Deep learning-based prediction of atrial fibrillation from polar transformed time-frequency electrocardiogram
Daehyun Kwon, Hanbit Kang, Dongwoo Lee, Yoon-Chul Kim
Abstract
Portable and wearable electrocardiogram (ECG) devices are increasingly utilized in healthcare for monitoring heart rhythms and detecting cardiac arrhythmias or other heart conditions. The integration of ECG signal visualization with AI-based abnormality detection empowers users to independently and confidently assess their physiological signals.
Introduction
Portable and wearable electrocardiogram (ECG) devices are increasingly utilized in healthcare for monitoring heart rhythms and detecting cardiac arrhythmias or other heart conditions [1]. These devices typically record single-lead ECG signals, which, while less comprehensive than the standard 12-lead ECG used in clinical settings.
Methods
This section outlines the ECG dataset used in our study, the steps for ECG signal preprocessing and polar transformation, and the development and validation of the deep CNN models.
Results
This section presents a qualitative comparison between polar transformed images and our proposed reverse polar transformed images, as well as visualization results for raw and P-T preprocessed ECG signals.
Discussion
To the best of our knowledge, this study presents the first demonstration of reverse polar- transformed spectrograms used as input to a deep CNN model for detecting atrial fibrillation (Afib). We focused on visualizing a 30-second ECG spectrogram in a polar representation.
Conclusions
This study introduces a novel reverse polar transformed spectrogram for visualizing 30-second ECG signals, which aids in the detection of atrial fibrillation (Afib). The reverse polar transformation enhances the visualization of cardiac arrhythmias by emphasizing irregular spacing between peaks at the periphery of the polar spectrogram.
Citation: Kwon D, Kang H, Lee D, Kim Y-C (2025) Deep learning-based prediction of atrial fibrillation from polar transformed time-frequency electrocardiogram. PLoS ONE 20(3): e0317630. https://doi.org/10.1371/journal.pone.0317630
Editor: Hirenkumar Kantilal Mewada, Prince Mohammad Bin Fahd University, SAUDI ARABIA
Received: June 7, 2024; Accepted: December 31, 2024; Published: March 10, 2025
Copyright: © 2025 Kwon et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data are publicly available from the PhysioNet/CinC Challenge 2017 database (https://physionet.org/content/challenge-2017/1.0.0/).
Funding: “This study was supported by the “Regional Innovation Strategy (RIS)” through the National Research Foundation of Korea (NRF) and funded by the Ministry of Education (MOE) (2022RIS-005).”
Competing interests: The authors have declared that no competing interests exist.
