Combining residual U-Net and data augmentation for dense temporal segmentation of spike wave discharges in single-channel EEG.

Manual annotation of spike-wave discharges (SWDs), the electrographic hallmark of absence seizures, is labor-intensive for long-term electroencephalography (EEG) monitoring studies. While machine learning approaches show promise for automated detection, they often struggle with cross-subject generalization due to high inter-individual variability in seizure morphology and signal characteristics. In this study we compare the performance of 16 machine learning classifiers on our own manually annotated dataset of 961 hours of EEG recordings from C3H/HeJ mice, including 22,637 labeled SWDs, and find that a 1D U-Net performs best. We then improve its performance by employing residual connections and data augmentation strategies combining amplitude scaling, Gaussian noise injection, and signal inversion to enhance cross-subject generalization. Our proposed model, AugUNet1D, achieves an average F1-score of 0.90 with balanced precision (0.91) and recall (0.90), representing a 29% relative improvement over the "Twin Peaks" algorithmic baseline and exceptional cross-subject performance. AugUNet1D, pretrained on our manually annotated data, along with the dataset itself, is made public for other users.