A Sleep Apnea Detection System Based on a One-Dimensional Deep Convolution Neural Network Model Using Single-Lead...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Obstructive sleep apnea (OSA) diagnosis traditionally relies on polysomnography (PSG), which is inconvenient and costly.
• There is a growing need for portable, affordable, and accessible diagnostic systems for OSA.
• Electrocardiogram (ECG) signals offer a potential data source for non-invasive sleep apnea detection.

Purpose of the Study:

• To develop and validate a novel sleep apnea detection system utilizing a deep convolutional neural network (CNN).
• To leverage single-lead 1D ECG signals for accurate and efficient OSA diagnosis.
• To compare the performance of the proposed CNN model against existing feature-based approaches.

Main Methods:

• A 1D deep CNN model was designed, featuring 10 CNN layers for feature extraction and fully connected layers for classification.
• The model was trained and validated using 70 ECG recordings from the MIT PhysioNet Apnea-ECG Database.
• Performance was evaluated using metrics such as accuracy, specificity, and sensitivity for both per-minute and per-recording classifications.

Main Results:

• The proposed CNN model achieved 87.9% accuracy, 92.0% specificity, and 81.1% sensitivity for per-minute apnea detection.
• For per-recording classification, the model demonstrated high performance with 97.1% accuracy, 100% specificity, and 95.7% sensitivity.
• The deep learning approach outperformed several traditional feature-engineering and feature-learning methods.

Conclusions:

• The developed 1D CNN model effectively detects obstructive sleep apnea using single-lead ECG signals.
• This system presents a promising, accurate, and potentially more accessible alternative to conventional PSG for OSA diagnosis.
• The findings highlight the potential of deep learning in analyzing physiological signals for medical diagnostics.