Abstract
Considering the variability of amplitude and phase patterns inelectrocardiogram (ECG) signals due to cardiac activity and individualdifferences, existing entropy-based studies have not fully utilized these twopatterns and lack integration. To address this gap, this paper proposes a novelfusion entropy metric, morphological ECG entropy (MEE) for the first time,specifically designed for ECG morphology, to comprehensively describe thefusion of amplitude and phase patterns. MEE is computed based on beat-levelsamples, enabling detailed analysis of each cardiac cycle. Experimental resultsdemonstrate that MEE achieves rapid, accurate, and label-free localization ofabnormal ECG arrhythmia regions. Furthermore, MEE provides a method forassessing sample diversity, facilitating compression of imbalanced trainingsets (via representative sample selection), and outperforms random pruning.Additionally, MEE exhibits the ability to describe areas of poor quality. Bydiscussing, it proves the robustness of MEE value calculation to noiseinterference and its low computational complexity. Finally, we integrate thismethod into a clinical interactive interface to provide a more convenient andintuitive user experience. These findings indicate that MEE serves as avaluable clinical descriptor for ECG characterization. The implementation codecan be referenced at the following link:https://github.com/fdu-harry/ECG-MEE-metric.