University of Malaga / EPFL (Atienza Lab)
Self-supervised deep learning system that detects ECG anomalies indirectly, by learning to forecast what a normal ECG signal should look like next. FADE is trained only on normal ECG segments using a novel morphology-inspired loss function; at inference time, a large mismatch between the forecast and the observed signal flags an anomaly, avoiding the need for labeled abnormal-beat datasets. Evaluated on the public MIT-BIH NSR and MIT-BIH Arrhythmia databases, FADE reached an average accuracy of 83.84% for anomaly detection and 85.46% for correctly classifying normal ECG, and the approach can be adapted to new recording contexts via domain adaptation.
Architecture
Other
Self-supervised sequence-forecasting network trained with a novel morphology-inspired loss to predict the near-future waveform of a normal ECG; large forecast error signals an anomaly (specific layer architecture not detailed in available abstracts)
Framework
PyTorch
Added to catalog
2026-08-13
48 half-hour excerpts of two-channel ambulatory ECG recordings from 47 subjects studied by the BIH Arrhythmia Laboratory; used as a cross-dataset, lead-missing/noise-robustness benchmark distinct from the pretraining data.
MIT-BIH Normal Sinus Rhythm Database
Binary flag distinguishing normal vs. anomalous ECG segments, derived from waveform-forecasting error