Wearable EMG-IMU System for Real-Time Detection of Compensatory Patterns in Patellofemoral Pain Syndrome

Patellofemoral pain syndrome (PFPS) is associated with compensatory movement patterns and neuromuscular deficits that often go undetected through conventional clinical assessments. This study investigated the effectiveness of an artificial intelligence (AI)-augmented wearable system that integrates surface electromyography (EMG) and inertial measurement unit (IMU) data for real-time detection of movement quality in individuals with PFPS. A cross-sectional study was conducted involving 45 participants (30 with PFPS, 15 controls), who completed functional tasks while EMG and IMU signals were recorded. A supervised machine learning model was developed to classify compensated versus optimal movement patterns and was compared to a baseline IMU-only model. The AI model demonstrated significantly higher classification accuracy (88.45 ± 3.12%) than the baseline (77.12 ± 4.56%, p < 0.001), with superior F1-score (87.92 ± 3.22%) and area under the curve (AUC = 0.93 ± 0.02). Deployment on an edge AI device yielded low inference latency (84.56 ± 5.34 ms) and reduced energy consumption (1.45 ± 0.23 mJ), supporting real-time operation. These findings highlight the potential of AI-integrated wearable systems for objective, efficient assessment of movement impairments in both clinical and remote rehabilitation settings.