XAI-ECGNet enables interpretable ECG multi-label classification

It achieves macro-AUROC of 0.9337, outperforming all baselines with four diagnostic classes exceeding 0.90 AUROC. Ablation shows SE attention is critical for BiLSTM gains, and post-hoc methods confirm clinically consistent feature learning. The model demonstrates genuine generalization on CPSC 2018 for NORM and CD, with interpretable label mismatch observed in STTC.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.