An ECG diagnostic classifier (PTB-XL), running in your browser
A 1-D convolutional network that classifies 12-lead ECGs into the five PTB-XL diagnostic superclasses, reproducing the benchmark of Strodthoff et al. (2021). Click a recording — it is classified live in your browser (ONNX Runtime Web).
Open PhysioNet data — PTB-XL (Wagner et al., Sci. Data 2020, CC-BY 4.0). Code: github.com/mlavanga/ecg-ptbxl.
Learned ECG feature space
loading model…900 PTB-XL test ECGs in the model's penultimate feature space (2-D), coloured by diagnosis. Click a recording → classified live in your browser.
Normal ECGMyocardial InfarctionST/T ChangeConduction DisturbanceHypertrophy
PTB-XL superdiagnostic benchmark
Reproducing Strodthoff et al. (2021), official test fold.
0.916
this model · macro-AUROC
0.928
paper
Normal ECG0.939
Myocardial Infarction0.905
ST/T Change0.927
Conduction Disturbance0.917
Hypertrophy0.894
Sample ECGs
Real PTB-XL test recordings (lead II shown).