Detection of Paroxysmal Atrial Fibrillation from Dynamic ECG Recordings Based on a Deep Learning Model

Abstract

Background and Objective: Atrial fibrillation (AF) is one of the most common arrhythmias clinically. Aging tends to increase the risk of AF, which also increases burden of other comorbidities including coronary artery disease (CAD), and even heart failure (HF). The precise detection of AF is a challenge due to its intermittence and unpredictability. Accurate detection of AF is still in need.

Methods: In the paper, we used a deep learning model to detect atrial fibrillation. Here, distinction was not made between AF and atrial flutter (AFL), both of which can be classified into atrial tachycardia and manifest a resemble pattern on electrocardiogram (ECG). Not only the method discriminated AF from normal, but also detected the onsets and offsets. The proposed model involved residual blocks and the Transformer. Datasets used for training was from CPSC2021 Challenge, collected by dynamic ECG devices.

Results: Tests on four public datasets validated the accuracy and availability of the proposed method. The best performance for AF rhythm detection attained an accuracy of 97.61%, a sensitivity of 97.76%, and a specificity of 97.52%, respectively. The best performance for onsets and offsets detection obtained a sensitivity of 95.90% and 87.70%.

Conclusions: The model had a great capacity to discriminate AF from normal and to detect the onsets and offsets. We visualized the features in heatmap and illustrated the interpretability. The model exactly focused on the ECG waveform which showed an obvious characteristic of AF.

Note:

Funding Information: This work was partially supported by the National Natural Science Foundation of China (Grant Nos. 61971089 and 61471081), Dalian Science and Technology Innovation Fund (2021JJ12WZ47), the Open Research Fund of State Key Laboratory of Bioelectronics, Southeast University and the National Key R&D Program of the Ministry of Science and Technology of China (2020YFC2004400).

Declaration of Interests: The authors declare no conflict of interests.

Keywords: Atrial fibrillation, Electrocardiogram, Deep Learning, the Transformer