Nighttime features derived from topic models for classification of patients with COPD.

Area of Science:

• Pulmonary Medicine
• Biomedical Engineering
• Sleep Science

Background:

• Nighttime symptoms are key indicators of disease impairment, especially in respiratory conditions like COPD.
• Current wearable sensor use for COPD sleep assessment is limited to basic metrics like movement and sleep continuity.
• A novel approach is needed to comprehensively describe sleep patterns in COPD patients.

Purpose of the Study:

• To develop and present a novel methodology for concisely describing sleep patterns in individuals with and without COPD.
• To identify higher-level sleep features that can differentiate between healthy individuals and COPD patients.
• To evaluate the utility of these features in assessing COPD disease severity and patient-reported dyspnea levels.

Main Methods:

• Multimodal sleep data was converted into a textual representation.
• Topic modeling was applied to a dataset of over 6000 nights of sleep data.
• Discovered latent sleep structures were used to discriminate between subject groups and assess disease metrics.

Main Results:

• The methodology successfully identified distinct sleep patterns associated with COPD.
• The discovered sleep characteristics effectively discriminated between healthy subjects and those with COPD.
• These novel features correlated with COPD disease severity and levels of dyspnea.

Conclusions:

• The developed approach offers a concise and effective way to describe sleep patterns in COPD.
• Latent sleep structures derived from nighttime data capture crucial aspects of sleep behavior affected by COPD and dyspnea.
• This method holds promise for improved assessment of COPD and related symptoms using wearable sensor data.