Review of Deep Learning-Based Atrial Fibrillation Detection Studies

Abstract

Atrial fibrillation (AF) is a common arrhythmia that can lead to stroke, heart failure, and premature death. Manual screening of AF on electrocardiography (ECG) is time-consuming and prone to errors. To overcome these limitations, computer-aided diagnosis systems are developed using artificial intelligence techniques for automated detection of AF. Various machine learning and deep learning (DL) techniques have been developed for the automated detection of AF. In this review, we focused on the automated AF detection models developed using DL techniques. Twenty-four relevant articles published in international journals were reviewed. DL models based on deep neural network, convolutional neural network (CNN), recurrent neural network, long short-term memory, and hybrid structures were discussed. Our analysis showed that the majority of the studies used CNN models, which yielded the highest detection performance using ECG and heart rate variability signals. Details of the ECG databases used in the studies, performance metrics of the various models deployed, associated advantages and limitations, as well as proposed future work were summarized and discussed. This review paper serves as a useful resource for the researchers interested in developing innovative computer-assisted ECG-based DL approaches for AF detection.

Convolutional Neural Network-Based ECG-Assisted Diagnosis for Coal Workers

“…As a branch of machine learning, deep learning [34], has achieved great success in determining abnormal ECGs and ECG waveforms, which greatly improving the accuracy of diagnosis. Fatma Murat [35] al. reviewed 24 relevant articles published in international journals and found that most of the studies used the CNN model and achieved good results.…”

Atrial fibrillation prediction by combining ECG markers and CMR radiomics

“…The second lead was used to extract the features as both old devices and the wearable devices are using a single lead. According to literature, the second lead provides the most valuable information 55 , 56 including P, QRS and T waves. For that reason, it is the most used within the single-lead ECG works and the one with better results from 12-lead ECG recordings 57 .…”

Golden Standard or Obsolete Method? Review of ECG Applications in Clinical and Experimental Context

“…The most challenging recent application using HRV features is atrial fibrillation (AFIB) detection (e.g., Smisek et al, 2018 ; Murat et al, 2021a ). Timely prediction of paroxysmal AFIB episodes using seven novel Poincaré map features achieves the accuracy over 86% for different ML models and even higher accuracy (98%) when combining with standard HRV features ( Parsi et al, 2021b ).…”

“…Many challenge participants applied the CNN or LSTM model to address the topic and achieved the best performance in binary (AFIB vs. non-AFIB) and multi-class (sinus rhythm, AFIB, atrial flutter, etc.) issues ( Murat et al, 2021a ). The growing trend in using DL is still present due to international challenges (PhysioNet/Computing in Cardiology 2020 and 2021, China Physiological Signal Challenge 2018–2021) focused on the algorithms for reliable QRS detection, supraventricular and ventricular premature contraction detection, AFIB detection or paroxysmal AFIB localization in ECG, multi-label classification of 24 different arrhythmias, or identification of the ECG leads with the highest discrimination ability.…”