A Hybrid Model of Inertial Sensors and Convolutional Neural Networks for Real-Time Ankle Injury Risk Monitoring in Professional Athletes

Introduction: Ankle injuries, particularly lateral ankle sprains, are among the most common musculoskeletal injuries in professional sports, often occurring during cutting maneuvers, landings, and jumps. Traditional methods for predicting these injuries are primarily based on clinical criteria and qualitative observation, lacking sufficient accuracy and speed for real-time prediction. This study aims to design and evaluate a hybrid model combining inertial sensors (IMUs) and one-dimensional convolutional neural networks (۱D-CNN) for real-time monitoring of ankle injury risk in professional athletes.Methodology: In this study, motion data from ۸۷ professional athletes across ۱۲ sports disciplines were utilized. Kinematic data were collected using inertial sensors attached to the shank, foot, and thigh. A hybrid model consisting of a one-dimensional convolutional neural network with three convolutional blocks and a pruned LSTM network was designed. The model was evaluated using ۵-fold cross-validation and its performance was compared with classical machine learning algorithms.Findings: The proposed hybrid model achieved ۹۴.۴% accuracy, ۸۷.۸% sensitivity, and ۹۰.۲% specificity in identifying high-risk movements leading to ankle injury. The area under the ROC curve (AUC) was ۰.۹۷۲. The model was able to predict ۷۶.۳% of injury events at least ۱۵۰ milliseconds before occurrence. The model's performance was significantly better than support vector machine (۸۶.۱% accuracy) and random forest (۸۴.۷% accuracy) algorithms (P<۰.۰۵).Conclusion: The proposed hybrid model can serve as a real-time monitoring tool for identifying high-risk movements and preventing ankle injuries in professional athletes, providing preventive intervention by issuing warnings at least ۱۵۰ milliseconds before injury occurrence.

کلیدواژه ها:

Ankle Injury ، Ankle Sprain ، Inertial Sensors ، Convolutional Neural Networks ، Injury Prevention.