CardioMix: Semi-Supervised ECG Segmentation via Cardiac Pattern-Guided Bidirectional Fusion
A new framework called CardioMix has been developed by researchers for semi-supervised segmentation of electrocardiograms (ECG), utilizing a bidirectional CutMix strategy informed by cardiac patterns. This innovative method tackles the issue of scarce annotated data by enhancing the labeled dataset with realistic variations derived from unlabeled data, while simultaneously providing more robust supervisory signals to the unlabeled samples. As a result, all augmented instances retain physiological relevance. CardioMix is created to function as a plug-and-play module, ensuring excellent compatibility. Further details can be found in arXiv:2605.15722.
Key facts
- CardioMix is a framework for semi-supervised ECG segmentation.
- It uses a bidirectional CutMix strategy guided by cardiac patterns.
- The framework enriches labeled data with variations from unlabeled data.
- It applies stronger supervisory signals to unlabeled data.
- All augmented samples remain physiologically meaningful.
- CardioMix is designed as a plug-and-play module.
- The research is published on arXiv with ID 2605.15722.
- The method addresses scarcity of annotated ECG data.
Entities
Institutions
- arXiv