Abstract
BACKGROUND: Elderly patients often face challenges in recovering from distal radius fractures (DRFs), and inadequately guided rehabilitation may lead to delayed healing or secondary injury. OBJECTIVE: This study aims to develop a system that monitors patients during rehabilitation exercises and provides real-time biofeedback to improve healing outcomes and minimize the risk of secondary injuries. METHOD: Volunteers performed a 10-second sequence including relaxation, flexion/extension, radial/ulnar deviation, and fisting. During these exercises, time series data of hand motions and corresponding electromyography (EMG) signals were collected. Musculoskeletal forces and healing outcomes were then computed to build a database for machine learning training and validation. Additionally, EMG signals and musculoskeletal forces were analysed to extract biomechanical insights into rehabilitation dynamics. RESULTS: Volunteers frequently performed unintended additional movements and exhibited distinct muscle activation patterns, even when instructed to perform specific target motions. Different exercises elicited varying mechanical stimuli, influencing cellular differentiation and ultimately impacting healing outcomes. Furthermore, machine learning predictions closely matched the simulation results. CONCLUSION: Individual variations in movement execution and muscle activation underscore the influence of personal movement patterns on rehabilitation outcomes. The simulation results highlight the importance of tailoring rehabilitation protocols to specific healing stages, as optimal exercise plans may vary across different phases of recovery. Furthermore, the strong concordance between simulation outcomes and machine learning predictions validates the prototype's potential to deliver real-time estimations of healing progress.