Multimodal injury risk prediction in tennis, 2024

This work aims to find more promising sources of data from various aspects in improving tennis players’ performance and injury prediction. The diverse sources of data in this work are collected from nine collegiate tennis players, including physiological metrics, training and match data, sleep data from wearable devices, self-reported information via daily questionnaires, jump assessments conducted by experts, and videos of match play. Accordingly, this work proposes a multimodal Predictive Athlete Readiness framework for Tennis (PART) for assessing both performance and injury risk in tennis players. From the multiple-source data, PART captures four characteristics: the overall wellness, injury risk, physical capability and playing style of tennis athletes using a range of machine learning models, including Linear Regression, XGBoost, Logistic Regression, Decision Tree, Random Forest, deep learning models and video analysis techniques such as motion analysis. Our framework integrates these four characteristics and provides a holistic assessment of the athlete’s condition by our defined Athlete Readiness Score (ARS), which learned by supervised learning. Our framework also provides advanced forecasts of specific body areas at risk, such as the upper body (e.g., elbows) or lower body (e.g., knees). This advanced injury risk assessment, combined with performance prediction, operates cohesively within the framework to deliver meaningful and precise guidance for tennis athletes. Our framework also shows promise for extension to recreational tennis players, who often suffer from injuries due to incorrect playing techniques.