Towards clinically relevant automatic assessment of upper-limb motor function impairment

This paper is to develop an automated assessment system of upper-limb motor function impairment for clinical environment. Although we had proposed the system in our previous work, there are some rooms to be improved. Using glove sensor was difficult due to stroke patient's hand contracture. Moreover, it was based on machine learning, and thus required huge effort to collect reference data to increase classification accuracy. To address those issues, three tests of Fugl-Meyer Assessment which were closely related the issues were chosen as target tests. Since Kinect v2 and force-sensing resister can provide hand-related information, the tests were automated without glove sensor. Fuzzy-logic classification table that is based on traditional FMA guidelines was implemented to rate the FMA score without machine learning. With a healthy subject, simple experiments were conducted to evaluate the proposed system with novel classification scheme. The results show a feasibility for more convenient automated assessment of upper-limb motor function impairment.