Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EB-GAME: A Game-Changer in ECG Heartbeat Anomaly Detection

JuneYoung Park, Da Young Kim, Yunsoo Kim, Jisu Yoo, Tae Joon Kim

TL;DR

The paper tackles ECG beat-level anomaly detection under data imbalance by proposing EB-GAME, a GAN-guided masked autoencoder that learns normal ECG patterns using only normal data. It introduces a novel wave-focused masking strategy and an MAE-based Generator–Discriminator setup, trained with a composite loss that includes reconstruction, adversarial, and contextual terms. On the MIT-BIH Arrhythmia dataset, EB-GAME achieves state-of-the-art metrics (AUROC ≈ 0.97, accuracy ≈ 0.97, high sensitivity and specificity) and outperforms several existing GAN-based ECG anomaly detectors. This unsupervised framework reduces dependence on abnormal-label availability and can generalize to other ECG datasets and signal anomaly detection tasks, enhancing clinically practical continuous monitoring.

Abstract

Cardiologists use electrocardiograms (ECG) for the detection of arrhythmias. However, continuous monitoring of ECG signals to detect cardiac abnormal-ities requires significant time and human resources. As a result, several deep learning studies have been conducted in advance for the automatic detection of arrhythmia. These models show relatively high performance in supervised learning, but are not applicable in cases with few training examples. This is because abnormal ECG data is scarce compared to normal data in most real-world clinical settings. Therefore, in this study, GAN-based anomaly detec-tion, i.e., unsupervised learning, was employed to address the issue of data imbalance. This paper focuses on detecting abnormal signals in electrocardi-ograms (ECGs) using only labels from normal signals as training data. In-spired by self-supervised vision transformers, which learn by dividing images into patches, and masked auto-encoders, known for their effectiveness in patch reconstruction and solving information redundancy, we introduce the ECG Heartbeat Anomaly Detection model, EB-GAME. EB-GAME was trained and validated on the MIT-BIH Arrhythmia Dataset, where it achieved state-of-the-art performance on this benchmark.

EB-GAME: A Game-Changer in ECG Heartbeat Anomaly Detection

TL;DR

The paper tackles ECG beat-level anomaly detection under data imbalance by proposing EB-GAME, a GAN-guided masked autoencoder that learns normal ECG patterns using only normal data. It introduces a novel wave-focused masking strategy and an MAE-based Generator–Discriminator setup, trained with a composite loss that includes reconstruction, adversarial, and contextual terms. On the MIT-BIH Arrhythmia dataset, EB-GAME achieves state-of-the-art metrics (AUROC ≈ 0.97, accuracy ≈ 0.97, high sensitivity and specificity) and outperforms several existing GAN-based ECG anomaly detectors. This unsupervised framework reduces dependence on abnormal-label availability and can generalize to other ECG datasets and signal anomaly detection tasks, enhancing clinically practical continuous monitoring.

Abstract

Cardiologists use electrocardiograms (ECG) for the detection of arrhythmias. However, continuous monitoring of ECG signals to detect cardiac abnormal-ities requires significant time and human resources. As a result, several deep learning studies have been conducted in advance for the automatic detection of arrhythmia. These models show relatively high performance in supervised learning, but are not applicable in cases with few training examples. This is because abnormal ECG data is scarce compared to normal data in most real-world clinical settings. Therefore, in this study, GAN-based anomaly detec-tion, i.e., unsupervised learning, was employed to address the issue of data imbalance. This paper focuses on detecting abnormal signals in electrocardi-ograms (ECGs) using only labels from normal signals as training data. In-spired by self-supervised vision transformers, which learn by dividing images into patches, and masked auto-encoders, known for their effectiveness in patch reconstruction and solving information redundancy, we introduce the ECG Heartbeat Anomaly Detection model, EB-GAME. EB-GAME was trained and validated on the MIT-BIH Arrhythmia Dataset, where it achieved state-of-the-art performance on this benchmark.
Paper Structure (18 sections, 10 equations, 1 figure, 2 tables)

This paper contains 18 sections, 10 equations, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Preliminary
  4. MAE.
  5. GAN-based Anomaly Detection.
  6. Let's GAME Change
  7. Wave Masking Strategy.
  8. Generator.
  9. Discriminator.
  10. Training objective.
  11. Masked Autoencoder Loss.
  12. Adversarial Loss.
  13. Contextual Loss.
  14. Experimental Results
  15. Datasets
  16. ...and 3 more sections

Figures (1)

  • Figure 1: Overview of the EB-GAME framework. The generator employs the MAE architecture, and through adversarial training with the discriminator, it is trained and performs anomaly detection.