CoRe-ECG: Advancing Self-Supervised Representation Learning for 12-Lead ECG via Contrastive and Reconstructive Synergy
Summary: arXiv:2604.11359v1 Announce Type: new
Abstract
Accurate interpretation of electrocardiogram (ECG) remains challenging due to the scarcity of labeled data and the high cost of expert annotation. Self-supervised learning (SSL) offers a promising solution by enabling models to learn expressive representations from unlabeled signals. Existing ECG SSL methods typically rely on either contrastive learning or reconstructive learning. However, each approach in isolation provides limited supervisory signals and suffers from additional limitations, including non-physiological distortions introduced by naive augmentations and trivial correlations across multiple leads that models may exploit as shortcuts.
Introduction
In recent years, there has been a growing interest in the application of self-supervised learning techniques to medical data, particularly in the realm of ECG analysis. The CoRe-ECG framework aims to address the limitations of existing methods by combining the strengths of contrastive and reconstructive learning paradigms. This innovative approach not only enhances the model’s ability to learn from unlabeled data but also improves the robustness of the learned representations.
Key Features of CoRe-ECG
- Unified Learning Paradigm: CoRe-ECG establishes a synergistic interaction between global semantic modeling and local structural learning.
- Global Representation Alignment: The framework aligns global representations during reconstruction, enabling instance-level discriminative signals to guide local waveform recovery.
- Frequency Dynamic Augmentation (FDA): FDA adaptively perturbs ECG signals based on their frequency-domain importance, enriching the learning process.
- Spatio-Temporal Dual Masking (STDM): STDM breaks linear dependencies across leads, increasing the challenge of reconstructive tasks and enhancing model resilience.
Performance and Results
The CoRe-ECG method has demonstrated state-of-the-art performance across multiple downstream ECG datasets, surpassing previous approaches in accuracy and reliability. Extensive ablation studies have been conducted to validate the necessity and complementarity of each component, confirming that the integrated approach significantly contributes to the overall effectiveness of the model.
Conclusion
CoRe-ECG represents a significant advancement in self-supervised representation learning for ECG data. By leveraging both contrastive and reconstructive learning strategies, this framework not only addresses the challenges posed by the scarcity of labeled data but also enhances the physiological relevance of the learned representations. As the field of medical AI continues to evolve, approaches like CoRe-ECG will play a crucial role in improving diagnostic accuracy and patient outcomes through more effective ECG analysis.