Can Machine Learning Enhance Computer Vision Predicted Wrist Kinematics Determined from a Low-Cost Motion Capture System?

Links

Abstract

Wrist kinematics provide insight into repetitive strain injury development. Markerless motion capture extends kinematic assessment beyond laboratory and clinical confines. Our method uses open-source computer vision software with machine learning to predict wrist angles. Sixteen participants performed range-of-motion tasks recorded with four GoPro video cameras and a VICON motion capture system. Using data obtained from the GoPro cameras, open-source software DeepLabCut and Anipose generated 2D and 3D datasets containing anatomical landmark locations used to determine flexion-extension (FE) and radial-ulnar deviation (RUD) wrist angles. Machine learning algorithms for 2D (2DML) and 3D (3DML) methods yielded Mean Absolute Error (MAE) results of FE (10.9°) and RUD (5.2°) for 2DML and 5.9° (FE) and 5.2° (RUD) for 3DML. The 2DML and 3DML methods excelled using deep neural networks to improve prediction accuracy compared to DeepLabCut and Anipose in isolation, which shows promise for the advancement of accessible, accurate motion capture technology

Cited Works (3)

Year	Entry
2021	Kanko RM, Laende E, Selbie WS, Deluzio KJ. Inter-session repeatability of markerless motion capture gait kinematics. J Biomech. May 24, 2021;121:110422.
1996	Cappozzo A, Catani F, Leardini A, Benedetti M, Della Croce U. Position and orientation in space of bones during movement: experimental artefacts. Clin Biomech (Bristol, Avon). March 1996;11(2):90-100.
2005	Wu G, van der Helm FCT, Veeger HEJ, Makhsous M, Van Roy P, Anglin C, Nagels J, Karduna AR, McQuade K, Wang X, Werner FW, Buchholz B. ISB recommendation on definitions of joint coordinate systems of various joints for the reporting of human joint motion, II: shoulder, elbow, wrist and hand. J Biomech. May 2005;38(5):981-992.