Hierarchical Bayesian inference for the EEG inverse problem using realistic FE head models: depth localization and...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biophysics

Background:

• Estimating brain activity from electromagnetic fields (EEG/MEG) is an ill-posed inverse problem.
• Accurate localization of deep brain sources remains a significant challenge for existing inverse methods.
• Hierarchical Bayesian modeling (HBM) offers a unified framework for current density reconstruction (CDR).

Purpose of the Study:

• To evaluate fully-Bayesian inference methods within HBM for brain source localization.
• To assess HBM's performance with realistic finite element (FE) head models and focal source configurations.
• To investigate HBM's effectiveness in depth localization and source separation in multi-source scenarios.

Main Methods:

• Utilized fully-Bayesian inference for Hierarchical Bayesian Modeling (HBM).
• Employed realistic, high-resolution finite element (FE) head models.
• Introduced Wasserstein distances for validating inverse methods in complex source scenarios.

Main Results:

• HBM demonstrated improved depth localization compared to Minimum Norm Estimation (MNE) and sLORETA.
• HBM showed enhanced separation of focal sources in multi-source configurations.
• Promising results were achieved for challenging scenarios where traditional methods exhibited errors.

Conclusions:

• Fully-Bayesian HBM is a promising framework for improving EEG/MEG source localization.
• HBM offers significant advantages over established CDR methods, particularly for depth accuracy and source separation.
• The introduced Wasserstein distances provide a robust measure for validating inverse methods in complex neurophysiological data analysis.