Neurofusionnet: a comprehensive framework for accurate epileptic seizure prediction from EEG data with hybrid...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epileptic seizures pose significant challenges for prediction and management.
• Current Electroencephalogram (EEG) analysis methods require sophisticated techniques for accurate seizure detection.

Purpose of the Study:

• To develop a comprehensive paradigm for epileptic seizure prediction using advanced EEG data processing.
• To introduce a novel deep learning detection model for improved accuracy and durability.

Main Methods:

• Multi-step methodology including pre-processing (bandpass filtering, Independent Component Analysis, Z-score normalization, Common Spatial Patterns), feature extraction (time-domain, frequency-domain via Fourier Transform, time-frequency via Wavelet Transform), and feature selection (Hybrid Chimp Enhanced Fox Optimization algorithm).
• Development of the NeuroFusionNet detection model integrating Improved ShuffleNet V2, SqueezeNet, EfficientNet V2, and Multi Head Attention (MHA) based GhostNet V2.

Main Results:

• The proposed methodology effectively processes EEG data, reducing noise and artifacts.
• The Hybrid Chimp Enhanced Fox Optimization algorithm efficiently selects discriminative features.
• The NeuroFusionNet model demonstrates capability in capturing complex patterns linked to epileptic episodes.

Conclusions:

• The developed paradigm offers a robust and accurate approach to epileptic seizure prediction.
• The NeuroFusionNet model represents a significant advancement in deep learning for neurological disorder detection.