MRehab: Mutlimodal data acquisition and modeling framework for assessing stroke and cardiac rehabilitation exercises

Abstract

Post-stroke rehabilitation is always stressful in-home settings due to the unaccustomed environment, irregular sleep, and undergoing rehabilitation exercises. Usually, the intensity and difficulty of the exercise are inherent complex problems for the patients to manage daily. Physical rehabilitation is essential for all stroke patients to recover. Therefore, an automated in-home rehabilitation system with feedback support both for patient and therapist could assist post-stroke patients in managing and assessing exercise daily to recover faster. This work proposes a data acquisition and analysis framework named “MRehab” that helps collect multimodal sensor signals while patients perform both voluntary and non-voluntary (prescribed) exercises. “MRe-hab” assesses the exercise and physiological states of the patients through signal processing and multiple machine learning models. This framework monitors repetition, patient fatigue, and exercise quality and recommends frequency and intensity.