An Automatic Rehabilitation Assessment System for Hand Function Based on Leap Motion and Ensemble Learning

For stroke patients, hand function assessment is an important part of the hand rehabilitation process. The hand function assessment, however, requires the patient to complete a series of actions under the guidance of the therapist who then scores the patient's performance. This type of assessment is both time-consuming and highly subjective. Therefore, in order to achieve a fast, objective and accurate assessment, this paper adopts a non-contact infrared imaging device, Leap Motion, to measure the patient's motion information and then uses these motion information to infer the hand's rehabilitation level. This paper improves the traditional way of hand function assessment from the following aspects. Only three coherent movements (finger opposition, lift wrist and stretch fingers) are required to complete the assessment, which makes the assessment time shorter and the assessment process easier. At the same time, an assessment algorithm based on the Ensemble Learning is proposed and integrated into the automatic hand function assessment system. In addition, the virtual reality game has been implemented in the assessment system to ensure a satisfactory interaction with patients, which makes the assessment process more interesting and convenient. Using this system, 50 stroke patients underwent clinical trials with the Brunnstrom and Fugl-Meyer assessment scales. The matching rate between the automatic assessment result and the manual Brunnstrom assessment result is 92%, while the matching rate with the Fugl-Meyer assessment result is 82%. Furthermore, Wilcoxon Signed-Rank test and Kappa test are also used to validate the consistency between the automatic assessment results and the manual assessment results. These experiments illustrate that this automatic assessment system is fast, comfortable and reliable.