Automatic range-of-motion assessment of the knee from videos

Details

Supervisors

Absil, Pierre-Antoine ; Pauwels, Andries

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en science des données, orientation technologies de l'information, à finalité spécialisée: analyse de données

Abstract

This master's thesis presents an approach for estimating knee range of motion (ROM) from videos of patients performing exercises. In collaboration with moveUP, a healthcare company, we explore the integration of machine learning models that use pose estimation techniques to predict keypoints on the human body, enabling accurate and efficient assessment of knee ROM. Our approach employs recent advancements in human pose estimation, integrating state-of-the-art methodologies to develop a dashboard for uploading videos, demonstrating the potential for real-time feedback. The results show high accuracy rates of 92% within a 10-degree error and 76% for a 5-degree error. This research contributes to the field of telemedicine by demonstrating the feasibility and benefits of integrating advanced pose estimation models into telerehabilitation systems, ultimately aiming to improve patient outcomes and streamline healthcare processes.