WGB-GLFI: A Novel Graph-Based Global-Local Feature Interaction Framework for Automated Seizure Detection

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epilepsy detection is challenging due to seizure unpredictability and risks to patient safety.
• Deep learning methods are increasingly used for electroencephalogram (EEG) analysis in epilepsy.
• Current deep learning models often lack unified spatial modeling and struggle with local feature details, limiting performance.

Purpose of the Study:

• To develop an advanced deep learning framework for enhanced epilepsy detection.
• To address limitations in spatial modeling and local feature extraction in existing methods.
• To improve the accuracy and robustness of seizure detection.

Main Methods:

• Proposed the Weighted Graph Building Global-Local Feature Interaction (WGB-GLFI) framework.
• Integrated a Weighted Graph Building (WGB) module for spatial connectivity.
• Incorporated a Global-Local Feature Interaction (GLFI) module for dynamic pattern analysis.

Main Results:

• The WGB-GLFI framework achieved high accuracy rates: 99.28% on CHB-MIT, 99.21% on Siena Scalp, and 99.30% on private datasets.
• Demonstrated robust and reliable seizure detection performance across multiple datasets.
• Effectively captured dynamic spatial relationships and integrated global-local features.

Conclusions:

• The WGB-GLFI framework significantly enhances epilepsy detection accuracy and robustness.
• Provides faster and more reliable seizure detection for improved patient intervention and quality of life.
• Represents a significant advancement in applying deep learning to clinical epilepsy management.