Automated Detection of Sleep Apnea Using Convolutional Neural Network from a single-channel ECG signal

Abstract

Sleep apnea (SA) is the most common sleep disorder to lead some serious cardiovascular diseases and neurological if left it alone. In this paper, a convolutional neural network (CNN) model with four 1D convolutional layers, two fully connected layers and one classification layer is presented to detect automatically SA from a single-channel electrocardiogram (ECG) signal, each convolutional layer is followed by rectified linear units (ReLU) activation function, max pooling and dropout operations. 70 ECG recordings from the Apnea-ECG dataset are used for evaluating the model. RR interval, which is time interval from one R wave to the next R wave, and R-peaks amplitudes from a single-channel ECG signal are employed as the input of the CNN model. We performed our experiment on single-channel ECG signal dataset and have achieved the advanced performance with overall classification accuracy of 87.9% and 97.1% on the per-segment classification and per-recording classification respectively. This model can effectively be used to detect SA from a single-channel ECG signal.