A hybrid deep learning architecture for patient-independent epileptic seizure identification that integrates a Hierarchical Temporal Separable Convolutional Network, a Dual-Stage Bidirectional Recurrent Neural Network, and a Multi-Head Attention Mechanism is proposed, enabling effective extraction of spatial, temporal, and contextual features from non-stationary EEG signals while addressing inter-patient variability.
There are approximately 50 million people worldwide living with epilepsy, highlighting the need for strong early warning systems to enable prompt clinical management. Despite the high performance of recently developed deep learning models based on Convolutional Neural Networks (CNNs) and Bidirectional Long Short-Term Memory (Bi-LSTM/GRU) networks, they are highly sensitive to ictal features and are not very effective in detecting the subtle temporal transitions leading to seizure onset. This article introduces an attention-improved deep learning model to predict pre-ictal seizures using EEG signals. The suggested model combines CNN-based feature extraction, Bi-LSTM/GRU temporal sequence modelling, and a learnable temporal attention mechanism to identify the early neural dynamics before the onset of seizures. The experimental dataset was constructed from four publicly available EEG repositories following the proposed temporal labeling strategy and comprises EEG recordings from 25 selected patients, 243 annotated seizure events, and 2,847 hours of continuous EEG recordings, using a patient-wise stratified 5-fold cross-validation protocol. Experimental evaluation demonstrates that the proposed framework achieves a predictive accuracy of 93% (±1.2%), sensitivity of 90% (±1.4%), specificity of 90% (±1.1%), an AUC-ROC of 0.93 (±0.012), a PR-AUC of 0.961 (±0.009), and a low false alarm rate of 0.12 ± 0.015 per hour. To confirm statistical reliability, all reported metrics are validated across 10 independent runs and supplemented with 95% confidence intervals and paired Wilcoxon signed-rank tests (p < 0.05) against all baselines. Attention weight analysis verifies that the model selectively targets temporally informative pre-ictal EEG regions and improves clinical interpretability.
Maleka Anjum, Shubhangi D. C.· Journal of Innovative Image...· 0 citations
Introduction Epilepsy is a common neurological disease, and accurate seizure detection is essential for clinical monitoring and scientific treatment. This study aims to construct an effective intelligent detection model to achieve precise automatic identification of epileptic EEG signals and assist clinical medical decisions. Methods To capture subtle local waveform variations and suppress redundant noise interference in EEG signals, this study adopts one-dimensional convolutional neural network (1D-CNN) layers for adaptive local feature extraction and a lightweight global temporal soft attention mechanism for critical feature enhancement. A hybrid classification model based on bidirectional long short-term memory (Bi-LSTM) and gated recurrent unit (GRU) is proposed for the binary classification of epileptic EEG signals. The synthetic minority oversampling technique (SMOTE) is applied only to the training data within each cross-validation fold to alleviate the class imbalance problem of EEG datasets. Results The proposed hybrid model achieves a binary classification accuracy of 99.23%, while delivering an especially balanced sensitivity (99.29%) and specificity (99.34%), with a difference (∆ Sens–Spec) of only 0.05%, verified on the public UCI epileptic seizure recognition data set. Discussion The CNN-Bi-LSTM-GRU and attention-integrated hybrid network can effectively distinguish seizure and non-seizure EEG signals. And a nearly equal sensitivity and specificity suggests robust and unbiased classification. Which is critical for clinical deployment. The proposed method achieves competitive performance compared with most recent mainstream algorithms, which can offer a potential automated detection reference to assist clinical analysis of epilepsy EEG signals.
Xingran Wang, Ting-Hao Gong, Xue-Jia Li et al.· Frontiers in Neuroscience· 0 citations
The use of pre-trained models reduced the training time and resources required, and the unique application of the ensemble learning approach produces more robust and reliable results compared to individual deep learning models.
