Background Compensatory movements are commonly observed in older adults with stroke during upper extremity (UE) motor rehabilitation, which could limit their motor recovery. Aim This study aims to develop a compensation-aware virtual rehabilitation system (VRS) that can detect compensatory movements and improve the outcome of UE rehabilitation in community-dwelling older adults with stroke. Methods The VRS development includes three main components: (1) the use of thresholds for determining compensatory movements, (2) the algorithm for processing the kinematic data stream from Kinect to detect compensation in real-time, and (3) the audio-visual feedback to assist older adults with stroke to be aware of the compensation. Two studies were conducted following the VRS development, where Study 1 identified the value of thresholds for determining compensatory movements in two planar motor exercises, and Study 2 provided preliminary validation for the developed VRS by comparing two groups undergoing VR training or conventional training (CT) in a community rehabilitation center. Results The VRS could effectively detect all determined compensatory movements and timely trigger feedback in response to the detected compensatory movements. The VR participants showed significant improvements in Fugl-Meyer Assessment-Upper Extremity (FMA-UE, p = 0.045) and Wolf Motor Function Test (WMFT, p = 0.009). However, the VR and CT groups had no significant differences in outcome measures. Conclusion The VRS demonstrates the ability to detect compensation and the potential of assisting older adults with stroke to improve motor functions. Suggestions are given for further improvements of the VRS to support the older adult with stroke to reduce compensation.

Background: Compensatory movements are commonly observed in older adults with stroke when they take motor practice for rehabilitation, which could limit their motor recovery.Aim: This study aims to develop one virtual rehabilitation system (VRS) that can detect and reduce compensatory movements to improve the quality of upper extremity (UE) movements and hence the outcome of rehabilitation in community-dwelling older adults with stroke. Method: To design and validate the algorithm of compensation detection equipped in VRS, a study was first conducted to recruit 17 healthy and 6 stroke participants to identify and quantify compensatory movements when they played rehabilitation games provided by the VRS. Then a pilot study was conducted to test the feasibility and efficacy of the VRS deployed in community, where 18 stroke participants were assigned to either virtual reality (VR) group or conventional treatment (CT) group, and each participant underwent 10 sessions of an additional 6 minutes of VR games or CT respectively, on top of their usual rehabilitation programme. Participants were assessed before and after interventions using Fugl-Meyer Assessment-Upper Extremity (FMA-UE), Wolf Motor Function Test(WMFT), Stroke Rehabilitation Motivation Scale (SRMS), Range of Motion (ROM) measurements and the number of compensatory movements.Results: VR group demonstrated a trend in reduction of trunk and upper-extremity compensations, increased intrinsic motivation scores, and statistically significant improvements in FMA-UE (p=0.045) and WMFT (p=0.009, p=0.0355) scores. There was, however, no significant difference in all outcome measures between two groups. Conclusion: The compensation-aware VRS demonstrates a trend towards reduced compensation and higher motivation level, which could be an effective adjunct to the conventional therapy with less supervision from a therapist as well as be potentially deployed in a community center or at an elder adult’s home.