Enriched Environments for Upper Limb Stroke Rehabilitation

1. Scientific Background

Stroke is the third leading cause of death in Western countries and upper limb (UL) sensorimotor dysfunction contributes significantly to the incidence of physical disabilities and handicaps (Olsen 1990). One explanation for poor arm recovery is the focus on task accomplishment rather than performance quality. This may reinforce alterative (compensatory) movement strategies instead of encouraging the reappearance of pre-morbid movement patterns (recovery). Although the rehabilitative goal is recovery of function, whether this is achieved through true motor recovery or compensation is still under debate. Indeed, for some patients with severe impairment, compensation should be encouraged to maximize functional ability. Alternatively, for those with good prognosis, motor recovery is emphasized for several reasons. First, given appropriate training, recovery can continue well into the chronic stage of stroke (e.g., Michaelsen \& Levin 2004). Second, while compensation may assist immediate performance, it may lead to longer-term problems such as pain and contracture (Ada et al. 1994; Levin 1997). Third, permitting compensations could encourage learned nonuse limiting the capacity for subsequent motor gains (Allred et al. 2005). While performance gains have been documented after repetitive training of isolated movements (e.g., Whitall et al. 2000), few studies have addressed whether patients can use explicit information to optimize motor skill acquisition and whether true behavioural recovery occurs. Points to consider in developing optimal training programs is that learning occurs when participants are motivated, practice a variety of related tasks and are given relevant feedback (Nudo \& Friel 1999; Winstein et al. 1999). In addition, patients may not benefit from variable practice until missing motor elements are recovered (Carr \& Shepherd 1987) and motor relearning is related to physical and cognitive impairment level (Cirstea et al. 2006).

To date, most studies have used only clinical measures to evaluate functional change (e.g., Jang et al. 2003) without considering how movement is performed. This study will focus on the ability to distinguish between whether functional improvement results from an increase in compensation or from true motor recovery. This can only be done by correlating functional improvement (clinical measures) with changes in arm motor patterns through detailed movement analysis (kinematics). Since there are a large number of possible kinematic indicators of improvement, the investigators need to identify which measures are most indicative of change. This is a first step in the determination of clinically meaningful outcome measures to be used in randomized control trials of the effectiveness of interventions. Clinical salience is essential to translate knowledge from research studies to evidence-based practice (van Peppen et al. 2007).

1. Study Design: 1. Detailed Plan of the Study

Two groups of 8 stroke subjects will participate in an A-B-A design in which data from post-intervention (within one week of the completion of the intervention) and follow-up (one month after the completion of the intervention)tests will be compared to data from two pre-tests (held at beginning and end of the first week of the study just prior to the start of the intervention). Multiple clinical and kinematic motor performance outcomes will be measured to determine which ones best describe arm motor recovery. This will be done by comparing changes in outcomes before and after training to mean scores of 8 age- and gender-matched healthy volunteers (recorded in a single session). Stroke survivors will practice movement in two different training environments providing different levels of motivation and feedback: Training Environment 1 will be created in a 2D video-based VR environment that will provide high motivation, Knowledge of Results (KR) about motor outcome (i.e., speed, precision) and non-specific Knowledge of Performance (KP) feedback about trunk movement. Environment 2 will be created in a physical environment that provides only KP feedback but no KR or additional motivation.

Training Protocol: After baseline evaluations (physical \& cognitive), stroke groups will practice pointing to 6 standardized targets placed just beyond reach (12 trials per target) during an acquisition phase of 15 sessions spaced over three weeks (3 sessions/wk, 72 trials/session). This practice regimen incorporates elements necessary for optimal motor learning: 1) Varied practice: Although the movements are not new motor tasks, they should be re-acquired during stroke recovery. Our aim is not teach subjects a novel task but to identify how a movement performed sub-optimally may be improved with practice. 2) Intensive practice: The number of repetitions per session was chosen according to studies by Cirstea and Levin (2000). Thus, 72 trials per session (3 blocks of 24 trials, 5 min rest between blocks) are considered as 'intensive' practice, which is a necessary element for a successful training program. Following the acquisition phase, evaluations which will be repeated after 3mos to evaluate retention and transfer of motor skills.

1. Methods: Subjects Eight stroke subjects per group will be recruited. They will be matched for age and for initial arm motor severity (± 5 pts on the Fugl-Meyer Arm Assessment). Eight age- and sex-matched healthy subjects will be recruited as a control group.

Subject Recruitment: Patients will be recruited from Haim Sheba Hospital discharge lists using a procedure in place since 2005. Screening will be done by clinical research associates. Potential participants meeting inclusion criteria 1-3 and exclusion criteria will be sent an explanatory letter inviting them to contact study investigators. Informed consent will be obtained from each participant. Control subjects will be recruited from the community.

