AI system detects age-related muscle loss through everyday movements

Sarcopenia — the age-related loss of muscle mass and strength — affects more than 10 percent of adults in their 60s and older, and its prevalence is rising rapidly. The condition impairs walking, raises the risk of falls and physical disability, and can trigger secondary complications including metabolic disorders, obesity, diabetes and reduced bone density. No approved treatment exists, and reliable diagnostic methods remain largely unavailable.

That may soon change: researchers say people will be able to monitor changes in their muscle function over the long term at home — using cameras, smart appliances or home healthcare devices — without visiting a hospital or clinic.

The Gwangju Institute of Science and Technology (GIST) announced that a research team led by professor Kang Ji-yeon of its Department of AI Convergence, working jointly with the Korea Institute of Science and Technology (KIST) and Bitgoeul Chonnam National University Hospital, has developed an AI technology called MAISE. The system tracks and analyzes changes in muscle function tied to sarcopenia progression using only the everyday movements of older adults.

The technology estimates the functional state of the ankle, knee and hip joints — and quantifies the degree of sarcopenia progression — based solely on routine movements such as rising from a chair or picking up an object, with no wearable sensors or expensive equipment required.

Sarcopenia is characterized by an abnormal decline in muscle mass and strength with age, increasing the risk of falls and fractures while reducing a person's ability to live independently.

By the time symptoms become noticeable, significant functional decline has often already occurred, making early detection critical. Current approaches rely mainly on indirect functional tests — grip strength, walking speed and chair-stand tests — or imaging studies to measure muscle mass, making it difficult to continuously monitor the gradual deterioration that unfolds in daily life.

To address this gap, the joint research team developed MAISE, an AI technology capable of assessing sarcopenia using only the movements older adults perform in everyday life.

A key feature is that the AI model was trained with embedded physics information, enabling it to estimate ground reaction force — the force the body receives from the floor when pushing against it, which is needed to calculate joint forces — without any dedicated measurement equipment.

Precise joint force analysis normally requires specialized equipment such as force plates, but the research team made it possible to estimate those forces from motion data alone, laying the groundwork for muscle function analysis in everyday environments.

As a result, the physics-informed model reduced the prediction error for the center of pressure — the point on the floor where a person's weight bears down — by up to 49.3 percent, and cut ground reaction force error by up to 6.5 percent, even on motion data from older adults not used in training.

The findings confirmed that the AI can analyze human movement more accurately and reliably estimate actual muscle function.

The team validated the system with 28 participants — a mix of older adults with sarcopenia and healthy older adults — who performed movements including rising from a chair, picking up an object and stepping onto a platform. The team then assessed how accurately the joint torque data analyzed by MAISE — the rotational force required when a joint moves — reflected actual muscle function and the degree of sarcopenia.

The joint torque indicators estimated by MAISE showed a clear correlation with the major clinical sarcopenia assessment metrics currently in use, including grip strength, walking speed and the time required to complete five chair stands.

The analysis also revealed distinct differences in joint torque patterns between the sarcopenia group and the healthy group, confirming that abnormal muscle function decline can be quantitatively assessed through everyday movements alone.

Professor Kang said the team plans to integrate a decision-making model that clinicians can easily interpret — one that presents functional decline in specific joints or movements in an explainable way — and to develop MAISE into a sarcopenia management platform linkable to personalized exercise prescriptions, rehabilitation robots and digital therapeutics.

The findings were published in the Journal of NeuroEngineering and Rehabilitation, an international journal in the field of rehabilitation engineering.