GIST Develops AI to Track Sarcopenia Using Only Everyday Movements [Reading Science]

Analyzing Joint Function by Simple Movements Such as Standing Up or Picking Up Objects

Quantitative Assessment of Sarcopenia Progression Possible Without Sensors or Expensive Equipment

A new artificial intelligence (AI) technology has been developed that can track the progression of sarcopenia simply by analyzing a patient’s movements. This innovation enables the quantitative assessment of muscle function in the ankle, knee, and hip joints using only everyday movements, without the need for additional sensors or expensive medical equipment. It is expected to facilitate early diagnosis and personalized management of sarcopenia in the era of population aging.

On June 9, Gwangju Institute of Science and Technology (GIST) announced that Professor Ji-Yeon Kang’s research team from the Department of AI Convergence, in collaboration with the Korea Institute of Science and Technology (KIST) and Bitgoeul Jeonnam National University Hospital, has developed an AI technology called 'MAISE (Motion-AI Integrated Surveillance for the Elderly)', which analyzes everyday movements of the elderly to track changes in muscle function due to the progression of sarcopenia.

The operation process of the AI technology 'MAISE' that analyzes sarcopenia using only daily movements. It estimates joint torque from movement data collected by a camera to assess the presence and progression of sarcopenia. Provided by the research team.

Sarcopenia is a disease characterized by a decline in muscle mass and strength due to aging, which increases the risk of falls and fractures and makes independent daily living more difficult. However, current diagnostic methods rely on grip strength tests, gait speed measurements, and muscle mass imaging, making it challenging to continuously monitor the gradual decline in function that occurs during daily life.

To address these limitations, the research team developed an AI framework capable of analyzing sarcopenia status using only everyday movements, such as standing up from a chair, picking up objects, or climbing stairs. The key lies in estimating ‘joint torque’—the force required to move joints—using only information about the person’s movements.

Estimating Joint Force Without Equipment

Conventionally, specialized equipment such as force plates was necessary to precisely analyze joint force. The research team implemented a method that allows the AI to learn the relevant physical laws so it can independently estimate ground reaction force and center of pressure information.

As a result, the physics-informed model reduced center of pressure prediction error by up to 49.3% and ground reaction force error by up to 6.5%, even when tested on elderly data not used in training. This demonstrates the technology’s potential to reliably analyze muscle function in real-world environments.

The research team validated the technology by having a total of 28 participants, including both sarcopenia patients and healthy elderly individuals, perform movements such as standing up from a chair, picking up objects, and stepping onto a platform.

Research team photo. (From left) Ji-Yeon Kang, Professor of AI Convergence at GIST (corresponding author), Jaebeom Cho, Master's student (first author), Ki-Hyun Kim, Doctoral candidate, Jun-Hyung Ha, Professor of Mechanical Engineering at Ulsan National Institute of Science and Technology (UNIST) (at the time of research, KIST), Kanghyun Ryu, PhD at Korea Institute of Science and Technology (KIST), Mingu Kang, Professor at Bitgoeul Chonnam National University Hospital. Provided by GIST

The analysis showed that the joint torque indicators estimated by MAISE exhibited a strong correlation with key clinical sarcopenia assessment indices, such as grip strength, gait speed, and the five-time chair stand test. Additionally, distinct joint torque patterns were observed between the sarcopenia and healthy groups, confirming that muscle function decline can be quantitatively assessed using only everyday movements.

Professor Ji-Yeon Kang stated, “This study demonstrates that it is possible to quantitatively assess sarcopenia by leveraging biomechanical information hidden in everyday movements. It presents the possibility of continuous, daily-life-based muscle function monitoring, beyond isolated hospital-based testing.”

She added, “We expect to further develop this into a camera-based, contactless monitoring technology that can contribute to the early detection and personalized management of sarcopenia.”

This research was supported by the Ministry of Science and ICT and was published in April in the international journal of rehabilitation engineering, the Journal of NeuroEngineering and Rehabilitation.

This content was produced with the assistance of AI translation services.