MuGEP: Multiplex Graph-Based Brain Network Modeling for Epileptic Seizure Prediction Using Intracranial EEG.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epilepsy seizure prediction is critical for patient safety and quality of life.
• Intracranial electroencephalogram (iEEG) provides detailed brain network information for epilepsy research.
• Current deep learning models for seizure prediction often fail to capture complex brain network dynamics and channel-specific information from iEEG signals.

Purpose of the Study:

• To develop an advanced framework, MuGEP, for more effective epileptic seizure prediction using iEEG data.
• To model diverse and fine-grained brain network relationships, including cross-frequency coupling, for improved prediction accuracy.

Main Methods:

• Proposed a Multiplex Graph-based brain network modeling framework (MuGEP).
• Designed a specialized Multiplex Brain Graph (MBG) representing frequency bands as nodes and Cross-Frequency Coupling (CFC) inspired relationships as edges across three subgraphs.
• Developed a novel MBG learning network using graph convolution networks and a joint fusion module to integrate intra- and inter-subgraph patterns.

Main Results:

• MuGEP demonstrated promising performance in seizure prediction tasks.
• The framework effectively captured complex brain network relationships and channel information from iEEG signals.
• Evaluations on the Kaggle and SWEC-ETHZ datasets confirmed the advantages of MuGEP.

Conclusions:

• MuGEP offers a significant advancement in seizure prediction by leveraging multiplex graph modeling of iEEG data.
• The proposed framework effectively exploits cross-frequency coupling and diverse network interactions for enhanced prediction accuracy.
• MuGEP holds potential for improving clinical management of epilepsy through accurate, automated seizure prediction.