The Iterative Reweighted Mixed-Norm Estimate for Spatio-Temporal MEG/EEG Source Reconstruction

Area of Science:

• Neuroimaging
• Biophysics
• Computational Neuroscience

Background:

• Magnetoencephalography (MEG) and electroencephalography (EEG) offer non-invasive brain activity analysis with high temporal and good spatial resolution.
• The bioelectromagnetic inverse problem is ill-posed, necessitating constraints for accurate source localization.
• Current methods using l1-norm constraints for spatial sparsity can introduce amplitude bias and suboptimal source selection.

Purpose of the Study:

• To develop a novel inverse solver that addresses amplitude bias and improves source selection in MEG/EEG analysis.
• To introduce a block-separable penalty with a Frobenius norm per block and a l0.5-quasinorm over blocks for enhanced source imaging.
• To present the iterative reweighted Mixed Norm Estimate (irMxNE) algorithm for solving the non-convex optimization problem.

Main Methods:

• Developed an inverse solver utilizing a block-separable penalty (Frobenius norm per block, l0.5-quasinorm over blocks).
• Proposed the iterative reweighted Mixed Norm Estimate (irMxNE) optimization scheme.
• Employed a block coordinate descent scheme and active set strategy to solve the non-convex problem efficiently.
• Compared irMxNE against dSPM and RAP-MUSIC using MEG data and simulations.

Main Results:

• The proposed irMxNE method effectively addresses amplitude bias inherent in standard l1-norm based methods.
• Demonstrated improved source recovery and stability compared to the standard Mixed Norm Estimate (MxNE).
• Empirical evidence from simulations and MEG data analysis supports the superiority of irMxNE.

Conclusions:

• The irMxNE method offers a significant advancement in sparse source imaging for MEG/EEG data.
• This approach provides more accurate and stable estimates of neuronal activation, overcoming limitations of previous techniques.
• The findings suggest irMxNE as a valuable tool for non-invasive brain activity analysis.