Classification of Obstructive Sleep Apnoea from single-lead ECG signals using convolutional neural and Long Short Term Memory networks

A Signal Segmentation‐Free Model for Electrocardiogram‐Based Obstructive Sleep Apnea Severity Classification

“…The presented model works as follows. First, feature maps of an input ECG signal were extracted in the Stem module using 2 identical convolutional layers, where filters ¼ 32, kernel size ¼ 27, stride ¼ 14, each abbreviated as Conv1D, (32,27,14). Next, high-level feature maps were extracted in the Body module using 3 cascaded CSPNet blocks, designated as CSPNet_Blocks.…”

“…As referenced previously, a single type of signal was used in ref., [7–21] the majority of which took ECG signals as input, and mostly used deep learning or convolutional neural network (CNN)‐based models to classify the OSA severity. ECG signals were directly used to train models and then classify the OSA severity in ref., [7–11] ECG‐derived respiration (EDR) signals or RR intervals, defined as the intervals between the R peaks of successive QRS complexes, were used in ref., [12–15] and alternatively spectral features of ECG signals were used in ref. [16–18]…”

“…For this sake, considerable efforts have been made to reduce the number of the required signals in a large volume of publications. [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23] For example, the OSA severity was classified using ECG [7][8][9][10][11][12][13][14][15][16][17][18] or SpO 2 signals alone, [19][20][21] using a combination of ECG and SpO 2 signals, [22] or even using multiple types of signals, i.e., ECG, SpO 2 , chest, and abdominal respiratory movement signals. [23] Recently, wearable and portable tools at home for OSA classification have been developed as an easy-to-use alternative to PSG.…”

“…[23] Recently, wearable and portable tools at home for OSA classification have been developed as an easy-to-use alternative to PSG. [24][25][26] As referenced previously, a single type of signal was used in ref., [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] the majority of which took ECG signals as input, and Obstructive sleep apnea (OSA) has been a common sleep disorder for years, and polysomnography (PSG) remains the gold standard for diagnosing OSA. Nevertheless, PSG is a time and money consuming test, and patients have to wait long for arranging a PSG test in a hospital.…”

“…mostly used deep learning or convolutional neural network (CNN)-based models to classify the OSA severity. ECG signals were directly used to train models and then classify the OSA severity in ref., [7][8][9][10][11] ECG-derived respiration (EDR) signals or RR intervals, defined as the intervals between the R peaks of successive QRS complexes, were used in ref., [12][13][14][15] and alternatively spectral features of ECG signals were used in ref. [16][17][18] Most of the above-referred deep learning models were trained using the MIT PhysioNet Apnea-ECG database.…”