Sarcopenia is a progressive condition in older adults associated with increased risks of falls, disability, hospitalization, and mortality. Despite its clinical relevance, early and accurate diagnosis remains limited due to reliance on indirect and subjective screening methods. This study introduces a passive, camera-based framework, Motion-AI Integrated Surveillance for the Elderly (MAISE), that seamlessly estimates joint torque from natural daily movements to provide a functional assessment of lower-limb muscle performance. The framework utilizes a CoP (Center of Pressure) limiter to enforce biomechanical constraints during ground reaction force estimation, facilitating precise torque calculation from motion data alone. A generalization evaluation was performed in 28 sarcopenic and non-sarcopenic older adults on movement tasks that were not included in the training dataset, allowing assessment of the model’s ability to generalize to unseen activities. For each lower-limb joint, Peak Torque (PT), Rate of Torque Development (RTD), and Power Range (PR) were extracted and analyzed in relation to established clinical metrics of Sarcopenia. These torque-based parameters were also compared between Sarcopenic and healthy groups to identify joint-specific functional differences. The proposed CoP limiter improved the accuracy of joint torque estimation without force plates by reducing center of pressure error by up to 49.3 percent and ground reaction force error by up to 6.5 percent on unseen data. Across various model configurations, the framework achieved RMSE values of 0.034 m for CoP, 0.752 N/kg for GRF, and 0.243 Nm/kg for GRM during testing. The computed torque metrics demonstrated strong correlations with conventional Sarcopenia indicators such as grip strength, gait speed, and chair-stand time. These correlations were both task- and joint-specific, allowing identification of localized muscle weaknesses and compensatory strategies. Significant differences in torque profiles were also observed between Sarcopenic and healthy individuals, particularly in tasks involving rapid force generation or postural control. The proposed framework enables quantitative assessment of joint-specific functional muscle performance using only motion data, providing preliminary insight toward personalized intervention. Its passive and non-contact implementation supports unobtrusive, longitudinal monitoring, which is essential for capturing the slow and often subclinical progression of Sarcopenia and for informing timely clinical decisions.
Jo et al. (2026) studied this question.