A Female Population-Averaged Musculoskeletal Model Outperforms Conventional Male-Based Generic Models in Simulating Female Gait

Most widely used lower-limb musculoskeletal models are derived from male anatomy and adapted to female participants solely by linear scaling, which may not capture sex-specific differences in pelvic and hip geometry. We developed a population-averaged, MRI-derived female lower-limb musculoskeletal model, built from MRI-based models of a cohort of 25 adult women using thin-plate-spline muscle-path mapping, bilateral symmetrisation, and wrapping-surface optimisation. We hypothesised that this average model, adapted to a new individual by standard linear scaling alone, would reproduce that individual’s MRI-based model’s joint kinematics, joint moments, muscle moment arms, and muscle-driven force and activation estimates more closely than a linearly scaled generic male-based model, and that this advantage would be concentrated in pelvis- and hip-dependent outputs rather than distributed evenly across all joints. Using 5-fold cross-validation, the scaled average-female model and the scaled male model were each compared against the held-out individual’s MRI-based model across gait kinematics, joint moments, muscle moment arms, muscle forces/activations, and joint reaction forces. The average-female model outperformed the male model in every output category (Holm-corrected p ≤ 7 ×10 ⁻⁵ ), supporting our primary hypothesis. Consistent with our secondary hypothesis, differences were largest and most sustained for pelvis tilt, hip flexion, and gluteal/adductor moment arms and forces, and smaller for knee and ankle kinematics. Some divergence remained localised to early-stance knee kinematics and patellofemoral loading. The population-averaged female musculoskeletal model is freely available on SimTK ( https://simtk.org/projects/aver_fem ) and is recommended for studies involving female participants, particularly when pelvic and hip biomechanics are the primary outcomes.