Volume 15, Issue 6 And S7 (Artificial Intelligence 2025)
J Research Health 2025, 15(6 And S7): 779-792 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Barzegar M M, Ahmadi Daryakenari N, Khodatars M. Explainable Epileptic Seizure Detection from Electroencephalography Signals via CNN–Bi-LSTM Attention Hybrid Model. J Research Health 2025; 15 (6) :779-792
URL: http://jrh.gmu.ac.ir/article-1-2987-en.html
URL: http://jrh.gmu.ac.ir/article-1-2987-en.html
1- Department of Biomedical Engineering, School of Electrical Engineering, Iran University of Science and Technology (IUST), Tehran, Iran.
2- Department of Epileptology, University Hospital Bonn, University of Bonn, Bonn, Germany.
3- Department of Medical Engineering, MMS.C. Islamic Azad University, Mashhad, Iran. ,khodatars1marjane@gmail.com
2- Department of Epileptology, University Hospital Bonn, University of Bonn, Bonn, Germany.
3- Department of Medical Engineering, MMS.C. Islamic Azad University, Mashhad, Iran. ,
Abstract: (1689 Views)
Background: Epilepsy is a chronic neurological disorder marked by recurrent daily seizures that threaten patient safety. Electroencephalography (EEG) is a crucial neuroimaging tool for epilepsy diagnosis, but manual interpretation of EEG signals is challenging for clinicians. To assist specialists, automated systems, such as computer-aided diagnosis systems (CADS) based on deep learning (DL) are essential.
Methods: The proposed CADS system was validated using the Turkish epilepsy dataset. In preprocessing, EEG signals were filtered, down-sampled, re-referenced using common average reference (CAR), and segmented into multiple temporal windows. A new feature extraction framework combining one-dimensional convolutional neural networks (1D-CNN), bidirectional long short-term memory (Bi-LSTM), and an attention mechanism was developed. All experiments were performed using 5-fold cross-validation. Post-hoc explainability was evaluated using explainable artificial intelligence (XAI) techniques, including t-distributed stochastic neighbor embedding (t-SNE) and shapley additive explanations (SHAP).
Results: The proposed CADS achieved a seizure diagnosis accuracy of 99.49%, demonstrating high robustness across the validation folds, with minimal variance between folds (±0.12%). Feature space visualization confirmed clear class separation, and SHAP analysis provided clinically meaningful explanations for model decisions.
Conclusion: The proposed DL architecture shows strong potential for reliable and interpretable automatic epileptic seizure detection from EEG. This CADS can significantly reduce the diagnostic burden on clinicians and support real-time decision-making in clinical environments.
Methods: The proposed CADS system was validated using the Turkish epilepsy dataset. In preprocessing, EEG signals were filtered, down-sampled, re-referenced using common average reference (CAR), and segmented into multiple temporal windows. A new feature extraction framework combining one-dimensional convolutional neural networks (1D-CNN), bidirectional long short-term memory (Bi-LSTM), and an attention mechanism was developed. All experiments were performed using 5-fold cross-validation. Post-hoc explainability was evaluated using explainable artificial intelligence (XAI) techniques, including t-distributed stochastic neighbor embedding (t-SNE) and shapley additive explanations (SHAP).
Results: The proposed CADS achieved a seizure diagnosis accuracy of 99.49%, demonstrating high robustness across the validation folds, with minimal variance between folds (±0.12%). Feature space visualization confirmed clear class separation, and SHAP analysis provided clinically meaningful explanations for model decisions.
Conclusion: The proposed DL architecture shows strong potential for reliable and interpretable automatic epileptic seizure detection from EEG. This CADS can significantly reduce the diagnostic burden on clinicians and support real-time decision-making in clinical environments.
Type of Study: Orginal Article |
Subject:
● Artificial Intelligence
Received: 2025/11/20 | Accepted: 2025/11/29 | Published: 2025/12/31
Received: 2025/11/20 | Accepted: 2025/11/29 | Published: 2025/12/31
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC-By-NC), which permits use, distribution, and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
Contact Information
