Anterior cruciate ligament (ACL) injuries are among the most serious and economically burdensome events in competitive football. This paper presents a wearable sensor-based framework that fuses triaxial inertial measurement unit (IMU) data, dual-channel surface electromyography (sEMG), and optical heart rate signals to assess ACL injury risk in near real time. Three families of classifiers — traditional machine learning, gradient boosting ensembles, and deep recurrent networks — were evaluated under a shared preprocessing and cross-validation protocol. CatBoost achieved the highest accuracy (92.3%, recall 0.93), while the proposed LSTM–GRU hybrid attained an F1-score of 0.92, reflecting strong temporal modelling of biomechanical sequences. Results suggest that multimodal sensor fusion can reliably separate normal from high-risk movement patterns, offering a viable route towards deployable athlete monitoring at both professional and grassroots level.
C et al. (Mon,) studied this question.