WRA-MTSI: A Robust Extended Source Imaging Algorithm Based on Multi-Trial EEG.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Reconstructing brain activity from electroencephalography (EEG) signals is vital for understanding brain function and dysfunction.
• Single-trial EEG data reconstruction is often unstable and variable due to signal non-stationarity and noise.
• Existing methods struggle with artifact mitigation and accurate source extent estimation in multi-trial EEG data.

Purpose of the Study:

• To develop a robust multi-trial EEG source imaging (ESI) method that leverages shared information across trials.
• To improve the stability and accuracy of brain activity reconstruction from noisy EEG data.
• To address limitations in current single-trial and multi-trial ESI techniques.

Main Methods:

• Proposes Wasserstein Regularization-based Multi-Trial Source Imaging (WRA-MTSI).
• Employs Wasserstein regularization for multi-trial source distribution similarity learning.
• Utilizes structured sparsity constraints and an alternating direction method of multipliers (ADMM) algorithm for efficient computation.

Main Results:

• WRA-MTSI effectively mitigates artifact influence in EEG data, outperforming single-trial ESI methods like wMNE, LORETA, SISSY, and SBL.
• Demonstrates superior performance in estimating source extents compared to state-of-the-art multi-trial ESI methods such as group lasso, the dirty model, and MTW.
• Validation through numerical simulations and real EEG data analysis confirms WRA-MTSI's efficacy.

Conclusions:

• WRA-MTSI offers a robust and effective solution for EEG source imaging with multi-trial noisy data.
• The method enhances the reliability of brain activity reconstruction in challenging EEG conditions.
• WRA-MTSI represents a significant advancement in analyzing complex brain signals.