Data Analytics Framework For A Game-Based Rehabilitation System

Stroke is a major cause of hemiparesis in United States. Constraint Induced Movement therapy (CI therapy) is an effective treatment for upper extremity hemiparesis; however it is inaccessible to most patients. To make it more accessible, we developed a game-based rehabilitation system incorporating the major rehabilitation principles from CI therapy. We introduce a data analytics framework for our rehabilitation system in this paper that can provide objective measures of motor performance during gameplay. We design techniques of preprocessing collected data and propose a series of kinematic measurements, which are used to assess the motor performance and supplement in-clinic measures of therapeutic effect. We also present contextual filtering techniques to enable comparing movement production under different conditions, e.g., self-paced versus game paced movement. We apply our data analytics framework on data collected from several participants. Our analysis shows that participants' motor movement improves over the period of treatment, with different participants showing different patterns of improvement, e.g., speed versus range of motion. Results of kinematic measurements during gameplay are highly consistent with in-clinic performance based on the Wolf Motor Function Test. Moreover, our fine-grained trend analysis reveals potential to detect fatigue, which is related to the duration of gameplay.