Sleep Quality Evaluation Based on Single-Lead Wearable Cardiac Cycle Acquisition Device

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Cardiology

Background:

• Polysomnography (PSG) is the gold standard for sleep disorder detection but is complex.
• Subjective sleep quality in healthy adults is often assessed via questionnaires.
• There's a need for simplified methods to assess sleep quality using wearable physiological signals.

Purpose of the Study:

• To develop a two-step method for classifying sleep quality as "good" or "poor" using single-lead wearable cardiac data.
• To compare the method's performance against subjective sleep questionnaire scores.
• To enable convenient daily sleep health monitoring.

Main Methods:

• Extracted heart rate variability (HRV) and ECG-derived respiration features.
• Constructed a sleep staging model (wakefulness, REM, light, deep sleep) using multi-classifier fusion.
• Developed a sleep quality evaluation model classifying "good" vs. "poor" sleep based on sleep staging results.

Main Results:

• Sleep staging model accuracy: 0.661 (Cardiology Challenge 2018) and 0.659 (SHHS Visit 1).
• Sleep quality evaluation model accuracy: 0.786 with an average F1-score of 0.771 on recording subjects.
• The models require only a single channel of wearable cardiac cycle signal.

Conclusions:

• A simplified, single-lead cardiac signal-based method can effectively assess sleep quality.
• The developed models are easily transferable to portable devices for daily sleep monitoring.
• This approach facilitates accessible and continuous sleep health management.