A Unified Hypergraph-Mamba Framework for Adaptive Electroencephalogram Modeling in Multi-view Seizure Prediction.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Seizure prediction from Electroencephalogram (EEG) signals is vital for epilepsy management.
• Current models face challenges in capturing dynamic inter-channel dependencies and spectral variations, particularly in cross-patient scenarios.

Purpose of the Study:

• To develop a novel framework, Unified Hypergraph-Mamba (UHM), for improved seizure prediction.
• To address limitations in existing models regarding high-order spatial and adaptive spectral modeling.

Main Methods:

• Integrated hypergraph-based spatial modeling with Mamba-based adaptive spectral modeling.
• Designed a hypergraph attention mechanism for high-order spatial interactions among EEG channels.
• Utilized a Mamba architecture for adaptive spectral modeling of preictal states.

Main Results:

• The UHM framework demonstrated superior performance compared to state-of-the-art baselines.
• Achieved enhanced sensitivity and Area Under the Curve (AUC) in both patient-specific and cross-patient settings.
• Successfully modeled spatiotemporal EEG dynamics for seizure prediction.

Conclusions:

• The UHM framework offers a unified approach to jointly model spatiotemporal EEG dynamics.
• This novel method significantly improves seizure prediction accuracy, especially in challenging cross-patient scenarios.
• The findings pave the way for more effective proactive intervention strategies in epilepsy.