Discriminating and Measuring Activities of Daily Living

Area of Science:

• Biomedical Engineering
• Neurology
• Artificial Intelligence

Background:

• Activities of Daily Living (ADLs) assessment is crucial for diagnosing neurodegenerative diseases like Alzheimer's.
• Current ADL assessment methods rely on subjective human input, lacking objective accuracy.
• Wearable sensors offer a potential solution for objective and continuous ADL monitoring.

Purpose of the Study:

• To explore automatic and objective quantification of ADLs using wearable sensor data.
• To detect basic activities such as eating, brushing teeth, walking, and inactivity.
• To compare contemporary time series classification approaches for ADL detection.

Main Methods:

• Utilized data from a wearable medical device recording eye- and head-movements.
• Focused on a subject-independent testing framework for robust evaluation.
• Applied and compared various time series classification methods, including a MUSE classifier.

Main Results:

• Achieved a peak mean accuracy of 84.72% for detecting four basic ADLs.
• Developed a system for classifying activity durations (30s, 60s, 90s) with perfect discrimination.
• Demonstrated the potential of wearable sensors for objective ADL quantification.

Conclusions:

• Objective ADL quantification via wearable sensors is feasible and accurate.
• This approach can aid in the diagnosis, monitoring, and care of individuals with degenerative brain conditions.
• Accurate ADL measurement may serve as a valuable biomarker for disease progression.