Epilepsy Detection in EEG Using Grassmann Discriminant Analysis Method

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Epilepsy diagnosis relies on analyzing electroencephalogram (EEG) signals, but low signal-to-noise ratio (SNR) presents significant challenges.
• Machine learning approaches have advanced epileptic detection from EEG, yet real-world application faces hurdles.

Purpose of the Study:

• To develop an automated method for epileptic detection using EEG signals.
• To address the challenge of low SNR in EEG data for improved diagnostic accuracy.

Main Methods:

• Utilized Fréchet mean-based Grassmann discriminant analysis (FMGDA) for EEG data dimensionality reduction and clustering.
• Mapped EEG features into Grassmann manifold space and employed Fréchet mean for cluster center representation.
• Implemented FMGDA to maximize between-class distance while minimizing within-class distance.

Main Results:

• The proposed FMGDA algorithm effectively reduced high-dimensional EEG data to a lower-dimensional representation.
• Experimental results demonstrated significant improvements in epileptic detection accuracy compared to existing Grassmann manifold methods.
• The method showed robust performance on benchmark EEG datasets, indicating its practical utility.

Conclusions:

• The FMGDA algorithm offers a promising automated approach for epileptic detection from EEG signals.
• This method enhances diagnostic accuracy by effectively handling low SNR EEG data.
• The findings suggest FMGDA is a valuable tool for advancing epilepsy diagnosis in clinical settings.