Real-Time Evaluation of Hand Motor Function Recovery in Home Use Finger Rehabilitation Device Using Gaussian Process Regression

Abstract

Continuous hand rehabilitation after discharge is important for hemiplegic patients to regain an independent finger movement. However, most patients cannot rehabilitate by themselves without therapists. For this problem, robotic rehabilitation has been investigated to support patients even at home. Most of the programs performed by these robots are focusing on the assistance for voluntary movement. However, the approach to the voluntary movement is not enough for regaining dexterous movement. Voluntary suppression of body parts that should not move is important. However, previous studies focusing on voluntary suppression are few. In this paper, we show a detailed program for voluntary suppression rehabilitation. The program is performed by our robotic finger rehabilitation device aiming at home use. In this program, a patient is requested to flex and extend an index finger independently. During moving, individual pressure sensors monitor the other fingers. If the device detects unnecessary movements such as compensatory movement at some fingers, the patient is notified that unnecessary movements are found there. The detection is based on 3σ range of healthy subject's finger pressure data which was constructed by using Gaussian Process Regression. Through experiments with hemiplegic patients, we have shown that the frequency of deviation of patients' data from 3σ range of healthy subjects decreases according to the degree of recovery.