Fully Automated EEG Source Imaging Using Structured Sparsity for Single and Multiple Synchronous Epileptic...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Accurate localization of epileptic zones from high-resolution electroencephalography (HR-EEG) is crucial for effective treatment.
• Existing EEG Source Imaging (ESI) methods face challenges with multifocal epilepsy and ill-posed inverse problems.
• Current ESI methods often rely on heuristic regularization parameter selection, impacting performance.

Purpose of the Study:

• To develop an efficient and automated EEG Source Imaging (ESI) method for accurate epileptic zone localization.
• To address the ill-posed nature of the inverse problem in ESI by imposing sparsity constraints.
• To evaluate the performance of the proposed method against traditional approaches using synthetic and real HR-EEG data.

Main Methods:

• An efficient ESI approach imposing sparsity on source-level activity and spatial gradients was developed.
• The method features automated, iterative adjustment of the regularization parameter based on noise levels.
• Performance was evaluated using realistic synthetic HR-EEG data (unifocal and multifocal epilepsy) and real HR-EEG data from absence seizures.

Main Results:

• The proposed method accurately reconstructs epileptic sources, outperforming traditional ESI techniques across various scenarios.
• The approach effectively reduces polarity artifacts, minimizing ghost sources and spatial discontinuities.
• Successful recovery of homogeneous and well-delineated epileptic regions was confirmed using real HR-EEG data.

Conclusions:

• The novel ESI method offers improved accuracy and robustness for localizing epileptic zones from HR-EEG data.
• Automated parameter selection enhances the reliability and applicability of ESI in clinical settings.
• This approach holds significant potential for advancing the diagnosis and surgical planning of epilepsy.