An Interpretable, Data-Driven, Hierarchical Multi-Domain Fusion Framework for Classification and Motor Function Scoring in Chronic Ankle Instability

Chronic ankle instability (CAI) is a common sports-related musculoskeletal disorder characterized by recurrent sprains and neuromuscular control deficits. It affects a wide range of individuals, from recreational to elite athletes, and poses a substantial healthcare burden. This study proposes an AI-enabled digital twin framework for sports health applications, offering both interpretability and clinical deployability. The framework identifies CAI and enables subtype stratification using a wearable electromyography (EMG) sensor-driven hierarchical multi-domain fusion model, generates fine-grained motor function scores through a probabilistic modeling approach, and further translates the generated scores into clinically interpretable functional stratification to support rehabilitation assessment. SHapley Additive exPlanations (SHAP) - based interpretability reveals key predictive biomarkers underlying model decisions, establishing a transparent and closed-loop framework for personalized rehabilitation. Validation on 150 participants, including CAI patients and healthy controls, confirms robust classification performance (Accuracy = 98.50%, AUC = 0.99), reliable discrimination of CAI subtypes (Accuracy = 87.70%, macro F1-score = 87.20%), and strong concordance between the generated scores and the clinical gold-standard scale (r = -0.908, p < 0.001). This non-invasive, personalized assessment framework supports long-term rehabilitation management of chronic conditions, offering an innovative and cost-effective digital health solution for sports medicine.