HRVConformer: Neonatal Hypoxic-Ischemic Encephalopathy Classification from the Heart Rate signals

In three linesHRVConformer is a hybrid Convolution-Transformer architecture for classifying neonatal hypoxic-ischemic encephalopathy from raw heart rate signals. Trained on 1,573 epochs (259 expert-annotated, rest weakly labelled), the model achieves 83.23% AUC and 74.56% accuracy on a 215-hour test set, outperforming ResNet50 and Transformer baselines.

Summary generated by Claude — human-verified