Cardiorespiratory Correlation Coefficient Matrices for CNN-Based Sleep Apnea Detection

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Signal Processing

Background:

• Sleep apnea is a serious health condition requiring accurate detection.
• Current identification methods may lack efficiency or accuracy.
• Physiological signals offer rich data for non-invasive monitoring.

Purpose of the Study:

• To develop a novel Convolutional Neural Network (CNN)-based approach for sleep apnea detection.
• To utilize physiological signal correlation coefficient matrices as a novel feature representation for apnea.
• To validate the proposed method's efficacy using simulation results.

Main Methods:

• Employing a CNN architecture, specifically a pre-trained ResNet-50 model.
• Utilizing transfer learning to leverage existing model capabilities.
• Generating physiological signal correlation coefficient matrices from ECG, heart rate, SpO₂, and respiratory signals.
• Applying data augmentation techniques to improve model generalization.

Main Results:

• The proposed CNN-based method achieved high classification performance in detecting sleep apnea.
• Matlab simulations validated the effectiveness of using correlation coefficient matrices.
• The approach demonstrated robustness through data augmentation.

Conclusions:

• The novel CNN approach using physiological signal correlation matrices is effective for sleep apnea detection.
• This AI-based method shows potential for integration into clinical decision support systems.
• Further research can explore real-world clinical validation.