the medium did not differ from effort in the fast condition (p>0.5)(Figure 4.3, Table 4.2). Among participants in the LH group, there was no difference in effort between the less-affected right and the more-affected left arm (mixed-effects linear regression analysis, p >0.5). There was an effect of condition on effort and post-hoc analysis revealed that both arms showed significant increases in effort from the no time to the medium (p<0.05), and to the fast (p<0.05), but effort in the medium and the fast conditions did not differ (p=0.758) (Figure 4.3, Table 4.2). Overall, the more-affected arm efforts between the participants in the RH group and the participants in the LH group were not significantly different (mixed-effects linear regression analysis, p > 0.1). The effects of the condition on effort for both stroke groups were significant and post-hoc analysis revealed that, for the participants in both the RH and the LH groups, the fast condition required more effort than the no time constraint (p<0.001), but the medium condition did differ from the no time (p>0.1) and from the fast conditions (p>0.1)(Figure 4.3, Table 4.2). 92 Among participants in the Control group, effort also varied with respect to condition, but there was no difference in effort between the left and the right arm (mixed- effects linear regression analysis, p >0.5). There was an effect of condition on effort and post-hoc analysis revealed that, for both arms, the fast condition required more effort than the no time constraint (p <0.01), but the medium did not differ from both the fast (p>0.1) and the no time constraint (p>0.1) (Figure 4.3, Table 4.2). 4.3.4. Success Rates Figure 4.3 shows success rates for both arms in the no time, the medium, and the fast conditions for the RH, the LH, and the Control groups. The participants in the RH group showed a significant decrease in success rate for the more-affected right arm as the movement duration constraint became short. The success rate in the medium was significantly lower than the success rate in no time (p<0.001), but higher than that in the fast condition (p<0.001) (Figure 4.3, Table 4.2). The participants in the LH group also showed a significant decrease in success rate for the more-affected left arm as the movement duration constraint became short. Success rate in the medium was significantly lower than success rate in the no time (p<0.001), but higher than success rate in the fast condition (p<0.001) (Figure 4.3, Table 4.2). Overall, success rates for the more-affected arm between the participants in the RH and the participants in the LH were not significantly different (mixed-effects linear regression analysis, p> 0.1). There were effects of condition on success rate for the more- affected arm in the participants in both the RH and the LH groups. Post-hoc analysis 93 revealed that success rates between the participants in the RH and the participants in the LH groups were not different in the medium (p >0.5), but those for participants in the LH group were lower than for the participants in the RH group in the fast condition (p < 0.05)(Table 4.2). Note that we assumed that success rates for the less-affected arms were 100% on our first approximation (see method). In addition, success rates for both right and left arm were almost perfect in Control with a few unsuccessful trials (62 out of 1372, 4.5%). 4.3.5. Difference between Arms in Movement Duration, Effort, and Success Rate Figure 4.4 shows z-transformed differences between arms in movement duration, effort, and success rate for each target in the fast condition in each group. The more- affected arm choice (for the RH and the LH groups) and the right arm choice (for the Control group) in the fast condition are also shown in Figure 4.4. Overall, it is appeared that the right arm (for the Control group) or the more-affected arm (for the Stroke group) are chosen when z-transformed differences are negative for effort and movement duration for all the RH, the LH, and the Control groups. In addition, for the RH and the LH groups, positive value of z-transformed difference in success seems to link to the more-affected arm choice (Figure 4.4A and 4.4D). 94 Figure 4.4. The z-transformed differences between arms in movement duration, effort, and success rate for each target in the fast condition. Each target is representing the averaged z-transformed value of difference between arms for all participants in each group for Effort (A), MD(C), and Success rate (D). The more-affected arm choice (for the RH and the LH groups) and the right arm choice (for the Control group) are shown in (A). 95 4.3.6. Predicting Arm Choice: Effort and Success Differentially Influence Arm Choice Table 4.3 shows the models with fixed-effect coefficients from the mixed-effects logistic-regression analysis for participants in the RH and the LH groups. We first tested the models with a single predictor. Model 1 with expected effort showed better fit (BIC = 3,521) compared with both Model 2 with expected success (BIC =5,677) and Model 3 with expected movement duration (BIC= 4,778)(Table 4.3). The fixed-effect coefficient in each model was statistically different from 0, but interaction with group was significant only in Model 3, indicating that the fixed-coefficient of expected success in the LH group was higher than that of the RH group (p< 0.05, Table 4.3). We then tested the model with expected effort, expected movement duration, and expected success (Model 8) and found that BIC significantly decreased (BIC = 3,232 in Table 4.3; comparison with Model 1, LRT p <0.001). The addition of interactions with group did not improve the fit, except that Model 10 including interaction between expected success × group improved the fit (BIC=3,323; comparison with Model 8 LRT p <0.05). The full model including all three predictors and their interactions did not further improve the fit (Model 12 in Table 4.3, BIC=3,339; comparison with Model 10, LRT p>0.5). Therefore, our final model was effort + movement duration + success + success × group (model 10 in Table 4.3). In this final model, the fixed-effect coefficients of expected effort and expected movement duration were statistically greater than 0 (p<0.001 for both parameters) and these coefficients did not differ between the RH and the LH groups (Figure 4.5A). However, the fixed-effect coefficient of expected success in the RH group was not 96 significant (not greater nor less than zero, p>0.1), while it was significantly greater than zero in the LH group (p<0.001)(Figure 4.5A), indicating expected success differentially influenced the more-affected arm choice between the RH and the LH group. The best-fit model for the RH group was effort + movement duration, whereas the best-fit model for the LH group was effort+ movement duration + success. The signs of the model parameters show that the probability of the more-affected arm choice increases when; 1) for the participants with RH, expected effort and expected movement duration are lower and shorter for the more-affected arm than those for the less-affected arm; 2) for the participants in the LH group, expected success is additionally higher for the more-affected arm than that for the more-affected arm choice (Figure 4.5A). Table 4.4 shows the models with fixed-effect coefficients from the mixed-effects logistic-regression analysis for participants in the Control group. The model with expected effort (Model 2 with BIC=717 in Table 4.4) significantly decreased BIC compared to the model with constant term only (Model 1 with BIC=1874 in Table 4.4). Adding expected movement duration (Model 3 in Table 4.4) further decreased BIC (BIC=696), but addition of expected success (Model 4 in Table 4.4) did not improve the fit (BIC=731). In addition, full model with all predictors (Model 5 with BIC=710 in Table 4.4) did not improve the fit (comparison to Model 3 LRT>0.05); therefore, our final model for the Control group is Effort + Success (Model 3 in Table 4.4). The fixed-effect coefficients in the final model were statistically significant (greater than 0, p<0.001 for both parameters)(Figure 4.5B). The probability of the right arm choice is increased by the following: lower expected effort for the right arm than for the left arm and shorter movement duration for the right arm than the left arm. 97 Figure 4.5. Comparison of the fixed-effect coefficients for best-fit models for the Stroke and the Control groups. A. The computational model for the Stroke group shows that effort and movement duration determine the more-affected arm choice for both the RH and the LH groups. Success additionally determines the more-affected arm choice for the LH group. The fixed-effect coefficients, that are statistically significantly different from zero, are indicated by * (p<0.001). The fixed-effect coefficient for success in the LH group is significantly greater than in the RH group, as noted by ** (p<0.01). B. The computational model for the Control group shows that effort and movement duration determine the right arm choice. The fixed-effect coefficients for effort and movement duration are significantly different from zero, are indicated by * (p<0.001). Note that sings for effort and movement duration are negative whereas sign for success is positive. 