Deep Learning Model for ECG-based Sleep Apnea Detection

Abstract

Sleep apnea occurs when breathing stops for more than 10 seconds at a time during the night. These occurrences must be correctly diagnosed. The recordings began with preliminary processing and segmentation of electrocardiogram (ECG) data. Deep learning and machine learning were used to make the diagnosis of sleep apnea. Each network was modified in the same way to be suitable for biosignal processing. The training, validation, and test sets were used to optimize model parameters and hyperparameters, while the test set was used to evaluate the model's performance on new data. Each recording was tested several times using a technique known as 5-fold cross-validation. Deep learning models had the highest detection accuracy rate of 88.13%. Sleep apnea and other sleep disorders can be difficult to diagnose, but this study demonstrates the effectiveness of various machine learning and deep learning algorithms.