Motion Sensor and Computer Vision Based Shoulder Joint Range of Motion Estimation for Rehabilitative Assessment

Musculoskeletal disorders (MSDs) are a leading cause of disability, impacting the lives of 1.71 billion people worldwide. These disorders include common conditions like lower back pain, osteoarthritis, and rheumatoid arthritis, which can make daily activities challenging. Managing MSDs involves performing physiotherapy exercises while maintaining proper postures, and/or involving treatments with medicines or surgical corrections. For MSD, such as frozen shoulder, the condition is characterised by stiffness and pain in the shoulder joint. To address the impact of targeted treatment for MSD through guided range of motion (ROM) enhancement, this paper proposes a remote monitoring system that uses noncontact computer vision (CV) based motion capture technology and wearable inertial measurement unit (IMU) sensors. This study focuses on comparing CV and IMU during abduction and adduction as part of a rehabilitation assessment. Using the MediaPipe pose module, real-time pose detection and landmark estimation enable joint angles calculation without physical markers. Thirty healthy participants performed predefined shoulder movements to evaluate these methods. Results showed agreement between the CV and IMU systems, with root mean square error (RMSE) values indicating minimal errors (mean RMSE [Formula: see text]) for lower angle ranges ([Formula: see text]–[Formula: see text]). However, as ROM increases from [Formula: see text] to [Formula: see text], the RMSE value also increases to [Formula: see text], and for angles greater than [Formula: see text], RMSE was notably as high as [Formula: see text], reflecting greater variability. These findings suggest that integrating CV and IMU technologies offers a reliable, noninvasive approach for tracking ROM, and it also provides immediate feedback to both patients and healthcare professionals.