Using Deep Learning to Detect Rehabilitation Exercise for Patients Diagnosed with Sarcopenia

Sarcopenia is a common condition among older adults, characterized by a progressive decline in skeletal muscle mass, strength, and functional capacity. This deterioration severely impacts patients’ ability to perform activities of daily living and reduces their overall quality of life. Traditional rehabilitation methods often require patients to frequently visit medical facilities and undergo training under professional supervision, posing logistical and practical challenges for many individuals. Consequently, developing an effective home-based rehabilitation monitoring system has become an urgent priority, particularly for homebound sarcopenia patients. This study aims to develop an innovative rehabilitation exercise monitoring system utilizing MMPose, a non-contact skeletal detection technology, to detect skeletal structures and assess the execution of prescribed rehabilitation exercises. The proposed system emphasizes convenience and exercise accuracy while providing a comprehensive analysis of movement patterns. Additionally, it employs time-series alignment to integrate real-time feedback and objective recommendations to optimize rehabilitation plans. By leveraging advanced technological methodologies, this study seeks to enhance rehabilitation outcomes, introduce novel assessment tools to the field of rehabilitation medicine, and alleviate the economic burden associated with recovery.