ABC-GAN: An Attention-Based Conditional GAN with DTW Loss for ECG Signal Generation to Address Class Imbalance

Q: What is the clinical evidence from this study?

Population: Imbalanced ECG datasets (Arrhythmias). Intervention: ABC-GAN (Attention-Based Conditional Generative Adversarial Network with DTW Loss) vs. Raw imbalanced data, standard oversampling techniques, and vanilla cGAN. Primary outcome: Downstream classifier accuracy (accuracy of 95%).

Q: What does this research mean for the field?

Augmenting imbalanced ECG datasets with synthetic signals generated by an attention-based conditional GAN with Dynamic Time Warping loss enhances downstream arrhythmia classification accuracy to 95%, outperforming standard oversampling and baseline conditional GANs. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Parmida Behain; Noushin Riahi

doi:10.66224/jist.51397.14.54.79

September 9, 2026Journal of Information Systems and Telecommunication (JIST)Open Access

ABC-GAN: An Attention-Based Conditional GAN with DTW Loss for ECG Signal Generation to Address Class Imbalance

View Full Paper

Ask AI

Bookmark

Share

Key result

ABC-GAN synthetic ECG augmentation improves downstream classifier accuracy to ~95% over standard oversampling techniques.

accuracy of 95%

Why the study?

The imbalance between normal and pathological ECG cases degrades the performance of deep learning diagnostic systems, especially for minority classes.

Does data augmentation with ABC-GAN improve downstream classifier accuracy for imbalanced ECG datasets?

Key Points

To introduce ABC-GAN, an attention-based conditional generative adversarial network utilizing dynamic time warping loss to generate high-fidelity, class-specific ECG waveforms for mitigating dataset imbalance.
Designed a conditional generative adversarial network integrating an attention mechanism in the generator to target critical ECG morphological segments such as the P-wave, QRS complex, and T-wave.
Incorporated Dynamic Time Warping (DTW) into the generator loss function to preserve temporal dynamics and align waveforms against phase variations.
Trained and evaluated the architecture across seven heartbeat classes using the MIT-BIH Arrhythmia dataset, followed by downstream classification testing.
ABC-GAN generated class-specific synthetic ECG waveforms that closely replicated real signal distributions and preserved essential temporal dynamics.
A downstream classifier trained on ABC-GAN-augmented data achieved an accuracy of 95%, markedly exceeding the performance of models trained on raw imbalanced data.
The synthetic data augmentation strategy outperformed both traditional oversampling techniques and a standard vanilla conditional GAN (cGAN).

Structured PICO

Does data augmentation with ABC-GAN improve downstream classifier accuracy for imbalanced ECG datasets?

P

Population

Electrocardiogram (ECG) datasets, specifically the MIT-BIH Arrhythmia dataset across seven different heartbeat classes.

I

Intervention

Data augmentation using ABC-GAN (an Attention-Based Conditional Generative Adversarial Network with Dynamic Time Warping loss).

C

Comparator

Raw imbalanced data, standard oversampling techniques, and vanilla Conditional GAN (cGAN).

O

Outcome

Downstream classifier accuracy

Main Result

Effect estimate: accuracy of 95%

Implication

ABC-GAN effectively mitigates data imbalance in ECG datasets by generating high-fidelity, class-specific synthetic signals, substantially improving diagnostic classification accuracy.

Abstract

The imbalance between normal and pathological cases in Electrocardiogram (ECG) datasets significantly degrades the performance of automated deep learning-based diagnostic systems, particularly for minority classes.This paper introduces ABC-GAN, a novel Attention-Based Conditional Generative Adversarial Network designed to mitigate this critical data imbalance by generating high-fidelity, class-specific synthetic ECG signals.Our proposed model incorporates two key innovations: an attention mechanism within the generator to focus on critical morphological features of the ECG waveform, and the integration of Dynamic Time Warping (DTW) as a distance metric in the generator's loss function to better preserve essential temporal dynamics.Trained and thoroughly evaluated on the MIT-BIH Arrhythmia dataset across seven different heartbeat classes, ABC-GAN successfully generates highly realistic signals that closely match the original data's distribution.When used to augment the training set, these synthetic signals significantly enhance classifier performance and generalization.A downstream classifier achieved an accuracy of 95%, representing a substantial improvement over the baseline trained on the raw imbalanced data and clearly outperforming both standard oversampling techniques and a vanilla Conditional GAN (cGAN).The overall findings demonstrate that ABC-GAN is a powerful and effective tool for data augmentation, fully capable of improving diagnostic accuracy in cardiac research and practical clinical applications.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Behain et al. (2026) studied Imbalanced ECG datasets (Arrhythmias). ABC-GAN (Attention-Based Conditional Generative Adversarial Network with DTW Loss) vs. Raw imbalanced data, standard oversampling techniques, and vanilla cGAN was evaluated on Downstream classifier accuracy (accuracy of 95%). Data augmentation using ABC-GAN generated high-fidelity synthetic ECG signals that improved a downstream classifier's accuracy to 95%, outperforming standard oversampling techniques.

synapsesocial.com/papers/6ac986baa26deea15a5120f8 https://doi.org/10.66224/jist.51397.14.54.79

View Full Paper

Ask AI

Bookmark

Share

ABC-GAN: An Attention-Based Conditional GAN with DTW Loss for ECG Signal Generation to Address Class Imbalance

Key Points

Structured PICO

Main Result

Implication

Abstract

Discussion

Cite This Study

ABC-GAN: An Attention-Based Conditional GAN with DTW Loss for ECG Signal Generation to Address Class Imbalance

Key Points

Structured PICO

Main Result

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading