Sleep apnea classification using ECG-signal wavelet-PCA features.

Area of Science:

• Biomedical Engineering
• Cardiology
• Sleep Medicine

Background:

• Polysomnography (PSG) is the gold standard for diagnosing sleep apnea but is inconvenient and costly.
• Existing ECG-based methods for apnea detection have limitations in performance.
• There is a need for more accessible and efficient sleep apnea screening tools.

Purpose of the Study:

• To develop and validate a method for classifying subjects as normal or having sleep apnea using single-channel ECG.
• To assess the feasibility of using ECG-derived features for minute-to-minute sleep apnea event detection.

Main Methods:

• ECG signals were decomposed using wavelet decomposition to extract detail coefficients (D3-D5).
• Approximately 15 features were extracted per minute of ECG data.
• Principal Component Analysis (PCA) and Support Vector Machine (SVM) were employed for feature reduction and classification.

Main Results:

• The proposed minute-to-minute classifier achieved 95.20% specificity and 92.65% sensitivity.
• Subject-based classification accuracy reached 94.3% in a dataset of 35 patients.
• The system demonstrated strong performance in differentiating between normal and apnea subjects based on ECG.

Conclusions:

• Single-channel ECG analysis combined with wavelet decomposition, PCA, and SVM is a promising approach for sleep apnea screening.
• The developed system offers a potential alternative to traditional PSG, improving convenience and reducing costs.
• This research provides a foundation for future development of user-friendly sleep apnea screening devices.