Systematic Review of Detecting Sleep Apnea Using Artificial Intelligence: An Insight to Convolutional Neural Network Method

Abstract

Background: Sleep apnea is a prevalent sleep disorder, especially in males and older ages. The common diagnostic methods, including polysomnography (PSG), are expensive, difficult to perform, and time-consuming. Numerous studies are focusing on developing easy-to-perform methods based on artificial intelligence (AI) for the early diagnosis of sleep apnea. This systematic review aimed to gather current methods based on the convolutional neural network (CNN) for the diagnosis of sleep apnea. Methods: Three international electronic databases (PubMed, Web of Science [WoS], and Scopus) were searched from 2010 to October 2023. All studies that have developed CNN-based methods for the diagnosis of sleep apnea and have accomplished the performance tests were included. Finally, the characteristics of the studies were extracted and summarized. Results: A total of 36 studies were included in this systematic review. Various physiological signals have been proposed to detect sleep apnea, including electrocardiogram (ECG), blood oxygen saturation (SpO2), sound signals, respiratory signals, electroencephalogram (EEG), and nasal airflow. Electrocardiogram was the most frequently used signal in the studies, followed by SpO2. The highest reported accuracy was achieved by SpO2 or ECG-based methods and with a one-dimensional CNN (1D-CNN) classifier. Using multiple signals did not necessarily increase the performance of test results. Conclusions: Diagnostic methods based on CNN can be used only as screening tools or home diagnosis of sleep apnea. These methods are easy to perform and can only reduce the diagnostic costs and waiting time for a sleep study in special scenarios. Nevertheless, PSG is still the gold standard for the diagnosis of sleep disorders.