ApneaNet: A hybrid 1DCNN-LSTM architecture for detection of Obstructive Sleep Apnea using digitized ECG signals

Gaurav Srivastava, Aninditaa Chauhan, Nitigya Kargeti, Nitesh Pradhan, Vijaypal Singh Dhaka

Abstract: Highlights•Efficient detection of Obstructive Sleep Apnea using lightweight 1DCNN-LSTM architecture — ApneaNet.•Deep feature extraction from digitized Electrocardiograms using modified 1-Dimensional CNN.•A hybrid 1-dimensional CNN-LSTM model inspired by Alexnet’s architecture is proposed with just 0.9 Million parameters.•A custom DeepCNN and LSTM-based model — ApneaNet, consisting of 13 layers taking efficiency into account.•Proposed Models demonstrated competitive results, achieving 90.87% on 70 data windows and 96.37% on 35 dataset windows.