Improving EEG Forward Modeling Using High-Resolution Five-Layer BEM-FMM Head Models: Effect on Source Reconstruction...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Electroencephalographic (EEG) source localization is crucial for understanding brain activity.
• Current methods often rely on simplified head models, potentially limiting accuracy.
• Advancements in computational power enable more complex modeling approaches.

Purpose of the Study:

• To evaluate the impact of head model complexity on EEG source localization accuracy.
• To compare a standard three-layer boundary element method (BEM) with a high-resolution five-layer BEM-FMM with adaptive mesh refinement (AMR).
• To quantify localization errors across the grey matter.

Main Methods:

• Generated noiseless 256-channel EEG data from 15 subjects.
• Utilized four anatomically relevant dipole positions and three conductivity sets.
• Compared a three-layer BEM inverse method with a five-layer BEM-FMM/AMR forward solver.
• Mapped localization errors across 4000 dipole positions in the grey matter.

Main Results:

• Average localization error for selected dipoles was ~5mm (±2mm) with ~12° (±7°) orientation error.
• Average source localization error across the entire grey matter was ~9mm (±4mm).
• Errors tended to be smaller in the occipital lobe.

Conclusions:

• Three-layer BEM models show robustness in noiseless conditions but can yield substantial errors (10-20mm).
• Higher-complexity models (five or more layers) are necessary for accurate EEG source reconstruction, especially with noisy data.
• BEM-FMM with AMR provides an efficient and accurate approach for high-resolution modeling.