CNN-LSTM Based Approach for Sleep Apnea Detection

Abstract

Obstructive Sleep Apnea (OSA) is a respiratory sleep disorder. OSA is affecting a large population all around the world. Many OSA disorders remain undiagnosed due to monitor device limitations. In this paper, we have proposed a sleep monitoring model based on Convolutional Neural Network (CNN) and single-channel Electrocardiogram (ECG) that may be applied to portable OSA monitor devices. In the proposed model, the convolutional layers in CNN learn various scale features and Long Short-Term Memory (LSTM) learns the dependencies which are long-term such as transition rules of OSA. The proposed model is evaluated on the dataset and achieved an accuracy of 97.72% using CNN-LSTM classifier. The outcomes showed that the suggested technique performs better than the benchmarks.