Gait2Hip-60: A Unified Deep Learning Benchmark for Predicting Hip Muscle Forces and Joint Moments from Multi-Cadence Gait Kinematics

In three linesUnified Gait2Hip-60 benchmark comparing LSTM, Transformer, and Mamba to predict hip muscle forces and joint moments from gait kinematics. Transformer outperforms other models (R²=0.819 for forces, R²=0.862 for moments). External validation on 9 femoral head osteonecrosis patients shows moderate generalization (R²=0.537–0.569).

Summary generated by Claude — human-verified