98 4.3.7. Model accuracy and simulation of choice data for the models To verify whether our final model sufficiently explains the more-affected or the less-affected arm choice, we tested goodness of fit using accuracy, which was calculated by comparing actual arm choice that we observed during the experiment (actual choice) to the model arm choice that the best model estimated from the model (model choice). The cut-off point for the model choice was set at 0.5; thus, above this value was the more- affected arm choice (for the Stroke group), while below this value was the less-affected arm choice. The Control group was the same; above cut-off was the right arm choice whereas below was the left arm choice. Higher accuracy indicated better model fitting. The accuracy for the RH group and the LH group, and the Control group were 85.34%, 82.09%, and 90.89 respectively. Thus, the Control group has greater model fit compared to the LH and the RH groups. Because the largest parameter, the effort parameter, is lower in the Stroke group compared to the Control group, there is more randomness in the arm choice in the Stroke group, explaining the decreased accuracy. Figure 4.6 shows the actual choice and model predicted choice. The model predicted choice was estimated from the best-fit model for each group (the RH, the LH, and the Control group). The best-fit model for the Stroke group (Figure 4.6B), which includes effort and movement duration as significant predictors for the RH group and additionally includes a success for the LH group, explains well the actual choice in the RH and the LH groups (Figure4. 6A). To verify the effect of success on the arm choice in the LH group, we removed the success term from the best-fit model. Figure 4.6C shows that the model choice from this reduced model is not as accurate in predicting choice compared to the best-fit model, especially for the no time and the medium conditions. The best-fit 99 model for the Control group, which includes effort and movement duration as significant predictors (Figure 4.6E), explains well the actual choice (Figure 4.6D). Adding success on the best-fit model does not change the model predicted choice compared to the best-fit model (Figure 4.6F). 4.3.8. Habitual vs. Adaptive arm choice strategies between RH and LH Figure 4.7 shows percentages of switch and stay for the RH and the LH groups after normalization with total failure of unsuccessful reaching in the constraint use session in the fast condition. Although the percentage of stay did not reach the statistically significant difference between the RH and the LH group (mean stay for the RH and the LH groups were 61.60 ± 10.67 and 43.54 ± 12.09, t-test, p > 0.1), the percentage of switch was significantly different between groups (mean switch for the RH and the LH groups were 21.73 ± 7.30 and 56.45 ± 12.09, t-test, p<0.05). This result indicates that the LH group switched to the less-affected arm more frequently in the fast condition than that of the RH group, to prevent unsuccessful reaching that often occurred after using the more-affected arm in the fast condition. 100 Figure 4.6. Actual choice and model predicted choice. Actual choice is measured from the experiment and model predicted choice is estimated from best-fit model. A. Actual choice in the RH (dark gray) and the LH (white) groups. B. Best-fit model estimates the more-affected arm choice well. C. Best-fit model without success estimates the more- affected arm choice less accurately for the LH group compared to best-fit model, especially in the no time and the medium conditions. D. Actual choice in the Control group. E. Best-fit model estimates the right arm choice well. E. Success added to best- fit model does not change model predicted choice compared to the best-fit model. 101 Figure 4.7. Habitual and adaptive use of affected arm in the RH and the LH groups. Dark gray is the RH group and white is the LH group. The LH group shows higher switch rate than the RH group, while they show similar stay rate. 102 Table 4.1. Summary of participant information Variable RH LH Control n 12 10 7 Age (years) 62.83 ± 13.95 56.90 ± 13.53 54.72±13.23 Gender (% female) 11M/1F 6M/4F 2M/5F Year Poststroke 2.18±1.67 3.43±4.84 NA Fugl-Meyer Motor 44.91±11.14 41.60±10.48 NA 103 Table 4.2. Choice, Movement Duration, Effort, and Success for the right and the left arms across conditions in the RH, the LH, and the Control groups. *right vs left; †paretic arm in stroke group vs intact arm in control; § right paretic vs left paretic; RH LH Control Predictors Condition Right Left Right Left Right Left Choice (%) No time 41.60 ± 1.75 58.40 ± 1.75* 54.44 ± 3.27* 45.56 ± 3.27 53.49 ±2.33* 46.51±2.33 Medium 44.32 ± 3.39 55.68 ± 3.39* 62.85 ± 3.26* 37.15 ± 3.26 57.39±2.30* 42.61±2.30 Fast 41.52 ± 3.42 58.48 ± 3.42* 72.68 ± 3.26* 27.32 ± 3.26 61.79±2.31* 39.21±2.31 Ave. 42.39 ±5.34 57.61 ±5.34* 63.17 ± 1.02* 36.83 ± 1.02 57.15 ±4.25* 43.75 ±4.25 Movement Duration (ms) No time 860 ± 12.40* †§ 516 ± 5.67 516 ± 5.19* 1098 ± 20.76 † § 493 ± 6.45* 516 ± 6.12 Medium 759 ± 9.09* † 484 ± 4.43 504 ± 4.47* 889 ± 12.89 † 465 ± 4.54* 468 ± 5.56 Fast 754 ± 17.24* † 467 ± 4.50 467 ± 3.97* 847 ± 13.25 † 468 ± 4.43* 423 ± 4.47 Ave. 792 ± 7.71* † 504 ± 2.94 504 ± 2.76* 791 ± 9.48 † 458 ± 3.15* 470 ± 3.30 Effort No time 1.07 ± 0.02* 1.02 ± 0.02 1.00 ± 0.02 1.01 ± 0.02 1.06 ± 0.02 1.06 ± 0.02 Medium 1.11 ± 0.02* 1.10 ± 0.02 1.06 ± 0.02 1.04 ± 0.02 1.08 ± 0.02 1.10 ± 0.02 Fast 1.15 ± 0.02* 1.11 ± 0.02 1.06 ± 0.02 1.06 ± 0.02 1.12 ± 0.02 1.12 ± 0.02 Ave 1.11 ± 0.01* 1.08 ±0.01 1.04 ± 0.01 1.04 ± 0.01 1.09 ± 0.01 1.09 ± 0.01 Success (%) No time 100 100 100 100 100 100 Medium 92.77 ± 0.64* † 100 100 89.15 ± 0.83* † 99.78 ± 0.31 99.56 ± 0.15 Fast 61.61 ± 1.40 *†§ 100 100 46.21 ± 1.47* †§ 96.04 ± 0.67 94.71 ± 1.10 Ave 85.19 ± 0.61* 100 100 78.01 ± 7.57* † 98.10 ± 0.37 98.62 ± 0.23 104 Mixed-effects logistic-regression results Dependent variable: More-affected arm choice Models (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Effort -2.970 *** (0.414) -2.801 *** (0.353) -2.810 *** (0.353) -3.007 *** (0.398) -2.999 *** (0.399) -2.642 *** (0.258) -2.842 *** (0.341) -2.639 *** (0.258) -2.834 *** (0.341) -2.845 *** (0.342) MD -2.572 *** (0.506) -0.488 ** (0.234) -0.393(0.320) -0.488 ** (0.235) -0.486 ** (0.235) -0.487 ** (0.235) -0.485 ** (0.235) -0.368(0.320) Success 0.204 * (0.111) 0.447 *** (0.142) 0.136(0.153) 0.454 *** (0.139) 0.454 *** (0.138) 0.158(0.147) 0.160(0.147) 0.161(0.147) Group -0.221(0.380) -0.235(0.378) -0.332(0.215) -0.440(0.383) -0.454(0.386) -0.327(0.418) -0.390(0.417) -0.521(0.423) -0.517(0.423) -0.579(0.425) -0.576(0.425) -0.594(0.427) Effort:group 0.670(0.610) 0.585(0.515) 0.598(0.517) 0.506(0.589) 0.496(0.589) 0.446(0.499) 0.434(0.500) 0.450(0.501) MD:group 0.771(0.747) -0.201(0.466) -0.247(0.467) Success:group 0.454 *** (0.164) 0.733 *** (0.223) 0.710 *** (0.215) 0.705 *** (0.215) 0.704 *** (0.215) Constant -0.619 ** (0.257) -1.115 *** (0.255) -0.336 ** (0.145) -0.706 *** (0.259) -0.699 *** (0.260) -0.618 ** (0.282) -0.602 ** (0.281) -0.718 ** (0.286) -0.718 ** (0.286) -0.703 ** (0.286) -0.704 ** (0.286) -0.695 ** (0.288) Observations 4,450 4,450 4,450 4,450 4,450 4,450 4,450 4,450 4,450 4,450 4,450 4,450 Log Likelihood -1,735.419 -2,364.251 -2,813.537 -1,700.518 -1,700.425 -1,649.644 -1,645.470 -1,623.729 -1,623.345 -1,619.637 -1,619.272 -1,619.133 AIC 3,482.838 4,740.501 5,639.074 3,417.036 3,418.849 3,315.287 3,308.941 3,265.459 3,266.690 3,259.274 3,260.545 3,262.265 BIC 3,521.242 4,778.905 5,677.478 3,468.241 3,476.455 3,366.492 3,366.547 3,323.065 3,330.697 3,323.281 3,330.952 3,339.073 Note: * p<0.05, ** p<0.01 *** p<0.001 Table 4.3. Mixed