ECG Generalization with Morphology-Rhythm Disentanglement
Research Paper#Medical AI, ECG Analysis, Deep Learning🔬 Research|Analyzed: Jan 3, 2026 16:07•
Published: Dec 29, 2025 10:14
•1 min read
•ArXivAnalysis
This paper addresses the challenge of generalizing ECG classification across different datasets, a crucial problem for clinical deployment. The core idea is to disentangle morphological features and rhythm dynamics, which helps the model to be less sensitive to distribution shifts. The proposed ECG-RAMBA framework, combining MiniRocket, HRV, and a bi-directional Mamba backbone, shows promising results, especially in zero-shot transfer scenarios. The introduction of Power Mean pooling is also a notable contribution.
Key Takeaways
- •Proposes ECG-RAMBA, a framework for ECG classification that disentangles morphology and rhythm.
- •Employs MiniRocket for morphological features, HRV for rhythm descriptors, and a bi-directional Mamba backbone for long-range context.
- •Introduces Power Mean pooling to improve sensitivity to transient abnormalities.
- •Demonstrates strong performance in zero-shot transfer, outperforming baseline models.
Reference / Citation
View Original"ECG-RAMBA achieves a macro ROC-AUC ≈ 0.85 on the Chapman--Shaoxing dataset and attains PR-AUC = 0.708 for atrial fibrillation detection on the external CPSC-2021 dataset in zero-shot transfer."