Deep Learning Approach for Sleep Apnea Detection Using Single Lead ECG: Comparative Analysis Between CNN and SNN

Sleep apnea, a sleep disorder causing interrupted breathing during sleep, leads to issues like fatigue and cardiovascular problems. Deep learning technology offers a potential solution for accurate and cost-effective sleep apnea detection, improving early diagnosis rates and management of sleep-related disorders. Recent studies are focusing on detecting sleep apnea using machine learning algorithms and neural network models. This study proposes a unique method for identifying sleep apnea using Deep 2D Convolutional Neural Network (CNN) and Spiking Neural Networks (SNN). Evaluation with a custom sleep dataset from PhysioNet apnea ECG database shows that the SNN model outperforms the deep 2D CNN model, achieving 89.2% accuracy, 88% f1 score, 87.3% recall, and 89% precision. This study highlights the critical role of deep learning techniques in improving the detection of sleep apnea, specifically through the analysis of ECG images. The outstanding performance of the Spiking Neural Network (SNN) model implies its potential as an effective tool for improving diagnostic accuracy and clinical management. In this study, we present findings on a unique deep learning model for sleep apnea diagnosis, demonstrating its comparative superiority over conventional methods, which could prove to be a valuable tool in the detection of future sleep-related diseases. The successful application of these advanced methodologies not only contributes to the scientific understanding of sleep apnea but also holds promising implications for enhancing healthcare outcomes and promoting patient well-being. This approach presents a reliable, fully automated deep learning solution that could greatly aid physicians in decision-making regarding sleep apnea diagnosis.