effect logistic regression model fits for Stroke group 105 Table 4.4. Mixed effect logistic regression model fits for the Control group. Note: *p<0.05, **p<0.01 ***p<0.001 Mixed-effects logistic-regression results Dependent variable: Right arm choice (1) (2) (3) (4) (5) Effort -4.896 *** (0.854) -4.582 *** (0.901) -4.893 *** (0.853) -4.563 *** (0.903) MD -0.948 *** (0.254) -0.963 *** (0.254) Success 0.022(0.097) -0.003(0.132) Constant 0.294 ** (0.117) 1.124 *** (0.406) 1.282 *** (0.433) 1.122 *** (0.404) 1.277 *** (0.430) Observations 1,372 1,372 1,372 1,372 1,372 Log Likelihood -929.786 -344.157 -326.704 -344.132 -326.506 AIC 1,863.572 696.314 665.408 700.263 669.012 BIC 1,874.020 717.210 696.752 731.608 710.804 106 4.4. Discussion In this study, we developed a novel computational model to understand the factors influencing arm choice in individuals with stroke and age-matched nondisabled participants. We found that effort and movement duration together influenced arm choice in both the participants with stroke and age-matched nondisabled participants. In addition, success influenced choice in the participants in the LH group, but not in participants in the RH group nor in the participants in the Control group. Arm choice patterns differed among the participants in the RH, the LH, and the Control groups. The participants in the RH group showed a decrease in choice of the more- affected right arm compared to the right arm choice in the Control group. However, their arm choice pattern was consistent across conditions (Figure 4.2 and Figure 4.3). In contrast, the participants in the LH group chose their more-affected left arm similar to the left arm choice in the participants in the Control group under no time constraint, but they showed a significant decrease in the more-affected left arm choice in the fast condition. The more- affected arm choices in the RH and the LH groups did not differ in either the no time constraint or the medium condition. However, in the fast condition, the participants in the RH group used their more-affected right arm more than the participants in LH group used the more-affected left arm. These results suggest that when choosing the more-affected arm, the participants in the LH group are affected by condition, which links to success rates, whereas the participants in the RH group are not. 107 The effect of success on arm choice has been studied for both nondisabled people and individuals with stroke. Stoloff et al. (2011) found that nondisabled young adults were less likely to use their right arm when the success rate for the right arm decreased through a reduction of the target size (Stoloff et al., 2011). Similarly, Schweighofer et al. (2015) found that nondisabled young adults showed an increase in right arm use when the task required movements with high accuracy during rapid reaching tasks. The effect of success on arm choice has also been studied for individuals with stroke. Emily (2011) found that individuals with stroke used their more-affected arm less when the tasks became more challenging from simple tasks that required simple reaching movements only to complex tasks that required both reaching and grasping movements (Emily, 2011). Although the author did not directly modulate task success by changing the experimental environment, task complexity (simple reaching vs. reaching and grasping) might indirectly link to the task success, thus, leading to a decrease in the more-affected arm use. In our experiment, interestingly, the participants in the RH group continuously used their more-affected arm in the fast condition, although the success rate of the more-affected arm in the fast condition was statistically significantly lower than the success rate of the less-affected arm (Table 4.2). In contrast, the participants in the LH group showed decreased use of the more- affected arm in the fast condition in which the success rate of the more-affected arm was significantly low. Possible explanations for consistent arm choice in the participants in the RH group are: 1) preference for the dominant-right arm was persistent even after stroke; 2) the participants in the RH group were less sensitive to the failure than were the participants in the LH group; or 3) our reaching task was simple so they did not experience the actual adverse failure, such as spilling hot coffee in the real world. These possibilities, however, are needed to be rigorously investigated throughout experiments. 108 Our results also showed that arm choice is influenced by movement duration for both individuals with stroke and the age-matched nondisabled. For equal expected success rates between the more- and the less-affected arms (or right and left arm for the Control group) as in the no time constraint, use of the arm with shorter expected movement duration is reasonable, because it relatively increases the cost less than using the arm with longer expected movement duration (Jordan and Wolpert, 1999). Therefore, our participants used the arm with short movement duration. This result can explain arm choice behaviors in which the right arm is chosen for the targets on the right side while the left arm is chosen for the targets on the left side. In addition, we found that our participants with stroke were able to move fast when instructed to do so. This result is consistent with the findings of a previous study showing that people with mild to moderate hemiparesis post-stroke were able to perform upper extremity reach–grasp–lift tasks substantially faster than their preferred movement speed (DeJong, Schaefer and Lang, 2012). Although our instruction was always the same, “move as rapidly and accurately as you can” across the conditions, the movement durations in the no time condition were significantly lower than those in the fast condition. We are assuming that, similar to others (Rigoux and Guigon, 2012; Shadmehr et al., 2010), movement duration in the no time constraint condition is determined by the minimization of the total cost, comprised of expected movement effort, duration, and success. For the case that success rates are 100% for both arms, the cost first decreases for slower movement due to lower effort and then increases due to the cost of time. The chosen movement duration is the duration that minimizes the cost. 109 Our result showed that participants take into account the biomechanical properties of the arm in decision tasks involving the arms. This result is in line with that of a previous study showing that nondisabled participants preferred to select a target, which required low biomechanical effort (Cos, Belanger and Cisek, 2011). When participants were asked to choose between two targets, they selected a target, aligned along the small axis of the arm’s inertia at the hand, thus requiring a low level of effort (Cos, Belanger and Cisek, 2011). During our experiment, we found that the effect of biomechanical properties of the arm on arm choice was persistent even after stroke. The participants in both the RH and the LH groups selected their more-affected arms when the more-affected arm took lower effort than the less-affected arm. This study has several limitations and leaves a number of open questions that need to be addressed in future work. First, we estimated effort using two-planar linked-arm (shoulder and elbow) with 2 degrees of freedom. However, our reaching task allowed more than 2 degrees of freedom. Specifically, participants in our experiments could rotate their shoulders internally or externally, which our 2 linked-arm model could not take into account. Nevertheless, effort estimated with 2 linked-arm successfully predicted actual arm choice. The model with effort alone predicted 86% of actual arm choice and the model with effort and movement duration predicted 90% of actual arm choice in the participants in the Control group. Experiments on the horizontal plane or inverse dynamics model with 3 degrees of freedom are needed to estimate effort precisely. Second, we estimated a biomechanical effort which reflects how much participants generate joint torques to move their arms, but this biomechanical effort cannot take into account abnormal muscle co- contractions commonly shown in patients with stroke (Dewald et al., 1995). In future work, 110 measurement of the actual metabolic effort including muscle co-contractions using expired gas and EMG systems (Huang and Ahmed, 2014) with measurement of the mental effort reflecting how people feel and perceive