HeartBeatAI: An Interpretable and Robust Deep Learning Framework for Multi-Label ECG Arrhythmia Detection
Researchers have developed HeartBeatAI, a deep learning framework designed to improve the accuracy and interpretability of multi-label ECG arrhythmia detection. The system integrates domain generalization techniques and multi-scale feature aggregation to capture subtle ECG anomalies. While achieving a high Macro F1-score of 98% on intra-source datasets, performance significantly degrades when tested on data from different institutions, indicating challenges for cross-institutional deployment. AI
IMPACT This framework could enhance diagnostic capabilities in healthcare by improving the accuracy and interpretability of ECG analysis for arrhythmia detection.