Improved Hybrid Local Binary Structural Pattern Shallow Graph Deep Convolutional Attention Neural Networks With...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epileptic seizure (ES) detection from electroencephalography (EEG) is challenging due to signal noise and individual brain variability.
• Existing methods often lack accuracy, are computationally expensive, and generalize poorly.

Purpose of the Study:

• To propose an advanced hybrid deep learning model for automated ES detection and diagnosis from EEG signals.
• To overcome limitations of traditional methods in terms of accuracy, noise handling, and computational efficiency.

Main Methods:

• Utilized subaperture keystone transform matched filtering (SAKTMF) for EEG signal denoising and artifact reduction.
• Employed second-order synchro-extracting transform combined with empirical wavelet transform (SOSTC-EWT) for precise time-frequency and spectral feature extraction.
• Developed an improved hybrid local binary structural pattern shallow graph deep convolutional attention neural network (IHLBSPSGDCAN2Nets) integrated with synergistic fibroblast optimization (SFO) for classification and hyperparameter tuning.

Main Results:

• Achieved an outstanding classification accuracy of 99.9% for ES detection.
• Demonstrated superior performance compared to conventional methods in terms of reliability, robustness, and efficiency.
• The proposed model effectively handled noise, extracted precise EEG features, and optimized performance with accelerated convergence and minimal computational cost.

Conclusions:

• The IHLBSPSGDCAN2Nets + SFO model offers a highly accurate and efficient solution for automated epileptic seizure detection from EEG signals.
• This advanced approach significantly improves upon existing methods, providing a more reliable tool for clinical diagnosis and research.