CutFEM-based MEG forward modeling improves source separability and sensitivity to quasi-radial sources: A...

Area of Science:

• Computational neuroscience
• Biophysics
• Medical imaging

Background:

• Magnetoencephalography (MEG) source analysis necessitates accurate computation of magnetic fields from brain activity.
• Estimating volume conduction effects within the human head is crucial for the MEG forward problem.

Purpose of the Study:

• Introduce the Cut finite element method (CutFEM) as a novel approach for solving the MEG forward problem.
• Evaluate CutFEM's performance against established methods like the boundary element method (BEM) and hexahedral finite element method (FEM).

Main Methods:

• Implemented CutFEM for MEG forward problem computation, noting its flexible meshing capabilities for complex geometries.
• Compared CutFEM with a 3-compartment BEM and a 6-compartment hexahedral FEM using somatosensory evoked fields (SEF) data from 19 participants.
• Reconstructed neural generators (M20 components) using both unregularized and regularized inversion techniques.

Main Results:

• Changing the forward model led to approximately 1 cm differences in source location and significant orientation variations.
• 6-compartment FEM approaches showed a significantly better goodness of fit to measured MEG data than the 3-compartment BEM.
• CutFEM demonstrated improved source separability compared to both BEM and hexahedral FEM, particularly for quasi-radial sources.

Conclusions:

• Head models with 6 compartments offer superior performance over 3-compartment models for MEG source reconstruction.
• The CutFEM approach represents a valuable advancement for MEG source analysis, especially for predominantly radial neural sources.