Sleep Apnea Detection from Single-Lead ECG A Comprehensive Analysis of Machine Learning and Deep Learning Algorithms

Sleep apnea is portrayed by successive episodes in which the wind stream to the lungs is diminished or halted for over ten seconds. A crucial first step in selecting the most effective medications and treatment options is the accurate diagnosis of sleep apnea episodes. Utilizing the 70-record PhysioNet ECG Sleep Apnea v1.0.0 dataset, this study explores deep learning and machine learning (ML) strategies. After pre-handling and sectioning the ECG information, techniques in view of deep learning and ML were utilized to analyze sleep apnea. To meet our requirement for biosignal processing, all networks were modified in the same way. Three arrangements of information were made: a test set for assessing model generalizability on untested information, an approval set for calibrating hyperparameters, and a training set for tweaking model boundaries. In a 5-fold cross-validation approach, this procedure was then repeated five times until all of the recordings in the test set were found. With a responsiveness of 84.26 percent, particularity of 92.27 percent, and exactness of 88.13 percent, separately, half breed deep models played out the best regarding discovery. With regards to distinguishing sleep apnea and other rest issues, this study reveals insight into the exhibition of different ML and deep learning calculations.