will allow us to estimate effort more accurately for patients with stroke. Third, in the constraint use sessions, participants reached each target in each condition only one time. Thus, movement duration and effort were measured and estimated from a single data point. Although this was done to prevent motor fatigue caused by excessive repetitive movements in individuals with stroke, repetitive measurements across multiple testing days would be required to obtain accurate movement duration and effort data. Our findings in this study may be important to establish rehabilitation strategies for patients with unilateral stroke. Individuals with right hemiparesis continue to use their more-affected arm as a habit when movement durations for the more-affected are short, whereas individuals with left hemiparesis flexibly use their more-affected arm depending on the task environment. Thus, we suggest that rehabilitation protocols should be differently targeted to individuals with RH and LH. For individuals with RH, the aim should be to decrease the more-affected arms’ movement duration and effort. Individuals with stroke can move fast when instructed to move fast, and movement quality is even better at a fast speed than that they move at a preferred speed (DeJong, Schaefer and Lang, 2012). Furthermore, individuals with stroke showed improvements in movement duration and smoothness after a two-day intensive reaching training, with the enhancements maintained up to 1 month (Park et al., 2015). Intervention emphasizing fast movement and re-training muscle co-activation patterns of shoulder and elbow to reduce abnormal elbow/shoulder joint torque coupling (Ellis et al., 111 2005) will be beneficial for the more-affected arm use. If individuals with RH can move their more-affected arm rapidly (short movement duration) and easily (less effortful) after treatments so that the differences between the more- and the less-affected arms in movement duration and effort decrease, they will be more likely to choose their more- affected arm in daily life. In contrast, for individuals with LH, additional treatment regimens will be needed to de-sensitize task success when the more-affected arm is used. As our results showed, individuals with LH are more sensitive to task success than are individuals with RH, they therefore need to be aware that it is more important to use their more-affected arm regardless of the task success. Task oriented-repetitive movements and accompanying patient education to encourage use of their more-affected arm (Winstein et al., 2016) will be helpful for habitual use of the more-affected arm in individuals with LH. Bimanual training will also be beneficial by preventing them from using their less-affected dominant right arm alone (Whitall et al., 2000). 112 Chapter 5. Summary and future study This dissertation work focuses on identifying factors differentially influence arm choice between patients with right hemiparesis and patients with left hemiparesis, and on developing an objective, reliable, and repetitive tool to measure use, non-use and performance of upper extremity in patients with stroke. Non-use is a discrepancy between what patients can do and what they actually do with their more-affected arms and non-use is often shown in individuals post-stroke. A novel, simple, objective, reliable, and valid instrument, the Bilateral Arm Reaching Test (BART) was developed to quantify this non-use (Chapter 2). The Actual Amount of Use Test (AAUT) was also assessed to validate the BART. For the BART and the AAUT, participants had two sessions: a spontaneous use session, in which participants were free to use either right or left arms, and a constraint use session, in which participants were asked to use their more-affected arm only. During the BART, participants were instructed to reach targets in 1.2 sec. Non-use was quantified by subtracting arm use in the spontaneous use session from the arm use in the constraint use session. Both the BART and the AAUT showed that all participants with stroke in the study did exhibit some degree of nonuse overall, although the range of nonuse observed was large. In addition, non-use measured with the BART had excellent test-retest reliability and good external validity with non-use measured with the AAUT, suggesting that the BART reliably quantified non- use of the more-affected arm in individuals post-stroke. 113 In Chapter 3, the BART was modified to assess use and performance of upper extremities in individuals post-stroke. The limitation of the BART was that the 1.2 sec time limit was too long for some patients so that they used their more-affected arm in much the same way a non-disabled person would, or too short for some patients so that they never used their more-affected arm. Therefore, additional movement duration constraints were added to develop the time-based BART system. Time-based BART included movement duration constraints shorter than 1.2 sec (fast condition with 0.5 sec time constraint), which aimed to distinguish participants with mild, moderate, and severe impairments, and also included movement duration constraints longer than 1.2 sec (no time constraint), which aimed to prevent patients with stroke from exhibiting ‘zero-use’ of the more-affected arm. Performance of upper extremity measured in the constraint use session, especially in the fast condition in the time-based BART, showed a strong correlation with performance measured with the Wolf Motor Function Test-time score. In addition, use of the upper extremity measured in the spontaneous use session, especially in the fast condition in the time-based BART, showed a strong correlation with use measured in the AAUT quality of movement score. Furthermore, both use and performance assessed in the fast condition of the time-based BART had excellent test-retest reliability; thus, the time-based BART system was sufficiently reliable to be an alternative to clinical tests to objectively and repeatedly measure use and performance in upper extremities in patients with stroke. However, the pointing task in the time-based BART mimics only one aspect of upper extremity use (e.g. reaching), but does not include other actions that might be part of daily use, such as stabilizing, supporting, grasping, tapping, etc. In future work, improved systems to automatically assess arm and hand use could, for instance, present 114 tools that allow for grasping at different spatial locations. A task-based rehabilitation robot could be modified for this purpose. In addition, the hardware (mini-bird magnetic sensors and projector mounted above table) and software (Matlab for data analysis) make the time based-BART a research tool only. Cheaper versions will need to be developed for use in the clinic. In Chapter 4, arm choice behavior in patients with right hemiparesis (RH) and patients with left hemiparesis (LH) was explored. The participants in RH group used their more-affected right-dominant arm more than the participants in LH group used their more- affected left-nondominant arm. Movement duration, success, and effort for both more- affected and less-affected arm were measured, calculated, and estimated in the constraint use session in the time-based BART system. Arm choice was measured in the spontaneous use session. A computational model of arm choice was developed to explain arm choice behavior using the difference between more-affected and less-affected arm in expected movement duration, expected effort, and expected success. The computational model revealed that the participants in the RH group took effort and movement duration into account when choosing whether to use the affected arm, while the participants in the LH group additionally took success into account. To be specific, the participants in the LH group did not use their more-affected arm if they were not successful at using it, whereas the participants in the RH group continued to use the more-affected arm regardless of their success or failure when using the more-affected arm. 115 Effort, which we estimated, however, is a biomechanical effort, which reflects how much participants generate joint torques to move their arms. This biomechanical effort cannot capture isometric movements, which require muscle contractions without actual movement. In addition, the biomechanical effort cannot take account into abnormal muscle co-contractions commonly shown in patients with stroke. In future work, measurement of actual metabolic effort involving abnormal muscle co-contractions via expired gas analysis, and mental effort reflecting how people feel and perceive together will allow us to estimate effort more accurately for patients with stroke. Overall, this dissertation emphasizes important aspects of spontaneous use of the more-affected arm for stroke rehabilitation by developing a new assessment tool and by establishing a computational model. One of the greatest advantages in computational arm choice modeling is that it can make a prediction for actual choice and identify factors influencing actual choice. In practical and clinical applications, our results can be used as a basis when clinicians establish rehabilitation strategies. Patients with left hemiparesis and patients with right hemiparesis need to have different treatment goals and schedules to improve spontaneous use of their more-affected arm. In addition, time-based BART can be useful in assessing and treating upper extremities in individuals with stroke. 