Cardiorespiratory Correlation Coefficient Matrices for CNN-Based Sleep Apnea Detection.

Sleep apnea is a disorder characterized by repetitive temporary cessation of breath, with significant health implications if untreated, thus requiring efficient and accurate identification methods. This paper proposes a novel CNN-based approach for apnea detection from physiological signals: ECG, heart rate, SpO₂, and respiratory signals. The novelty of this work consists in the employment of physiological signal correlation coefficient matrices to represent apnea manifestations. For classification, we employ a pre-trained ResNet-50 model with transfer learning and apply data augmentation to enhance model generalization. Matlab simulation results demonstrate high classification performance, validating the proposed method for sleep apnea detection.Clinical Relevance- This demonstrated the potential of AI-based physiological signal correlation analysis for decision support system in sleep apnea diagnosis.