Machine learning-based muscle mass estimation using gait parameters in community-dwelling older adults: A...

Area of Science:

• Gerontology
• Biomechanical Engineering
• Machine Learning

Background:

• Loss of skeletal muscle mass (sarcopenia) is linked to poor health outcomes in older adults.
• Current assessment methods for muscle mass may not be universally accessible or simple.
• Gait analysis shows potential for estimating geriatric risks, but its link to muscle mass is unexplored.

Purpose of the Study:

• To apply machine learning to gait parameters for distinguishing low skeletal muscle mass in older adults.
• To identify specific gait parameters most effective in detecting low muscle mass.

Main Methods:

• Sixty-six community-dwelling older adults participated.
• Thirty-two gait parameters were extracted from 3D skeletal models during comfortable walking.
• Skeletal muscle mass was measured via bioimpedance analysis; low muscle mass was defined by the Asia Working Group for Sarcopenia guidelines.
• An eXtreme gradient boosting (XGBoost) model was used for classification.

Main Results:

• The XGBoost model achieved a c-statistic of 0.7, with 59.5% sensitivity and 81.4% specificity.
• The most influential gait parameters for detecting low muscle mass were stride length, hip dynamic range of motion, and trunk rotation variability.

Conclusions:

• Machine learning-based gait analysis offers a viable method for assessing skeletal muscle mass in community-dwelling older adults.
• This approach can aid in the early detection of sarcopenia through non-invasive gait measurements.