116 References Andrews, K., & Stewart, J. (1979). Stroke recovery: he can but does he? Rheumatol Rehabil, 18(1), 43-48. Barker, R. N., & Brauer, S. G. (2005). Upper limb recovery after stroke: the stroke survivors' perspective. Disabil Rehabil, 27(20), 1213-1223. Brown, E. (2011). Hand preference after stroke: The development and initial evaluation of a new performance-based measure. Thesis, Department of Kinesiology, University of Waterloo. Chen, S., Lewthwaite, R., Schweighofer, N., & Winstein, C. J. (2013). Discriminant validity of a new measure of self-efficacy for reaching movements after stroke-induced hemiparesis. J Hand Ther, 26(2), 116-122; quiz 123. Chen, S., Wolf, S. L., Zhang, Q., Thompson, P. A., & Winstein, C. J. (2012). Minimal Detectable Change of the Actual Amount of Use Test and the Motor Activity Log: The EXCITE Trial. Neurorehabil Neural Repair, 26(5), 507-514. Choi, Y., Gordon, J., Kim, D., & Schweighofer, N. (2009). An Adaptive Automated Robotic Task-Practice System for Rehabilitation of Arm Functions After Stroke. IEEE Transactions on Robotics, 25(3), 556-568. Choi, Y., Gordon, J., Park, H., & Schweighofer, N. (2011). Feasibility of the adaptive and automatic presentation of tasks (ADAPT) system for rehabilitation of upper extremity function post-stroke. J Neuroeng Rehabil, 8, 42. Cirstea, M. C., & Levin, M. F. (2000). Compensatory strategies for reaching in stroke. Brain, 123 ( Pt 5), 940-953. Cirstea, M. C., Ptito, A., & Levin, M. F. (2003). Arm reaching improvements with short-term practice depend on the severity of the motor deficit in stroke. Exp Brain Res, 152(4), 476-488. Cos, I., Belanger, N., & Cisek, P. (2011). The influence of predicted arm biomechanics on decision making. J Neurophysiol, 105(6), 3022-3033. 117 Dejong, S. L., & Lang, C. E. (2012). Comparison of unilateral versus bilateral upper extremity task performance after stroke. Top Stroke Rehabil, 19(4), 294-305. DeJong, S. L., Schaefer, S. Y., & Lang, C. E. (2012). Need for speed: better movement quality during faster task performance after stroke. Neurorehabil Neural Repair, 26(4), 362-373. Dewald, J. P., Pope, P. S., Given, J. D., Buchanan, T. S., & Rymer, W. Z. (1995). Abnormal muscle coactivation patterns during isometric torque generation at the elbow and shoulder in hemiparetic subjects. Brain, 118 ( Pt 2), 495- 510. Dobkin, B. H. (2005). Clinical practice. Rehabilitation after stroke. N Engl J Med, 352(16), 1677-1684. Duff, S. V., He, J., Nelsen, M. A., Lane, C. J., Rowe, V. T., Wolf, S. L., . . . Winstein, C. J. (2015). Interrater reliability of the Wolf Motor Function Test-Functional Ability Scale: why it matters. Neurorehabil Neural Repair, 29(5), 436-443. Duncan, P., Studenski, S., Richards, L., Gollub, S., Lai, S. M., Reker, D., . . . Johnson, D. (2003). Randomized clinical trial of therapeutic exercise in subacute stroke. Stroke, 34(9), 2173-2180. Ellis, M. D., Holubar, B. G., Acosta, A. M., Beer, R. F., & Dewald, J. P. (2005). Modifiability of abnormal isometric elbow and shoulder joint torque coupling after stroke. Muscle Nerve, 32(2), 170-178. Emily, B. (2011). Hand preference after stroke. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). "Mini-mental state". A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res, 12(3), 189-198. Fugl-Meyer, A. R., Jaasko, L., Leyman, I., Olsson, S., & Steglind, S. (1975). The post- stroke hemiplegic patient. 1. a method for evaluation of physical performance. Scand J Rehabil Med, 7(1), 13-31. Fullerton, K. J., McSherry, D., & Stout, R. W. (1986). Albert's test: a neglected test of perceptual neglect. Lancet, 1(8478), 430-432. Guigon, E., Baraduc, P., & Desmurget, M. (2007). Computational motor control: redundancy and invariance. J Neurophysiol, 97(1), 331-347. Haaland, K. Y., Mutha, P. K., Rinehart, J. K., Daniels, M., Cushnyr, B., & Adair, J. C. (2012). Relationship between arm usage and instrumental activities of daily living after unilateral stroke. Arch Phys Med Rehabil, 93(11), 1957-1962. 118 Habagishi, C., Kasuga, S., Otaka, Y., Liu, M., & Ushiba, J. (2014). Different strategy of hand choice after learning of constant and incremental dynamical perturbation in arm reaching. Front Hum Neurosci, 8, 92. Han, C. E., Arbib, M. A., & Schweighofer, N. (2008). Stroke rehabilitation reaches a threshold. PLoS Comput Biol, 4(8), e1000133. Han, C. E., Kim, S., Chen, S., Lai, Y. H., Lee, J. Y., Osu, R., . . . Schweighofer, N. (2013). Quantifying arm nonuse in individuals poststroke. Neurorehabil Neural Repair, 27(5), 439-447. Hidaka, Y., Han, C. E., Wolf, S. L., Winstein, C. J., & Schweighofer, N. (2012). Use it and improve it or lose it: interactions between arm function and use in humans post-stroke. PLoS Comput Biol, 8(2), e1002343. Hoff, B. R. (1992). A computational description of the organization of human reaching and prehension. PhD Thesis. University of Southern California. Huang, H. J., & Ahmed, A. A. (2014). Older adults learn less, but still reduce metabolic cost, during motor adaptation. J Neurophysiol, 111(1), 135-144. Johnson, M., Paranjape, R., Strachota, E., Tchekanov, G., & McGuire, J. (2011). Quantifying learned non-use after stroke using unilateral and bilateral steering tasks. IEEE Int Conf Rehabil Robot, 2011, 5975457. Jordan, M. I., & Wolpert, D. M. (1999). Computational motor control. The Cognitive Neurosciences. Kopp, B., Kunkel, A., Flor, H., Platz, T., Rose, U., Mauritz, K. H., . . . Taub, E. (1997). The Arm Motor Ability Test: reliability, validity, and sensitivity to change of an instrument for assessing disabilities in activities of daily living. Arch Phys Med Rehabil, 78(6), 615-620. Lang, C. E., Bland, M. D., Bailey, R. R., Schaefer, S. Y., & Birkenmeier, R. L. (2013). Assessment of upper extremity impairment, function, and activity after stroke: foundations for clinical decision making. J Hand Ther, 26(2), 104- 114;quiz 115. Mathiowetz, V., Volland, G., Kashman, N., & Weber, K. (1985). Adult norms for the Box and Block Test of manual dexterity. Am J Occup Ther, 39(6), 386-391. Mayo, N. E., Wood-Dauphinee, S., Cote, R., Durcan, L., & Carlton, J. (2002). Activity, participation, and quality of life 6 months poststroke. Arch Phys Med Rehabil, 83(8), 1035-1042. 119 Michaelsen, S. M., & Levin, M. F. (2004). Short-term effects of practice with trunk restraint on reaching movements in patients with chronic stroke: a controlled trial. Stroke, 35(8), 1914-1919. Michielsen, M. E., Selles, R. W., Stam, H. J., Ribbers, G. M., & Bussmann, J. B. (2012). Quantifying nonuse in chronic stroke patients: a study into paretic, nonparetic, and bimanual upper-limb use in daily life. Arch Phys Med Rehabil, 93(11), 1975-1981. Morris, D. M., Uswatte, G., Crago, J. E., Cook, E. W., 3rd, & Taub, E. (2001). The reliability of the wolf motor function test for assessing upper extremity function after stroke. Arch Phys Med Rehabil, 82(6), 750-755. Oldfield, R. C. (1971). The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia, 9(1), 97-113. Park, H., Kim, S., Winstein, C. J., Gordon, J., & Schweighofer, N. (2015). Short-Duration and Intensive Training Improves Long-Term Reaching Performance in Individuals With Chronic Stroke. Neurorehabil Neural Repair. Rigoux, L., & Guigon, E. (2012). A model of reward- and effort-based optimal decision making and motor control. PLoS Comput Biol, 8(10), e1002716. Rinehart, J. K., Singleton, R. D., Adair, J. C., Sadek, J. R., & Haaland, K. Y. (2009). Arm use after left or right hemiparesis is influenced by hand preference. Stroke, 40(2), 545-550. Roby-Brami, A., Feydy, A., Combeaud, M., Biryukova, E. V., Bussel, B., & Levin, M. F. (2003). Motor compensation and recovery for reaching in stroke patients. Acta Neurol Scand, 107(5), 369-381. Schenker, M., Burstedt, M. K., Wiberg, M., & Johansson, R. S. (2006). Precision grip function after hand replantation and digital nerve injury. J Plast Reconstr Aesthet Surg, 59(7), 706-716. Schweighofer, N., Han, C. E., Wolf, S. L., Arbib, M. A., & Winstein, C. J. (2009). A functional threshold for long-term use of hand and arm function can be determined: predictions from a computational model and supporting data from the Extremity Constraint-Induced Therapy Evaluation (EXCITE) Trial. Phys Ther, 89(12), 1327-1336. Schweighofer, N., Xiao, Y., Kim, S., Yoshioka, T., Gordon, J., & Osu, R. (2015). Effort, success, and nonuse determine arm choice. J Neurophysiol, 114(1), 551-559. 