Multi Neural Networks Investigation based Sleep Apnea Prediction

Abstract

Sleep apnea (SA) is recognized as the most important and common type of sleep disorders with several short term and long term side effects on health and prediction of sleep apnea events before they happened can help to prevent these side effects. There are several studies on automated SA detection but not too much works have been done on prediction of apnea's individual episodes. This paper investigated the application of artificial neural networks (ANNs) to predict sleep apnea. Three types of neural networks were investigated: Elman, RBF and feed-forward back propagation on data from 5 patients. Based on the obtained results, generally on all of experiments the best performance is obtained by the feed-forward neural network with average of Area-Under-Curve (AUC) statistic equal to But this superiority was not hold in all individual experiments and each of neural networks were be able to obtain the best result in some cases. This result showed the necessary of more investigation on methods such as dynamic neural networks selections instead of using a fixed model.