Automatic prediction of obstructive sleep apnea event using deep learning algorithm based on ECG and thoracic...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Sleep Medicine

Background:

• Obstructive sleep apnea (OSA) is a prevalent sleep disorder associated with numerous health complications.
• Accurate and accessible detection methods for OSA are crucial for timely intervention and management.

Purpose of the Study:

• To develop and validate an automated deep learning model for obstructive sleep apnea (OSA) event detection.
• To investigate the efficacy of combining electrocardiogram (ECG) and thoracic movement signals for enhanced OSA detection.

Main Methods:

• Utilized retrospective polysomnography (PSG) data from 420 cases.
• Employed the ResNeSt34 deep learning algorithm to construct models using ECG alone and combined ECG with thoracic movement signals.
• Evaluated model performance using accuracy, precision, recall, F1-score, ROC, and AUC metrics.

Main Results:

• The model integrating both ECG and thoracic movement signals achieved superior performance compared to the ECG-only model.
• Combined signal model yielded accuracy (89.0%), precision (88.8%), recall (89.0%), F1-score (88.2%), and AUC (92.9%).
• ECG-only model achieved accuracy (84.1%), precision (83.1%), recall (84.1%), F1-score (83.3%), and AUC (82.8%).

Conclusions:

• An automated OSA event detection model leveraging combined ECG and thoracic movement signals with the ResNeSt34 algorithm demonstrates high reliability.
• This model is a promising tool for effective and efficient screening of obstructive sleep apnea.