Training Environments: Motivation and Feedback: Training will be done in the Dept. of Occupational Therapy located at the Sheba Medical Center. Training environments will allow us to evaluate the effect of combining different degrees of motivation and feedback on motor outcome. Environment 1 is highly motivating (novel and fun) since training is presented as a game in which the learner tries to beat his own score in subsequent sessions. Targets are presented as balls in a video-based VR system (Gesture Xtreme). The patient sits in front of a 36 inch display of the 6 target scene. The patient's image is captured by a video camera and inserted into the scene displayed on the monitor. The task is to point to each of the targets which appear in a random sequence. Feedback will provide pertinent information for motor learning (KR, KP, game score). Environment 2 incorporates the same number and disposition of targets as Env. 1 but they will be presented in a physical environment on a wooden frame in front of the subject. KR and KP feedback will be provided as verbal cues by the experimenter as is done usually in the clinic. However, they will be no additional motivational information (game score).

Preliminary Outcome Measures: Clinical Before and after training in both stroke groups, blinded evaluators will measure clinical scores of arm motor impairment and function (Fugl-Meyer Arm Score, Reaching Performance Scale, Box and Blocks test, Wolf Motor Function Test, Motor Activity Log) as well as cognitive function. All tests are valid and reliable and are regularly used in clinical practice.

Kinematic testing: Test Task and Transfer Task:

The investigators will also record kinematic outcome measures characterizing arm and trunk movements during reaching (elbow extension, shoulder horizontal adduction, shoulder flexion, scapular movements, trunk displacement).

Our aim is to focus on segmental and joint kinematics during a Test and a Transfer Task consisting of reaching movements. Test Task will be recorded. The Test target will be placed in line with the patient's sternum at a distance just beyond the subject's arm length. The Test Task is similar to movements made to one of the practiced targets. To assess motor learning, the investigators will determine if elements learned in one task transfer to other similar tasks. Thus, the investigators will also assess movements made in a Transfer Task, to a target placed in front of the ipsilateral shoulder 5 centimeters (cm) higher than the topmost row of trained targets. The Transfer Task is therefore a new movement, not practiced during the training. Rigid body segmental kinematics will be recorded from sets of 4 passive reflective markers (0.5 cm diameter spheres) attached to the trunk, upper arm and forearm segments. This will enable the computation of three translational and three rotational degrees of freedom per segment. Joint kinematics will be collected from markers fixed on the sternum, acromion, elbow and wrist via exo-skeletal frames. Marker motion will be recorded with a calibrated 3 camera opto-electronic motion-capture system (ProReflex MCU-240, Qualisys) on suitable computer software (Qualisys, Göteborg, Swe). Data collection (100 Hertz, 2-5 seconds) will be triggered by movement of the hand from a central position by release of a mechanical switch. The accuracy of the measurements of each marker is within an error of <0.2 cm.

As a first step, the investigators will determine movement times, accuracy, smoothness, segment and joint rotations and the main components of multi-joint coordination. Cartesian coordinates (x,y,z) for each segment will be obtained from the segmental marker sets. Raw data from at least three markers per segment will be used after interpolation of missing markers (5th-order polynomial). For each trial, movement times and peak velocities will be determined from endpoint tangential velocity traces. Movement time is the time between the point at which the tangential velocity exceeds and remains above 5% of its maximum and then returns to and remains below this level. Movement accuracy in terms of constant extent errors will be computed as the mean distance (d) between the final endpoint position (x, y, z) and the position of the target (x0, y0, z0). Constant and variable (SD) directional errors will also be defined. Movement smoothness will be computed from the tangential velocity trace using the index of curvature (ratio of endpoint path length to length of a straight line joining the initial and final positions) and the number of peaks in the trajectory path. For segment and joint rotations, vectors will be calculated within reference coordinate systems for the upper and forearm segments.

From the resulting segment centroid motion and segment rotation data different outcome variables will be calculated. The most relevant ones will define changes in performance with respect to the endpoint motion and with respect to the whole-arm posture, particularly, the evolution of the movement over time (time derivatives) and the fragmentation of the segmental path and posture of the arm while moving towards the different targets. Additional aspects of movement will be explored in order to find the most appropriate descriptors of arm motor performance and changes after the proposed clinical interventions.

Change scores will be calculated for each subject to determine an Index of Improvement (IP) and an Index of Learning (IL). The IP is defined as the change in each variable as a ratio of post- to pre-test values. An IP of 1 indicates no change whereas negative and positive values indicate that the value decreased or increased respectively. For some variables, positive ratios will indicate improvement (endpoint velocity, endpoint smoothness, angle measures, phase amplitudes) while for others, improvement will be demonstrated by negative ratios (movement time, movement precision). The IL, used to determine if subjects retain improvements after training for each follow-up epoch, is defined as the change in the variable at retention- compared to post-test. IL values of 1 indicate that the improvement is maintained (no change), of negative value that the parameter decreased and of positive value that the parameter increased.

Statistical analysis: For our first aim, changes in outcome measures will be determined by comparing movement outcomes before and after the acquisition and retention phases by two-repeated measures mixed design ANOVA (MANOVA) and comparing raw means (Time 1, Time 2, Time 3, Time 4) and change scores (IP, IL) between groups. The investigators will determine which kinematic measures undergo the most change and thus may be most indicative of motor recovery (e.g., increase in elbow and/or shoulder movement, decrease in trunk movement) by investigating which measures are the most different between training groups. The investigators expect changes in the group training in Environment I to be greater than those in from the group training in Environment 2.