Multi-Feature Automatic Extraction for Detecting Obstructive Sleep Apnea Based on Single-Lead Electrocardiography...

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence

Background:

• Obstructive sleep apnea (OSA) is a common sleep disorder linked to cardiovascular diseases.
• Nocturnal polysomnography (PSG) is the standard OSA diagnostic tool but is costly and time-consuming.
• Electrocardiogram (ECG)-based methods offer potential but often require complex feature engineering.

Purpose of the Study:

• To develop an innovative, automated method for OSA classification using ECG data.
• To overcome the limitations of conventional OSA diagnostic techniques.
• To enhance the accuracy and convenience of sleep apnea diagnosis.

Main Methods:

• A composite deep convolutional neural network (CNN) model was developed for OSA classification.
• A multimodal strategy was employed for automatic feature extraction from ECG data.
• ECG data was transformed into scalogram and Gramian angular field matrix images to enrich features.

Main Results:

• The proposed model achieved high diagnostic performance for OSA.
• Achieved accuracy of 96.37%, sensitivity of 94.67%, specificity of 97.44%, and AUC of 0.96.
• The model demonstrated exceptional results on the PhysioNet Apnea-ECG database.

Conclusions:

• The developed deep learning model shows significant potential for efficient and accurate OSA diagnosis.
• This approach offers a promising, convenient alternative to traditional PSG for sleep apnea detection.
• The multimodal feature extraction strategy enhances diagnostic capabilities for OSA.