National and Kapodistrian University of Athens / ETH Zurich (Pantelidis, Ruiperez-Campillo et al.)
Explainable Inception-style 1D CNN for multi-label arrhythmia detection from 12-lead ECGs, integrating Grad-CAM visualization to highlight the waveform segments driving each prediction. Trained on MIMIC-IV-ECG and externally validated on PTB-XL across atrial fibrillation, sinus tachycardia, conduction disturbances (RBBB/LBBB/LAFB), long QT, Wolff-Parkinson-White pattern, and paced-rhythm detection, with all metrics exceeding 90% internally and strong generalization on external validation.
Architecture
CNN (1D)
Custom Inception-style 1D CNN (three sequential Inception blocks, each with two Inception modules of parallel multi-kernel 1D convolutions) for multi-label 12-lead ECG classification, with Grad-CAM explainability
Framework
TensorFlow
Added to catalog
2026-08-10
CC BY 4.0
License for model weights only. Associated code may be licensed seperately, check code source for specific terms.
800,035 12-lead ECG-report pairs from 161,352 subjects at Beth Israel Deaconess Medical Center.
52% male / 48% female; age range 0-95 (median ~62). 21,837 10-second 12-lead ECG records.
Discrimination of atrial fibrillation vs. sinus tachycardia vs. non-arrhythmia from 12-lead ECG
Multi-label detection of conduction disturbances (RBBB/LBBB/LAFB), long QT, Wolff-Parkinson-White pattern, and paced rhythm