Abstract:

Sleep apnea patients have frequent episodes of stopping or lowering airflow to the lungs for more than 10 seconds. The accurate diagnosis of sleep apnea episodes is a critical first step in developing effective medicines and management techniques. The Physio Net ECG Sleep Apnea v1.0.0 dataset contains 70 recordings, and this study examines machine learning and deep learning techniques on those recordings. After pre-processing and segmenting ECG signals, methods for deep learning and machine learning were used to diagnose sleep apnea. To meet our bio signal processing requirement, all networks were modified in the same way. The data was divided into three sets: a training set for fine-tuning model parameters, a validation set for fine-tuning hyper parameters, and a test set for assessing the generalizability of the models on untested data. The process was then repeated five times in a 5-fold cross-validation strategy until all of the recordings were found in the test set. Hybrid deep models were found to have the best detection performance, with the best accuracy, sensitivity, and specificity of 88.13%, 84.26%, and 92.27%, respectively. This study sheds light on how various machine learning and deep learning algorithms perform in terms of detecting sleep apnea and other sleep episodes.

Description:

sleep apnea, electrocardiogram (ECG), detection, deep learning, long shortterm memory, convolutional neural network