Biomarkers

Area of Science:

• Gerontology and Biomedical Engineering
• Human-Computer Interaction
• Machine Learning for Healthcare

Background:

• One in four individuals with Alzheimer's disease (AD) wish to age in place independently.
• Remote caregiving for AD patients faces challenges in monitoring activities of daily living (ADLs).
• Existing monitoring technologies have limitations in detecting granular ADL information.

Purpose of the Study:

• To evaluate the potential of structural vibration technology for detecting ADLs.
• To differentiate between coarse-grained (activity labels) and fine-grained (specific actions) ADL detection.
• To address the gap in remote monitoring for AD patients.

Main Methods:

• Utilized geophone sensors and machine learning (ML) to analyze vibration patterns.
• Collected data in a simulated apartment testbed with sensors in key living areas.
• Ten participants performed 30 activity sequences, with data manually labeled against video ground truth.

Main Results:

• Successfully detected activities like walking, talking, and medication intake in real-time.
• Identified specific fine-grained actions within broader activities.
• Vibration data visualization showed distinct clusters for different activities, indicating unique patterns.

Conclusions:

• Structural vibration sensing offers an innovative, unobtrusive method for ADL detection.
• The technology can identify both the occurrence of activities and specific contributing actions.
• Vibration sensing shows promise for intervention and outcome assessment in Alzheimer's disease clinical trials.