Biomechanical Analysis of Musculoskeletal Systems

Abstract

Musculoskeletal disorders (MSDs) -- including osteoarthritis (OA), anterior cruciate ligament (ACL) injuries, and lumbar disc degeneration -- affect 1.71 billion people globally and are the leading contributor to disability-adjusted life years. Biomechanical analysis quantifies the forces, moments, and stresses acting on joints and tissues during movement, providing mechanistic insight into injury aetiology and rehabilitation effectiveness. However, traditional inverse dynamics requires expensive laboratory-grade motion capture systems ($100,000-500,000), force plates, and expert operators, limiting clinical accessibility. This study developed and validated five biomechanical analysis approaches -- marker-based optical motion capture (OMC) with inverse dynamics, inertial measurement unit (IMU) motion estimation, markerless video-based pose estimation (OpenPose + DeepLabCut), finite element analysis (FEA) of knee joint contact mechanics, and a proposed deep learning musculoskeletal model (DL-MSK) that predicts joint kinematics, kinetics, and muscle forces directly from wearable IMU data -- across three clinical biomechanical tasks: gait analysis in knee OA patients (n = 48), ACL injury risk screening in athletes (n = 72), and spinal loading assessment in manual workers (n = 36). DL-MSK achieved the highest agreement with the OMC gold standard for knee joint moments (RMSE 0.12 +- 0.03 Nm/kg; r = 0.94) and hip contact forces (RMSE 0.28 +- 0.06 BW; r = 0.91), significantly outperforming IMU-based inverse dynamics (RMSE 0.24 +- 0.06 Nm/kg; p < 0.001) and video-based estimation (RMSE 0.32 +- 0.08 Nm/kg; p < 0.001). For ACL risk screening, DL-MSK identified high-risk landing biomechanics (knee valgus moment > 0.40 Nm/kg) with 91.4% sensitivity and 88.6% specificity, enabling field-based screening without laboratory infrastructure. These results establish deep learning musculoskeletal models as a clinically viable alternative to laboratory motion capture for biomechanical assessment.