Localization of Active Brain Sources From EEG Signals Using Empirical Mode Decomposition: A Comparative Study

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) source localization is crucial for clinical applications like epilepsy and ADHD studies.
• Distributed-source models estimate neural activity by solving the inverse problem, often enhanced by spatio-temporal constraints.
• Frequency band separation is vital for analyzing specific neural oscillations but poses challenges in maintaining temporal resolution.

Purpose of the Study:

• To improve frequency-band identification and temporal resolution in EEG pre-processing for more accurate brain source reconstruction.
• To compare the efficacy of Empirical Mode Decomposition (EMD) variants against traditional methods for EEG inverse problem solving.
• To assess the impact of mitigating EMD's mode-mixing problem on the spatio-temporal accuracy of reconstructed brain activity.

Main Methods:

• EEG signals were decomposed into frequency bands using Empirical Mode Decomposition (EMD) and Wavelet Transform (WT).
• Three EMD variants—masking EMD, Ensemble-EMD (EEMD), and multivariate EMD (MEMD)—were explored to address the mode-mixing issue.
• Spatial reconstruction accuracy was evaluated using the Wasserstein metric on simulated and real EEG data.

Main Results:

• Masking EMD and MEMD effectively mitigated the mode-mixing problem in EMD, leading to improved spatio-temporal reconstruction of brain sources.
• EMD variants (masking EMD, EEMD) outperformed standard EMD, Multiple Sparse Priors, and wavelet packet decomposition in spatial reconstruction accuracy.
• All tested EMD variants offered substantially better spatial resolution than standard EMD by reducing mode-mixing.

Conclusions:

• EMD-based pre-processing, particularly using masking EMD and EEMD, significantly enhances the accuracy and spatial resolution of EEG brain source reconstruction.
• Addressing the mode-mixing problem is critical for optimizing EMD's utility in analyzing neural activity.
• Further research into EMD variants holds promise for advancing clinical EEG analysis and brain source imaging.