Brain Activity Mapping from MEG Data via a Hierarchical Bayesian Algorithm with Automatic Depth Weighting

Area of Science:

• Biophysics
• Neuroimaging
• Computational Neuroscience

Background:

• Magnetoencephalography (MEG) inverse solving aims to localize neural activity.
• Existing methods like wMNE, dSPM, and sLORETA have limitations in accurately identifying active brain regions.
• A systematic evaluation of the iterated alternating sequential (IAS) algorithm's performance is lacking.

Purpose of the Study:

• To propose novel statistical protocols for quantifying MEG inverse solver performance.
• To systematically evaluate the accuracy and precision of the IAS algorithm in identifying active brain regions.
• To compare the IAS algorithm against standard methods (wMNE, dSPM, sLORETA).

Main Methods:

• Development of Monte Carlo sampling-based statistical protocols.
• Generation of simulated active brain regions across an atlas.
• Performance measurement based on the concentration of reconstructed activity within simulated regions.
• Bayes factor analysis to test hypotheses of correct versus erroneous source attribution.

Main Results:

• The IAS algorithm demonstrates strong performance in identifying active brain regions.
• The proposed protocols provide a robust, bias-free method for evaluating inverse solvers.
• IAS shows comparable or superior performance to wMNE, dSPM, and sLORETA for both cortical and subcortical sources.
• The methodology effectively handles single or multiple simultaneous active regions without prior knowledge of their number.

Conclusions:

• The IAS MEG inverse solver is a reliable tool for accurately localizing neural activity.
• The novel statistical protocols offer a standardized approach for benchmarking MEG inverse solutions.
• The IAS algorithm shows promise for advancing neuroimaging research by improving source localization accuracy.