Multi-Resolution Graph Based Volumetric Cortical Basis Functions From Local Anatomic Features.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Clinical MRI provides high-resolution anatomical data, but scalp electroencephalography (EEG) source localization struggles with low resolution and ill-posed problems.
• Current volumetric EEG methods simplify anatomical structures through grid coarsening, limiting fine-scale analysis.
• Improved dimensionality reduction is crucial for managing computational and storage complexity in EEG source localization.

Purpose of the Study:

• To develop and validate a novel approach for near-arbitrary spatial scaling in volumetric EEG source localization.
• To enhance the accuracy and computational efficiency of mapping electrical activity in the brain.

Main Methods:

• A voxelwise brain parcellation was used, with sub-parcels identified via local cortical connectivity using iterated graph cuts.
• Spatial basis functions were constructed within each parcel using leadfield matrix decomposition or spectral graph theory.

Main Results:

• Volumetric basis functions demonstrated up to a 30% improvement in reconstruction accuracy compared to traditional methods.
• Computational complexity was reduced by over two orders of magnitude.
• Accurate localization of seizure onset regions was achieved using real patient data from epilepsy surgical candidates.

Conclusions:

• Spatial dimensionality reduction using volumetric basis functions significantly enhances EEG source localization accuracy.
• This approach substantially decreases computational requirements, making advanced algorithms more feasible.
• The method enables anatomically driven, multi-resolution volumetric reconstruction for modern neuroimaging.