Deep Learning Approaches for Early Detection of Obstructive Sleep Apnea Using Single-Channel ECG: A Systematic Literature Review

In this paper, we aimed to analyze and review various deep-learning approaches for obstructive sleep apnea (OSA) detection using single-channel ECG. The aim is to investigate an efficient and robust system for the early detection of OSA using single-channel ECG in different deep-learning approaches. The methodology we implemented was reviewing the literature in preferred reporting items for systematic reviews and meta-analyses (PRISMA) which includes research conducted during the last decade from 2012 to 2022. We explored various sources for collecting research articles relevant to OSA detection and then a total of 1110 papers are chosen. PRISMA framework facilitates the eligibility criteria to down-sample the articles which are most suitable for our review. Over a decade there is sharp growth in deep learning-based classification techniques for sleep apnea detection and particularly after the year 2017. In the year 2022 drastic increase in the using deep learning has been reported. During the years 2017 to 2022, deep learning approaches were used for classification where LSTM, CNN, RNN, pre-trained networks, and hybrid deep neural architectures were implemented. It is explored thoroughly in the research reported during the year 2022 to detect SA using ECG in deep learning since it is simple and requires fewer computational tasks. This paperwill lead future researchers to pursue theirwork by also procuring relevant background knowledge to pursue an efficient method and approach for OSAdetection usingECGmodality. The best deep learning models were the hybrid model which combines both RNN and CNN for the robustness of the proposed model.