Sleep-Wake Classification using Statistical Features Extracted from Photoplethysmographic Signals.

Area of Science:

• Biomedical Engineering
• Sleep Science
• Machine Learning

Background:

• Sleep quality significantly impacts overall health.
• Polysomnography (PSG) is the gold standard for sleep-wake stage detection but is inconvenient.
• Non-invasive, accessible methods for sleep monitoring are needed.

Purpose of the Study:

• To develop and evaluate an automated system for sleep-wake stage classification.
• To utilize finger-tip photoplethysmography (PPG) signals for sleep monitoring.
• To compare the performance of K-nearest neighbors (KNN) and Support Vector Machine (SVM) models.

Main Methods:

• Extracted statistical features from PPG signals.
• Employed supervised machine learning models: KNN and SVM.
• Trained and tested models on a dataset of sleep-wake events (80% training, 20% testing).

Main Results:

• The best performing model, medium Gaussian SVM, achieved 72.36% overall accuracy.
• Weighted KNN reached 70.53% accuracy, and quadratic SVM reached 71.33% accuracy.
• The results demonstrate the capability of PPG statistical features in recognizing physiological state changes.

Conclusions:

• Finger-tip PPG signals coupled with statistical features can be used for automated sleep-wake stage detection.
• Machine learning models like KNN and SVM show promise for non-clinical sleep monitoring.
• This approach offers a convenient alternative to traditional PSG for sleep studies.