AI-Based Prediction of Grip Strength Fatigue and Asymmetry in Korean Coast Guard officers Using IoT-Enabled Dynamometric Data: Toward Personalized Rehabilitation and Training Systems

Abstract

Background: Handgrip strength (HGS) is a widely recognized biomarker of physical capability and overall health. However, conventional assessments rely on static peak values and fail to capture temporal variations that reflect fatigue accumulation and muscular asymmetry. Recent advances in Internet of Things (IoT) technologies now enable continuous biomechanical monitoring, offering new opportunities for precision occupational health management. Objective: This study aimed to examine dynamic grip performance among Korean Coast Guard officers using an IoT-enabled handgrip device and to evaluate the feasibility of artificial intelligence (AI) models in predicting fatigue risk and interlimb asymmetry. Methods: A total of 160 participants completed bilateral grip trials using a continuous IoT-based dynamometer that recorded mean force, asymmetry, fatigue index, coefficient of variation, and other derived parameters. Random Forest and Gradient Boosting algorithms were trained to classify participants into high- and low-fatigue risk groups. Model performance was evaluated using AUC, F1-score, and accuracy metrics, while explainable AI analysis (SHAP) identified key predictors. Results: Both models demonstrated strong predictive performance (AUC = 0.86–0.88; accuracy > 0.83). Fatigue index and asymmetry were identified as the most influential predictors, followed by years of service and mean handgrip strength. Continuous data analysis revealed that temporal grip variability provides valuable insights into neuromuscular efficiency beyond absolute force measurements. Conclusion: IoT-enabled continuous grip monitoring combined with interpretable AI offers a novel approach for detecting occupational fatigue and muscular imbalance. These findings suggest that dynamic digital biomarkers can enhance preventive ergonomics, inform personalized rehabilitation, and support the development of real-time fatigue management systems for high-demand professions.