Combining residual U-Net and data augmentation for dense temporal segmentation of spike wave discharges in single-channel EEG
- 1School of Data Science, University of Virginia
- 2School of Medicine, University of Virginia
Abstract
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.
Related
- Event segmentation and encoding rely on information accumulated over different temporal scales
- Resurrecting host–microbiome interactions: combining resurrection ecology and microbiome transplants to test temporal local adaptation
- Spike-wave discharges reflect widespread, but non-uniform, cortical hypersynchrony during immobility
- EEGBind: Detecting Source-Level Interictal Epileptiform Discharges via EEG-Centric Multimodal Binding
- The deficiency of htr6 gene results in aberrant brain development and epileptiform discharges in zebrafish larvae.
BibTeX
@article{sengupta2026combining,
title = {{Combining residual U-Net and data augmentation for dense temporal segmentation of spike wave discharges in single-channel EEG}},
author = {Saurav Sengupta and Scott Kilianski and Suchetha Sharma and Sakina Lashkeri and Ashley McHugh and Mark Beenhakker and Donald E. Brown},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-62064-7},
url = {https://doi.org/10.1038/s41598-026-62064-7},
note = {CC BY},
}
RIS
TY - JOUR
TI - Combining residual U-Net and data augmentation for dense temporal segmentation of spike wave discharges in single-channel EEG
AU - Saurav Sengupta
AU - Scott Kilianski
AU - Suchetha Sharma
AU - Sakina Lashkeri
AU - Ashley McHugh
AU - Mark Beenhakker
AU - Donald E. Brown
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-62064-7
UR - https://doi.org/10.1038/s41598-026-62064-7
AB - 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.
SN - 2045-2322
ER -