A Research Study On The Diagnostic Capabilities Of Deep Learning Regarding Sleep Apnea

Abstract

Apnea of the Sleeping Disorders is a breathing disorder in which the patient stops breathing repeatedly for ten seconds or more while sleeping. In this study, we focus on a model or technique that employs deep learning using the CNN (computational neural network) approach. The principal disadvantages of type 1 full-night polysomnography compared to type 4 sleep investigations are the time commitment and the space requirements of a sleep lab. An alternative to costly and bulky polysomnography is a portable and affordable SPO2 sensor-based deep convolutional neural network model for sleep apnea detection. In all, 190,000 samples from 50 patients' SPO2 sensors were used. The accuracy of deep convolutional neural networks for snoring and apnea detection will be around approximately 92.3085% when using a loss rate of 2.3 and a cross-entropy cost function.