Journal of Neuroengineering and Rehabilitation Open Access Intervention to Enhance Skilled Arm and Hand Movements after Stroke: a Feasibility Study Using a New Virtual Reality System

Background: Rehabilitation programs designed to develop skill in upper extremity (UE) function after stroke require progressive practice that engage and challenge the learner. Virtual realty (VR) provides a unique environment where the presentation of stimuli can be controlled systematically for optimal challenge by adapting task difficulty as performance improves. We describe four VR tasks that were developed and tested to improve arm and hand movement skills for individuals with hemiparesis.


Background
Neurorehabilitation after stroke may include interventions designed to improve functional upper extremity (UE) skills through task-related practice. While amount of practice is an important variable for motor learning [1], variations in direction, timing and speed are needed to optimize the development of skill [2]. Virtual reality (VR) is a promising modality for the creation of favorable practice environments for neurorehabilitation [3][4][5][6][7][8].
The purpose of this pilot trial was to determine the feasibility of providing individualized, progressive practice of skilled UE arm and hand movements after stroke using VR based tasks. We developed 4 tasks that allowed control of multiple parameters for the purpose of promoting motor skill learning by varying movement direction and speed. We investigated the feasibility of implementing an intervention tailored to each individual's level of movement ability and rate of progression over 12 practice sessions. Preliminary results are reported for two participants with different motor severity.

Participants
Six individuals with hemiparesis were recruited; two with different motor severity were selected for case presentation. Potential participants were screened for inclusion: 1) stroke at least 1 month prior; 2) more than 18 years of age; 3) Mini-Mental Status Exam score ≥ 24; 4) no significant range of motion limitations in the hemiparetic UE; and 5) voluntary movement control to perform the VR tasks. Table 1 includes demographic details for Subjects 102 (severe impairment) and 103 (moderate impairment).

Virtual Reality System and Environment
All tasks were displayed using a desktop personal computer and shutter glasses (StereoGraphics) to provide a threedimensional view of stimuli. To interact with the VR environment in three of the tasks, a 6 degree-of-freedom (DOF) magnetic tracker (Flock of Birds, Ascension Technology) was attached to the participant's hand or to a held object. The fourth task, 'Pinch', was performed using two PHANToM devices (SensAble Technologies) reconfigured to work together. PHANToM 1 was a Premium 1.5/3 DOF model fit with a thimble gimbal replacing the stylus and attached to the end of the index finger. PHANToM 2 was a 6 DOF model with the stylus placed in the web space of the hand and secured to the thumb with an elastic band ( Figure 1A). VR tasks were programmed using C++ with Open GL and Ghost libraries.
Four VR 'games' developed at the University of Southern California Integrated Media Systems Center were adapted to address specific motor deficits common after stroke and to provide a challenging and engaging practice environment. 'Reaching' requires the participant to reach for static cubes and 'hit' one cube at a time in a participantselected order (Figure 2A). 'Ball Shooting' requires the participant to reach and intercept a ball shot from a wall. Both of these tasks were mapped to the individual by presenting stimuli in relation to his/her shoulder location ( Figure 2B). 'Rotation' [9][10][11] enables forearm pronation and supination movements ( Figure 3). 'Pinch' enables a precision grasp between the thumb and index finger and requires the participant to pick up a cube ( Figure 1B). Summary feedback was provided to the participant after the completion of each practice block (10 to 20 trials) in the form of trial success rate and total time.

Outcome Measures
Behavioral assessments were administered pre-, mid-, and post-training. Severity of motor deficit was determined with the UE portion of the Fugl-Meyer (FM) [12], an impairment-based measure. Functional ability was evaluated with the Functional Test of the Hemiparetic UE (FTHUE) [13] where the individual completes progressively more difficult functional tasks and the Box and Block test (B&B) [14] which requires one to grasp and move 2.5 cm blocks over a 10.8 cm tall barrier. The Stroke Impact Scale (SIS) was administered pre-and post-training to assess participation and health status [15].

Procedure
Each participant attended 12 training sessions lasting 1-2 hrs/day over 3 weeks. A physical or occupational therapist was present during each session to run diagnostic tests and chose practice blocks and task parameters with the goal to maintain a moderate level of difficulty. If necessary, the therapist provided assistance for task completion, protected joint structures, and/or promoted movement quality.

VR Task Performance
Both participants completed all 12 VR practice sessions. Subject 102 (more impaired) was unable to perform  Both visual and auditory feedback indicates successful collision of the "virtual hand" with a cube. B) Interface for practice trial configuration. Pitch angle, yaw angle, and percentage of arm length (distance from the acromion to the radial styloid with the elbow extended) were chosen for each cube within a practice block. Practice blocks were designed to address reaching ability using arm lengths ranging from 10% to 120%. A similar interface was used to develop 'Ball Shooting' practice blocks.

Outcome Measures
Physical practice in the virtual environment generalized to different behavioral changes for the two participants (Table 3). Subject 103 showed no change in impairment score (UE FM) but did show functional improvements in grasp and release (B&B, 20% improvement). FTHUE score was unchanged likely due to the ceiling effect at pre-test. Subject 102 did not change impairment level (UE FM) or functional grasp and release (B&B). However, Subject 102 demonstrated a 30% improvement on the FTHUE by completing 4 additional tasks after training. Subject 102 reported less difficulty with arm and hand use after training as measured by the Hand Domain of the SIS, while Subject 103 reported no change.

Discussion
In this report, we describe a newly developed VR system designed to promote UE movement skill in individuals recovering from hemiparesis. Two participants with differing motor severity were able to engage in VR based practice and improve performance over 12 training sessions. We were able to successfully tailor and progress practice content and task difficulty based on each participant's level of movement ability and rate of performance improvement. The feedback provided by the system was 'Rotation' Task Figure 3 'Rotation' Task. The virtual environment consists of two cube configurations that are identical in composition but different in orientation. The participant rotates and laterally moves the green cubes to superimpose them onto the static blue cubes by matching their orientation. Movement of the green cubes is controlled by a magnetic marker attached to a cylinder held in the paretic hand or directly onto the dorsum of the hand at the 3 rd metacarpal head. Blocks were configured to require progressively greater amounts of supination ranging from 15° to 150° (from a start position of full pronation). Each practice block contained 20 trials, 10 requiring supination and 10 requiring pronation. A maximum of 60 seconds was allowed for each trial. useful to the supervising therapist in setting goals, monitoring change in performance, grading task difficulty, and demonstrating performance change to the participant.
Others have reported improvement in UE movement capability in individuals recovering from stroke after training in a virtual environment. Merians et al. [6,16] found improvements in hand function following 2 to 3 weeks of training on VR tasks. The tasks used in those studies focused primarily on hand and finger ability. Our system includes only one task that addresses hand function ('Pinch'), specifically a thumb and index finger pinch, with additional requirements that the grasp be coordinated with a reach movement. Holden et al. [5,17] also demonstrated improved UE function in individuals post-stroke after training reaching movements in a virtual environment. The system used by Holden et al. [5,17] made use of a "virtual teacher" to demonstrate optimal task completion and provide guidance to the user. We did not provide guidance during task performance but provided summary feedback at the completion of each practice block (10 to 20 trials) in order to engage the participant in anticipatory motor planning and problem solving throughout practice.

Conclusion
The VR system and tasks described in this pilot study provided a challenging practice environment that allowed individually-tailored practice progression. Future work is underway to further validate task design and configuration, develop hypothesis-driven algorithms for optimal task progression, evaluate transfer and persistence of training to real world activities, and incorporate more gaming features.