A Novel Co-Designed Multi-Domain Entropy and Its Dynamic Synapse Classification Approach for EEG Seizure Detection

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Signal Processing

Background:

• Automated electroencephalography (EEG) seizure detection is crucial for clinical applications.
• Existing methods struggle with comprehensive feature extraction and generic classifiers, limiting real-world effectiveness.
• A need exists for advanced techniques to improve the accuracy and robustness of EEG-based seizure detection.

Purpose of the Study:

• To propose DySC-MDE, an end-to-end co-designed framework for enhanced EEG seizure detection.
• To develop a novel multi-domain entropy (MDE) representation for characterizing nonlinear EEG dynamics.
• To introduce a dynamic synapse classifier (DySC) tailored to the MDE features for adaptive information fusion.

Main Methods:

• Constructed a multi-domain entropy (MDE) representation using amplitude-sensitive permutation entropy (ASPE) and its variants (RCMASPE, HASPE-DWT, TSMASPE).
• Developed a dynamic synapse classifier (DySC) with parallel processing pathways and adaptive synaptic gating for heterogeneous feature fusion.
• Conducted extensive experiments on two public datasets (Bonn and CHB-MIT) using cross-validation for performance evaluation.

Main Results:

• Achieved high accuracy (97.50%-98.93%) and F1-scores (97.58%-98.87%) in binary classification tasks on the Bonn and CHB-MIT datasets.
• Demonstrated strong performance in a three-class task with an F1-score of 96.83%, indicating robust generalization.
• The co-designed framework significantly improved the analysis of complex epileptic EEG signals.

Conclusions:

• The joint optimization of nonlinear dynamic feature representations and structure-aware classifiers enhances EEG seizure detection.
• DySC-MDE offers a novel and effective direction for robust and accurate automated seizure detection.
• The proposed framework shows significant potential for clinical application in epilepsy management.