Unnati Chaurasia, Shilpa Sj, H. Pathak et al.· Discover Artificial Intellig...· 0 citations
A Deep Hybrid Neural Network framework that combines Convolutional Neural Networks (CNN) with the Aquila Optimizer (AO) for the automatic detection of epileptic seizures utilizing EEG data in MATLAB is introduced.
Swati Chowdhuri, Tiyasha Mondal· International Journal of Eng...· 0 citations
Epileptic Seizure Detection plays a crucial role in identifying irregularities in brain activity patterns, such as seizures, focal onset, and generalized seizures, which often go unnoticed until they escalate into more severe conditions like prolonged seizures or cognitive impairments. Therefore, early prediction of epileptic seizure activities is vital for enabling timely medical responses, optimizing treatment strategies, and improving overall patient care and management. The difficulty of accurately and promptly detecting epileptic seizures from EEG data is addressed in this work. This effort is made more difficult by noise, signal pattern fluctuation, and the requirement for quick analysis. The limitations of current detection techniques, which frequently struggle with accuracy and dependability in clinical settings, define the issue. Addressing this critical need, the research introduces a novel hybrid model that combines Modified Shuffle Net V2 (MSNetV2) and Deep Convolutional Neural Network (DCNN) architectures through Electroencephalogram (EEG) signal data. The novelty of the research is to decrease false positives and increase detection accuracy by combining an advanced hybrid architecture model along with robust feature extraction techniques and sophisticated preprocessing techniques. This innovative model is designed specifically for early epileptic seizure detection and employs a comprehensive methodology that includes several key processes. Initially, the EEG signal data are preprocessed using the Modified Wiener Filtering (MWF) technique to remove noise and improve signal clarity. The preprocessed signals are then subjected to feature extraction to identify the most pertinent features, which are further enhanced to improve the dataset. Here, the min-max normalization procedure is used to carry out the data augmentation process. To identify epileptic episodes, the two network architectures independently process the augmented features before feeding them into the hybrid MSNetV2-DCNN model. The study uses detailed simulations and experimental evaluations to verify the efficacy of the suggested paradigm. The study comes to the conclusion that the MSNetV2-DCNN model exhibits a reliable and effective technique for epileptic seizure identification, underscoring its potential for practical use in traffic management situations as well as medical diagnostics. The proposed work achieves the highest accuracy of approximately 94% in 90% of the training data, which is the highest among the other conventional models.
Automated epileptic seizure detection from electroencephalogram (EEG) signals remains a critical challenge for real-world clinical deployment due to the complex, nonstationary, and multi-scale nature of neural dynamics. Existing deep learning approaches, including convolutional and transformer-based models, often fail to jointly capture spectral–temporal dependencies while maintaining robustness across heterogeneous datasets and noisy clinical environments. In this work, we propose BrainXNet, a novel multi-scale spectro-temporal attention framework that unifies local feature extraction, frequency-aware representation learning, and global temporal modeling within a single architecture. The proposed model integrates (i) multi-scale convolutional pathways to capture transient and long-duration EEG patterns, (ii) a spectral attention module that dynamically emphasizes clinically relevant frequency bands, and (iii) a temporal transformer encoder for modeling long-range dependencies across EEG sequences. Extensive evaluations on two large-scale benchmark datasets, CHB-MIT and TUH Seizure Corpus, demonstrate that BrainXNet achieves state-of-the-art performance, reaching accuracies of 99.1% and 98.4%, respectively. Beyond in-dataset performance, the proposed framework exhibits strong cross-dataset generalization, maintaining over 94% accuracy in transfer settings, and demonstrates high robustness under noisy conditions. Ablation studies further confirm the complementary contributions of each architectural component. These results highlight the effectiveness of explicitly modeling multi-scale spectro-temporal dynamics for EEG analysis and position BrainXNet as a promising candidate for reliable, real-time clinical seizure detection systems. This work bridges the gap between high-performance experimental models and practical deployment in diverse healthcare environments.
Mostafa Gamal, Mustafa Abdel-Wanes· Scientific Reports· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
