SCNN: Scalogram-based convolutional neural network to detect obstructive sleep apnea using single-lead...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Sleep apnea affects nearly 1 billion people globally, posing a significant health burden.
• Current diagnostic methods like polysomnography are costly and inconvenient.
• There is a need for accessible and accurate sleep apnea detection methods.

Purpose of the Study:

• To develop a novel deep learning model for detecting obstructive sleep apnea (OSA).
• To utilize single-lead electrocardiogram (ECG) signals for OSA detection.
• To improve the efficiency and accessibility of sleep apnea diagnosis.

Main Methods:

• A scalogram-based convolutional neural network (SCNN) was proposed.
• Continuous wavelet transform (CWT) converted ECG signals into scalograms.
• Empirical mode decomposition (EMD) and CWT were used to create hybrid scalograms for feature extraction.

Main Results:

• The SCNN model achieved high accuracy (94.30%) and F1-score (95.85%) on the Apnea-ECG dataset for per-segment classification.
• The model demonstrated strong performance on the UCDDB dataset with 81.86% accuracy.
• Achieved 100.00% accuracy in per-recording classification for the Apnea-ECG dataset.

Conclusions:

• The proposed SCNN model effectively detects OSA using single-lead ECG signals.
• This novel approach outperforms existing ECG-based OSA detection methods.
• The method offers a promising, cost-effective alternative to traditional sleep apnea diagnostics.