EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Interpretable Deep Learning for Alzheimer’s Detection: EEG Meets Transformers and Multi-Resolution Fusion

Nandini Kumari, Neha Prerna Tigga, Fady Alnajjar

Procedia Computer Science · 2026 · 283 · 4117-4126

DOI: 10.1016/j.procs.2026.06.472 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Procedia Computer Science, 2026. Verfasst von Nandini Kumari und 2 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid and accurate diagnosis of the Alzheimer Disease (AD) is a significant challenge to the clinical field. Non-invasive and therefore less expensive, electroencephalogram (EEG) signals have a lot of potential in AD classification, yet class imbalance and their minimal interpretability of deep learning models are limiting. In the present research, we are proposing a framework that combines data augmentation, sophisticated deep learning structures and explainable AI to enhance the way classification works. The pre-processed EEG signals were augmented with CycleGAN to equitably compare the minority classes of the Frontotemporal Dementia (FTD) and Healthy Controls (HC) with the AD. For EEG-based AD classification two models were proposed and tested, a Temporal-Spatial Transformer Network (TSTN) and a Multi-Resolution Spatio-Temporal Fusion Network (MRSTFNet). Experimental findings indicated that TSTN was always better than MRSTFNet and reached a maximum accuracy of 86.68% with an equal balance between precision, recall, and F1-score, which proves its strength in detecting AD. To make models transparent, LIME-based explanation was used, showing that frontal and temporal EEG regions played the largest role in predicting the AD, whereas central and occipital regions exhibited counterbalance. In general, the suggested framework has high predictive power, tackles the problem of class imbalance, and improves interpretability enough to be a proof-of-concept methodological contribution to clinical decision support in AD diagnosis.

Zitieren

APA

Kumari, N., Tigga, N. P., & Alnajjar, F. (2026). Interpretable Deep Learning for Alzheimer’s Detection: EEG Meets Transformers and Multi-Resolution Fusion. Procedia Computer Science, 283, 4117-4126. https://doi.org/10.1016/j.procs.2026.06.472

Vancouver

Kumari N, Tigga NP, Alnajjar F. Interpretable Deep Learning for Alzheimer’s Detection: EEG Meets Transformers and Multi-Resolution Fusion. Procedia Computer Science. 2026;283:4117-4126. doi:10.1016/j.procs.2026.06.472

BibTeX
@article{Kumari2026,
  author = {Kumari, Nandini and Tigga, Neha Prerna and Alnajjar, Fady},
  title = {Interpretable Deep Learning for Alzheimer’s Detection: EEG Meets Transformers and Multi-Resolution Fusion},
  journal = {Procedia Computer Science},
  year = {2026},
  volume = {283},
  pages = {4117--4126},
  doi = {10.1016/j.procs.2026.06.472}
}
Kostenlos starten →