Wearable AI for on-device frailty assessment

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Wearable Technology

Background:

• Continuous wearable monitoring offers insights into chronic conditions but faces challenges with energy demands and large datasets.
• Existing systems often require off-device inference, necessitating robust network infrastructure and power, limiting clinical integration.
• Machine learning can condense data into actionable trends, but on-device processing is key for practical, long-term use.

Purpose of the Study:

• To develop and validate an edge AI device framework for continuous, on-device biosignal analysis.
• To assess the clinical utility and model stability of the framework for gait-based frailty assessment.
• To demonstrate autonomous, longitudinal analysis of high-sampling-rate biosignals over extended wear periods.

Main Methods:

• Integrated artificial intelligence (AI) with clinical-grade biosignal acquisition at the edge for on-device inference.
• Utilized the framework for gait-based frailty assessment in vivo (N=16).
• Validated clinical utility, model stability, and on-device inference through further in vivo trials (N=14) and ten-day extended wear experiments.

Main Results:

• The device framework achieved clinical-grade fidelity for on-device inference over extended durations.
• Gait-based frailty assessment results matched gold standard diagnostic tools.
• Demonstrated continuous operation without wearer intervention and autonomous longitudinal analysis.

Conclusions:

• The developed edge AI framework enables practical, long-term wearable health monitoring.
• On-device inference with clinical-grade fidelity is feasible for continuous biosignal analysis.
• This technology has the potential to significantly enhance diagnostic and screening tools for chronic conditions.