120 Shadmehr, R., Orban de Xivry, J. J., Xu-Wilson, M., & Shih, T. Y. (2010). Temporal discounting of reward and the cost of time in motor control. J Neurosci, 30(31), 10507-10516. Spaulding, S. J., McPherson, J. J., Strachota, E., Kuphal, M., & Ramponi, M. (1988). Jebsen Hand Function Test: performance of the uninvolved hand in hemiplegia and of right-handed, right and left hemiplegic persons. Arch Phys Med Rehabil, 69(6), 419-422. Steenbergen, B., Van Thiel, E., Hulstijn, W., & Meulenbroek, R. G. J. (2000). The coordination of reaching and grasping in spastic hemiparesis. Human Movement Science, 19, 75 –105. Sterr, A., Elbert, T., Berthold, I., Kolbel, S., Rockstroh, B., & Taub, E. (2002). Longer versus shorter daily constraint-induced movement therapy of chronic hemiparesis: an exploratory study. Arch Phys Med Rehabil, 83(10), 1374- 1377. Sterr, A., Freivogel, S., & Schmalohr, D. (2002). Neurobehavioral aspects of recovery: assessment of the learned nonuse phenomenon in hemiparetic adolescents. Arch Phys Med Rehabil, 83(12), 1726-1731. Stewart, J. C., & Cramer, S. C. (2013). Patient-reported measures provide unique insights into motor function after stroke. Stroke, 44(4), 1111-1116. Stoloff, R. H., Taylor, J. A., Xu, J., Ridderikhoff, A., & Ivry, R. B. (2011). Effect of reinforcement history on hand choice in an unconstrained reaching task. Front Neurosci, 5, 41. Sunderland, A., & Tuke, A. (2005). Neuroplasticity, learning and recovery after stroke: a critical evaluation of constraint-induced therapy. Neuropsychol Rehabil, 15(2), 81-96. Taub, E., Crago, J., & Uswatt, G. (1998). Constraint-Induced Movement Therapy: A New Approach to Treatment in Physical Rehabilitation. Rehabilitation Psychology, 43(2), 152-170. Taub, E., Crago, J. E., Burgio, L. D., Groomes, T. E., Cook, E. W., 3rd, DeLuca, S. C., & Miller, N. E. (1994). An operant approach to rehabilitation medicine: overcoming learned nonuse by shaping. J Exp Anal Behav, 61(2), 281-293. Taub, E., Miller, N. E., Novack, T. A., Cook, E. W., 3rd, Fleming, W. C., Nepomuceno, C. S., . . . Crago, J. E. (1993). Technique to improve chronic motor deficit after stroke. Arch Phys Med Rehabil, 74(4), 347-354. 121 Taub, E., & Uswatte, G. (2003). Constraint-induced movement therapy: bridging from the primate laboratory to the stroke rehabilitation laboratory. J Rehabil Med(41 Suppl), 34-40. Taub, E., & Uswatte, G. (2006). Constraint-Induced Movement therapy: Answers and questions after two decades of research. NeuroRehabilitation, 21(2), 93-95. Taub, E., Uswatte, G., & Elbert, T. (2002). New treatments in neurorehabilitation founded on basic research. Nat Rev Neurosci, 3(3), 228-236. Taub, E., Uswatte, G., Mark, V. W., & Morris, D. M. (2006). The learned nonuse phenomenon: implications for rehabilitation. Eura Medicophys, 42(3), 241- 256. Taub, E., Uswatte, G., & Pidikiti, R. (1999). Constraint-Induced Movement Therapy: a new family of techniques with broad application to physical rehabilitation--a clinical review. J Rehabil Res Dev, 36(3), 237-251. Todorov, E., & Jordan, M. I. (2002). Optimal feedback control as a theory of motor coordination. Nat Neurosci, 5(11), 1226-1235. Uswatte, G., Foo, W. L., Olmstead, H., Lopez, K., Holand, A., & Simms, L. B. (2005). Ambulatory monitoring of arm movement using accelerometry: An objective measure of upper-extremity rehabilitation in persons with chronic stroke. Arch Phys Med Rehabil, 86(7), 1498-1501. Uswatte, G., Miltner, W. H., Foo, B., Varma, M., Moran, S., & Taub, E. (2000). Objective measurement of functional upper-extremity movement using accelerometer recordings transformed with a threshold filter. Stroke, 31(3), 662-667. Uswatte, G., & Taub, E. (2005). Implications of the learned nonuse formulation for measuring rehabilitation outcomes: lessons from constraintinduced movement therapy. Rehabil Psychol 2005, 50, 34-42. Uswatte, G., Taub, E., Morris, D., Barman, J., & Crago, J. (2006). Contribution of the shaping and restraint components of Constraint-Induced Movement therapy to Treatment Outcome. NeuroRehabilitation, 21(2), 147-156. Uswatte, G., Taub, E., Morris, D., Light, K., & Thompson, P. A. (2006). The Motor Activity Log-28 - Assessing daily use of the hemiparetic arm after stroke. Neurology, 67(7), 1189-1194. Uswatte, G., Taub, E., Morris, D., Vignolo, M., & McCulloch, K. (2005). Reliability and validity of the upper-extremity Motor Activity Log-14 for measuring real- world arm use. Stroke, 36(11), 2493-2496. 122 Uswatte, G., Taub, E., Morris, D., Vignolo, M., & McCulloch, K. (2005). Reliability and validity of the upper-extremity motor activity Log-14 for measuring real- world arm use. Stroke, 36(11), 2493-2496. van Beers, R. J., Haggard, P., & Wolpert, D. M. (2004). The role of execution noise in movement variability. J Neurophysiol, 91(2), 1050-1063. van der Lee, J. H., Beckerman, H., Knol, D. L., de Vet, H. C., & Bouter, L. M. (2004). Clinimetric properties of the motor activity log for the assessment of arm use in hemiparetic patients. Stroke, 35(6), 1410-1414. van Kordelaar, J., van Wegen, E. E., Nijland, R. H., de Groot, J. H., Meskers, C. G., Harlaar, J., & Kwakkel, G. (2012). Assessing longitudinal change in coordination of the paretic upper limb using on-site 3-dimensional kinematic measurements. Phys Ther, 92(1), 142-151. Wade, D. T. (1992). Measurement in neurological rehabilitation. Curr Opin Neurol Neurosurg, 5(5), 682-686. Whitall, J., McCombe Waller, S., Silver, K. H., & Macko, R. F. (2000). Repetitive bilateral arm training with rhythmic auditory cueing improves motor function in chronic hemiparetic stroke. Stroke, 31(10), 2390-2395. Winstein, C. J., Miller, J. P., Blanton, S., Taub, E., Uswatte, G., Morris, D., . . . Wolf, S. (2003). Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke. Neurorehabil Neural Repair, 17(3), 137-152. Winstein, C. J., Rose, D. K., Tan, S. M., Lewthwaite, R., Chui, H. C., & Azen, S. P. (2004). A randomized controlled comparison of upper-extremity rehabilitation strategies in acute stroke: A pilot study of immediate and long-term outcomes. Arch Phys Med Rehabil, 85(4), 620-628. Winstein, C. J., Wolf, S. L., Dromerick, A. W., Lane, C. J., Nelsen, M. A., Lewthwaite, R., . . . Interdisciplinary Comprehensive Arm Rehabilitation Evaluation Investigative, T. (2016). Effect of a Task-Oriented Rehabilitation Program on Upper Extremity Recovery Following Motor Stroke: The ICARE Randomized Clinical Trial. JAMA, 315(6), 571-581. Wolf, S. L., Catlin, P. A., Ellis, M., Archer, A. L., Morgan, B., & Piacentino, A. (2001). Assessing Wolf motor function test as outcome measure for research in patients after stroke. Stroke, 32(7), 1635-1639. 123 Wolf, S. L., Lecraw, D. E., Barton, L. A., & Jann, B. B. (1989). Forced use of hemiplegic upper extremities to reverse the effect of learned nonuse among chronic stroke and head-injured patients. Exp Neurol, 104(2), 125-132. Wolf, S. L., McJunkin, J. P., Swanson, M. L., & Weiss, P. S. (2006). Pilot normative database for the Wolf Motor Function Test. Arch Phys Med Rehabil, 87(3), 443-445. Wolf, S. L., Thompson, P. A., Winstein, C. J., Miller, J. P., Blanton, S. R., Nichols-Larsen, D. S., . . . Sawaki, L. (2010). The EXCITE stroke trial: comparing early and delayed constraint-induced movement therapy. Stroke, 41(10), 2309-2315. Wolf, S. L., Winstein, C. J., Miller, J. P., Taub, E., Uswatte, G., Morris, D., . . . Investigators, E. (2006). Effect of constraint-induced movement therapy on upper extremity function 3 to 9 months after stroke: the EXCITE randomized clinical trial. JAMA, 296(17), 2095-2104.

