Fig. 1: ECG-CLIP diagnostic performance across cardiac conditions comparing internal testing and external validation.

a area under receiver operating characteristic curve (AUROC), b area under precision-recall curve (AUPRC), c accuracy, d sensitivity, e specificity, and f F1 score. In all panels, blue lines represent internal testing performance on the MIMIC-IV-ECG dataset, while red lines represent external validation performance on the LSAS-ECG dataset. Each radar plot displays 14 cardiac conditions arranged circumferentially: AF atrial fibrillation, SVT supraventricular tachycardia, ST sinus tachycardia, SB sinus bradycardia, PVC premature ventricular complexes, PAC premature atrial complexes, AVB atrioventricular block, 1dAVb first-degree atrioventricular block, RBBB right bundle branch block, LBBB left bundle branch block, LVH left ventricular hypertrophy, STTC ST-T changes, PQTI prolonged QT interval, and LowQRS low QRS voltages. Performance values range from 0 to 1.0 on the radial axis, with higher values indicating better performance.