EEG Microstate Features as an Automatic Recognition Model of High-Density Epileptic EEG Using Support Vector Machine

Area of Science:

• Neuroscience
• Medical Diagnostics
• Signal Processing

Background:

• Epilepsy is a significant nervous system disorder requiring accurate diagnosis.
• Video electroencephalogram (EEG) is a key diagnostic tool for epilepsy.
• Microstate analysis offers a novel approach to understanding EEG patterns.

Purpose of the Study:

• To investigate the utility of EEG microstate analysis for diagnosing epilepsy.
• To compare the effectiveness of microstate parameters versus other EEG features in epilepsy classification.
• To develop an automated classification system for epileptic EEG using machine learning.

Main Methods:

• Recruited patients with focal epilepsy and healthy controls for resting EEG recording.
• Applied microstate analysis to EEG data, extracting parameters like duration, occurrence, and coverage.
• Utilized a Support Vector Machine (SVM) classifier with various EEG features (linear, non-linear, microstate parameters).

Main Results:

• Microstate parameters in the gamma sub-band achieved 87.18% accuracy for interictal epilepsy recognition.
• The SVM classifier demonstrated effective epilepsy classification using both microstate parameters and EEG features.
• Microstate parameters showed superior classification performance compared to other extracted EEG features.

Conclusions:

• EEG microstate analysis, particularly microstate parameters, is a promising method for epilepsy diagnosis.
• The SVM classifier effectively utilizes microstate parameters for automated epileptic EEG classification.
• This approach aids in the accurate identification and diagnosis of epilepsy.