Abstract (if available)

Abstract Most individuals with upper extremity disability resulting from a stroke face difficulties effectively using their more-affected arm and hand in daily activities. Spontaneous use of the more-affected arm in daily life is a meaningful indicator of motor recovery for stroke rehabilitation. However, the spontaneous use of the more-affected arm varies among patients. Some patients with stroke continue to use the more-affected arm after treatment, while others avoid using their more-affected arm in the real world even though they have a residual capacity to use it. Therefore, it is important to understand the mechanisms underlying arm choice in individuals with post-stroke as a basis for effective rehabilitation strategies. ❧ This dissertation has two aims. The first is to develop a simple, objective, and replicable tool to assess movements in upper extremities in individuals with stroke. The Bilateral Arm Reaching Test (BART) system has been proposed to assess spontaneous use, non-use, and performance of upper extremities. The BART shows excellent test-retest reliability and strong external validity with clinical assessments for non-use of the more-affected arm in individuals with stroke. In addition, we modified the BART system (time-based BART) by adding additional movement duration constraints and evaluated performance and use of the more-affected arm. The fast condition in the time-based BART system shows excellent test-retest reliability and strong validity with Actual Amount of Use Test (AAUT) and Wolf Motor Function Test (WMFT) for spontaneous use and performance assessments. ❧ The second aim of this dissertation is to understand the mechanisms that determine arm choice for both patients with stroke and age-matched nondisabled people by clarifying individual factors related to arm choice during reaching tasks. Lesion side (Right hemiparesis vs. Left hemiparesis) leads to different arm choice patterns and mechanisms. Computational modeling of arm choice revealed that patients with right-hemiparesis took effort and movement duration into account when choosing the more-affected arm, while patients with left-hemiparesis additionally took success into account. The results of this dissertation may ultimately further the development of rehabilitation strategies customized for each individual with stroke.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Learning reaching skills in non-disabled and post-stroke individuals

