Dementia Scale Classification with Sequential Model from Sleep Activity Data.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Dementia is a brain disorder impacting cognitive function and sleep patterns.
• Monitoring sleep activity offers a potential non-invasive method for assessing cognitive status changes.
• Early detection of cognitive decline is crucial for timely intervention and management.

Purpose of the Study:

• To develop a machine learning model for dementia classification using sleep activity data.
• To explore the feasibility of using low-burden sleep monitoring for cognitive assessment.
• To identify specific sleep features associated with lower cognitive scores.

Main Methods:

• Collected sleep activity data (heart rate, respiration, sleep depth) from 124 elderly participants using a single sensor.
• Utilized the Mini Mental State Estimation (MMSE) test to determine cognitive status.
• Applied statistical analysis and sequence modeling (LSTM) for time-series analysis and binary classification.

Main Results:

• Identified significant differences in sleep patterns between participants with high and low cognitive status.
• Achieved a maximum macro F1 score of 0.67 in classifying dementia using LSTM models.
• Demonstrated the potential of sleep activity data for predicting dementia classification.

Conclusions:

• Sleep activity data exhibits distinct patterns correlating with cognitive status in the elderly.
• Machine learning models, particularly LSTMs, can effectively utilize sleep data for dementia classification.
• Sleep monitoring presents a promising, low-burden approach for dementia screening and prediction.