Epileptic Seizure Detection Using Brain-Rhythmic Recurrence Biomarkers and ONASNet-Based Transfer Learning

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epilepsy management requires real-time monitoring of abnormal brain activity.
• Electroencephalogram (EEG) signals offer potential for detecting seizures.
• Developing accurate EEG-based seizure detection systems is crucial for clinical applications.

Purpose of the Study:

• To propose a single-channel seizure detection system utilizing brain-rhythmic recurrence biomarkers (BRRM) and an optimized neural network architecture (ONASNet).
• To evaluate the efficiency and performance of the proposed BRRM-ONASNet system against established neural network models and existing methods.
• To demonstrate the utility of analyzing nonlinear EEG features for epilepsy detection and computational medical assistance.

Main Methods:

• Brain-rhythmic recurrence biomarkers (BRRM) were employed to capture nonlinear dynamics from EEG signals.
• An optimized neural network architecture, ONASNet, was developed using a modified neural network searching strategy.
• Transfer learning was utilized to adapt ONASNet for EEG data analysis, extracting features from various brain rhythms simultaneously.

Main Results:

• The BRRM-ONASNet system demonstrated superior performance compared to other transfer-learning models in terms of learning capability, stability, model size, and prediction latency.
• ONASNet-based models required fewer computational resources and exhibited shorter prediction latencies.
• The BRRM-ONASNet method achieved 100% accuracy on the Bonn University EEG dataset, outperforming existing seizure detection methods.

Conclusions:

• The proposed method, integrating nonlinear feature analysis from phase-space representations with deep neural networks, offers novel insights into EEG decoding.
• The BRRM-ONASNet system represents a computationally efficient and highly accurate tool for epileptic seizure detection.
• This approach contributes significantly to the development of advanced computational medical assistance for epilepsy management.