Gait-Based AI Models for Detecting Sarcopenia and Cognitive Decline Using Sensor Fusion.

Area of Science:

• Gerontology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Sarcopenia and cognitive decline (CD) significantly impact aging populations' quality of life.
• Early detection is hindered by the limitations of traditional in-person screening methods.
• Developing accessible, regular screening tools is crucial for timely intervention.

Purpose of the Study:

• To develop AI algorithms for detecting sarcopenia and CD using gait analysis.
• To integrate sensor and computer vision (CV) data for enhanced predictive accuracy.
• To explore the potential of multimodal gait analysis for early disease detection.

Main Methods:

• A cross-sectional case-control study involving 42 older adults (≥60 years).
• Gait patterns assessed using foot/lumbar sensors and CV data during usual walking.
• Machine learning models developed using extracted gait variables to predict sarcopenia and CD.

Main Results:

• AI models demonstrated high predictive accuracy for both CD and sarcopenia.
• The best CD model achieved an F1-score of 0.914 (95% sensitivity, 92% specificity).
• Combined sensor and CV model for sarcopenia yielded an F1-score of 0.748 (100% sensitivity, 83% specificity).

Conclusions:

• Gait analysis via sensor and CV fusion effectively screens for sarcopenia and CD.
• A multimodal approach significantly improves model accuracy for early detection.
• This technology shows promise for home-based screening and intervention in aging populations.