Medical Rehabilitation Evaluation using Image Processing of Patients’ Gait

An automatic image processing method to evaluate human gait for patients walking during their rehabilitation process is proposed in this paper. Our present study evaluates abnormal walking of the patients due to some lower limb injuries. In order to detect joint and key points positions of the human body from video data we have used the MoveNet Thunder model. With body parts identified, angles from symmetric articulations of the body are calculated in various frames to evaluate their magnitudes and changes during rehabilitation. Because the image acquisition differs for different patients, some statistical parameters for each stride are calculated and used in further analysis. A rehabilitation degree is defined and computed as the average percentage of the relevant angles of the body segments in relation to the normality domains of those angles.