PDF

Hemisphere-specific deficits in the control of bimanual movements after stroke

PDF

Deficits and rehabilitation of upper-extremity multi-joint movements in individuals with chronic stroke

PDF

Relationship between brain structure and motor behavior in chronic stroke survivors

PDF

The effects of fast walking, biofeedback, and cognitive impairment on post-stroke gait

PDF

Using ecological momentary assessment to study the impact of social-cognitive factors on paretic hand use after stroke

PDF

Planning of unconstrained reach actions after unilateral sensorimotor stroke

PDF

Design of adaptive automated robotic task presentation system for stroke rehabilitation

PDF

Hemispheric specialization of reach-to-grasp actions

PDF

Experimental and computational explorations of different forms of plasticity in motor learning and stroke recovery

PDF

Engagement of the action observation network through functional magnetic resonance imaging with implications for stroke rehabilitation

PDF

Computational model of stroke therapy and long term recovery

PDF

Impact of enhanced efficacy expectation on motor learning in individuals with Parkinson’s disease

PDF

The brain and behavior of motor learning: the what, how and where

PDF

Development and implementation of a modular muscle-computer interface for personalized motor rehabilitation after stroke

PDF

Quantification of lower extremity dynamic capability: implications for anterior cruciate ligament injury and change of direction ability

PDF

Trunk control during dynamic balance: effects of cognitive dual-task interference and a history of recurrent low back pain

PDF

Modeling motor memory to enhance multiple task learning

PDF

Contextual interference in motor skill learning: an investigation of the practice schedule effect using transcranial magnetic stimulation (TMS)

PDF

Behavioral and neurophysiological studies of hand dexterity in health and Parkinson's disease

Asset Metadata

Creator Kim, Sujin (author)

Core Title Arm choice post-stroke

School School of Dentistry

Degree Doctor of Philosophy

Degree Program Biokinesiology

Publication Date 08/02/2016

Defense Date 05/05/2016

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag arm choice,left and right hemiparesis,non-use,OAI-PMH Harvest,Performance,stroke,upper extremity,use

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Schweighofer, Nicolas (committee chair), Finley, James (committee member), Gordon, James (committee member), Monterosso, John (committee member), Winstein, Carolee (committee member)

Creator Email sujink@usc.edu,veritaspt@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-289824

Unique identifier UC11281190

Identifier etd-KimSujin-4704.pdf (filename),usctheses-c40-289824 (legacy record id)

Legacy Identifier etd-KimSujin-4704.pdf

Dmrecord 289824

Document Type Dissertation

Format application/pdf (imt)

Rights Kim, Sujin

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA