Advanced machine learning for real-time tibial bone force monitoring in runners using wearable sensors

Abstract

This study explores the innovative integration of machine learning (ML) with wearable sensor technologies for the real-time monitoring of tibial bone force in runners. Utilizing three distinct approaches—a linear regression model based on the Vertical Average Loading Rate (VALR), a physics-based method, and a sophisticated ML algorithm—this research was conducted with 10 participants equipped with wearable sensors. The results revealed the ML model's superior performance over both the physics-based and VALR techniques, achieving Mean Absolute Percentage Errors (MAPEs) of 6.7 percent and 11.3 percent respectively. This was accomplished through extensive training across various datasets. Additionally, the ML approach demonstrated remarkable reliability in 25 different running session simulations, underscoring its effectiveness in complex musculoskeletal analytics. These findings not only highlight the potential of ML in enhancing wearable technology for biomechanical data analysis but also emphasize the necessity for further comprehensive studies in this field.