Voxel-Wise Brain Graphs From Diffusion MRI: Intrinsic Eigenspace Dimensionality and Application to Functional MRI.

Area of Science:

• Neuroimaging
• Graph Theory
• Computational Neuroscience

Background:

• Conventional brain graphs use atlas-defined regions as nodes.
• This limits the resolution and specificity of anatomical-functional relationships.

Purpose of the Study:

• To develop and validate high-resolution, subject-specific brain graphs using individual voxels as nodes.
• To investigate the spectral properties of these graphs and their relationship to brain function.

Main Methods:

• Modeled brain structure using voxel-to-voxel connections derived from diffusion MRI data.
• Analyzed graph Laplacian spectral properties and inter-subject variability.
• Applied graph signal processing to relate anatomical graphs to functional MRI data.

Main Results:

• Graph Laplacian eigenmodes revealed detailed spatial profiles corresponding to white matter pathways.
• Demonstrated that the intrinsic dimensionality of these high-resolution graphs is significantly lower than their full dimensions.
• Showed that brain activity (task and resting-state fMRI) can be approximated by low-frequency components of the anatomical graph.

Conclusions:

• Proposed voxel-based graphs provide a novel scaffold for studying brain organization and function at an individual level.
• Spectral graph theory offers powerful tools for analyzing these high-resolution brain structures.
• This approach enhances understanding of the anatomical underpinnings of brain activity.