An Intelligent Assessment System for Human Motor Functions of Stroke Patients

Assessment is an essential part of the rehabilitation process for post-stroke patients, while due to the low accuracy and long duration of traditional rehabilitation assessment methods, as well as the fact that the assessment mainly relies on the subjective judgment of doctors, there is a lack of efficient, high-precision and objective intelligent assessment methods. Facing the above needs, this study developed a lightweight human motor function dynamic analysis system based on the Fugl-Meyer assessment scale to evaluate the different rehabilitation stages of post-stroke patients. We use a cell phone as the lightweight device to dynamically track the changes of patients’ motion in multiple sensitive motion paradigms and identified motion vector centroids by sliding window convolution to perform normalization of temporal features. Based on this, we process temporal information by simulating RGB image features skillfully and use a multimodal decision fusion model consisting of a convolutional neural network (CNN) and a long and short-term memory (LSTM) network to achieve quantitative scoring of patients’ rehabilitation degrees. Experimentally verified by 12 participants from the China Rehabilitation Research Center (CRRC), the system proved to be effective in assessing the rehabilitation level of stroke patients, and significantly improved the efficiency and precision of the existing assessment methods.