Predicting anterior cruciate ligament changes during unexpected side cutting based on inertial sensors

Abstract

In sports and athletic competitions, efficient lateral cutting maneuvers (LCMs) and directional changes are essential for optimal performance. This study involved 18 male athletes who were screened by professional physicians. A Random Forest classification model was employed to predict anterior cruciate ligament (ACL) injuries during outdoor LCMs and running. Results indicated that during 90° LCMs, compared to LCMs in other directions and running, the knee joint experienced greater flexion angles, increased abduction moments, and elevated coronal- plane knee contact forces. OpenSim calculations further demonstrated significantly higher ACL strain values during 90° LCMs relative to other movement conditions. Training and validation of inertial sensor data using the Random Forest model revealed high sensitivity and accuracy in identifying key time-domain and frequency-domain features, specifically "kurtosis," "skewness," "variance," "energy," and the "25th percentile."