Deep Learning-Based Sarcopenia Classification through Gait Video Analysis with a Single Mobile Camera

As the global population ages and life expectancy increases, early detection and continuous monitoring of sarcopenia-an age-related decline in muscle mass and strength-are critical for promoting healthy aging. Traditional assessment methods rely on expensive, specialized medical equipment and expert intervention, limiting their practicality for everyday use. To address these challenges, this study proposes a novel vision-based approach for identifying sarcopenia using gait. A total of 92 elderly individuals participated, including 60 patients with sarcopenia and 32 healthy controls. Digital cameras captured each participant's walking motion, from which 2D skeleton sequences were extracted. Our deep learning model, trained on these 2D skeleton sequences along with additional gait-related features, classified sarcopenia and healthy controls with 82.88% sample-wise accuracy and 94.44% subject-wise accuracy on the test dataset.Clinical relevance- This study represents a significant advancement in employing a vision-based approach for the early detection and monitoring of sarcopenia. It provides a promising, accessible solution for at-home muscle health management, eliminating the need for specialized expertise and equipment.