Improved forward EEG calculations using local mesh refinement of realistic head geometries

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Accurate modeling of human head structures (scalp, skull, brain) is crucial for electrophysiological studies.
• Surface mesh generation from medical images enables realistic simulations.
• The boundary element method (BEM) is a common technique for these simulations.

Purpose of the Study:

• To develop and evaluate a semi-automatic method for constructing realistic head surface meshes.
• To assess the impact of mesh density on the accuracy of dipole modeling using BEM.
• To determine optimal mesh parameters for simulating deep and superficial sources.

Main Methods:

• Semi-automatic construction of scalp, skull, and brain meshes from MRI data.
• Evaluation of spherical and realistic head models using BEM.
• Comparison of numerical and analytical solutions for spherical geometries.
• Assessment of models against a highly refined reference mesh for realistic geometries.
• Definition and analysis of global and local mesh densities (triangles/cm2).

Main Results:

• Comparable results were obtained for both spherical and realistic geometries.
• Meshes with a global density of 0.5 triangles/cm2 yielded acceptable results for deep dipoles (>20-30 mm depth).
• Superficial dipoles (2-3 mm < depth < 20-30 mm) required local mesh refinement to 5-8 triangles/cm2 for comparable accuracy.

Conclusions:

• The study provides guidelines for mesh density selection in BEM simulations of human head models.
• Local mesh refinement is essential for accurate localization of superficial neural sources.
• The developed method facilitates the creation of realistic head models for neuroscience research.