Toward Trustworthy EEG-Based Epileptic Seizure Classification: A Hybrid Explainable AI Framework with Statistical Validation
Artificial Intelligence and Applications · 20.10.2026
DOI: 10.47852/bonviewaia620211142 ↗Kurz gefasst
Fachartikel in der Zeitschrift Artificial Intelligence and Applications, online erschienen am 20.10.2026. Verfasst von Mohamed Al Saidi und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Automated classification of epileptic seizure activity from electroencephalography (EEG) signals remains challenging because of the nonstationary nature of EEG and the need for reliable and interpretable clinical decision-support systems. Although recent machine learning and deep learning approaches have reported promising performance, many provide limited statistical validation and model explainability. This study presents a methodologically integrated hybrid artificial intelligence framework for binary EEG seizure classification using 591 ictal and non-seizure segments from the patient-nonspecific Meta-EEG derivative of the CHB-MIT dataset. The framework combines a one-dimensional convolutional neural network (CNN), trained on standardized multichannel EEG signals, with an XGBoost classifier trained on statistical features extracted using level-4 db4 wavelet packet decomposition (WPD). The branch probabilities were combined using validation-selected equal-weight fusion, with a weight of 0.5 assigned to each branch. SHapley Additive exPlanations and Integrated Gradients were used to interpret the XGBoost-WPD and CNN branches, respectively. Statistical evaluation included permutation testing, bootstrap confidence intervals, exact binomial testing, and paired McNemar comparisons. On the held-out test set, the hybrid model achieved 94.38% accuracy, 93.31% balanced accuracy, 90.00% sensitivity, 96.61% specificity, 91.53% F1-score, and 98.19% Receiver Operator Characteristics- Area Under Curve (ROC-AUC). McNemar’s test showed a significant difference from the standalone CNN (p = 0.0117), but not from XGBoost-WPD or InceptionTimeMini. These findings support the potential value of complementary temporal and time–frequency representations while highlighting the need for cross-validation and external evaluation. Received: 29 June 2026 | Revised: 20 September 2026 | Accepted: 29 September 2026 Conflicts of Interest Noushath Shaffi is the Associate Editor for Artificial Intelligence and Applications, and he was not involved in the editorial review or the decision to publish this article. The authors declare that they have no conflicts of interest in this work. Data Availability Statement The data that support the findings of this study are openly available in Zenodo at https://doi.org/10.5281/zenodo.6062372, reference number [9]. Author Contribution Statement Mohamed Al Saidi: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization. Noushath Shaffi: Conceptualization, Methodology, Validation, Formal analysis, Writing – original draft, Writing – review & editing, Supervision, Project administration. Vimbi Viswan: Validation, Writing – review & editing, Supervision. Abdelhamid Abdesselam: Methodology, Writing – review & editing, Supervision. Rachid Hedjam: Writing – review & editing, Supervision.
Zitieren
APA
Saidi, M. A., Shaffi, N., Viswan, V., Abdesselam, A., & Hedjam, R. (2026). Toward Trustworthy EEG-Based Epileptic Seizure Classification: A Hybrid Explainable AI Framework with Statistical Validation. Artificial Intelligence and Applications. https://doi.org/10.47852/bonviewaia620211142
Vancouver
Saidi MA, Shaffi N, Viswan V, Abdesselam A, Hedjam R. Toward Trustworthy EEG-Based Epileptic Seizure Classification: A Hybrid Explainable AI Framework with Statistical Validation. Artificial Intelligence and Applications. 2026. doi:10.47852/bonviewaia620211142
BibTeX
@article{Saidi2026,
author = {Saidi, Mohamed Al and Shaffi, Noushath and Viswan, Vimbi and Abdesselam, Abdelhamid and Hedjam, Rachid},
title = {Toward Trustworthy EEG-Based Epileptic Seizure Classification: A Hybrid Explainable AI Framework with Statistical Validation},
journal = {Artificial Intelligence and Applications},
year = {2026},
doi = {10.47852/bonviewaia620211142}
}
EUVIMED