Automated Parcellation of the Cortex Using Structural Connectome Harmonics.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Brain Imaging

Background:

• Dividing the human cortex into distinct regions is a key challenge in neuroscience.
• Existing brain parcellations often rely on anatomical or functional magnetic resonance imaging (fMRI) data.
• There is a need for parcellations based solely on structural data from diffusion MRI.

Purpose of the Study:

• To develop a novel method for human cortical parcellation using only structural connectivity data.
• To demonstrate that harmonic modes of a whole-brain structural connectome graph can yield meaningful cortical regions.
• To create subject-specific and hierarchical parcellations with tunable group-level correspondence.

Main Methods:

• Construction of a novel high-resolution, vertex-level graph model of the whole-brain structural connectome using diffusion MRI.
• Analysis of harmonic modes derived from the structural connectome graph.
• Development of a multi-layer graph formulation and application of hierarchical clustering to the harmonic modes.

Main Results:

• The harmonic modes of the structural connectome graph successfully generated cortical parcellations that qualitatively agree with established atlases.
• The multi-layer graph formulation enabled the creation of subject-specific parcellations at various resolutions.
• The method ensures tunable group-level correspondence for the generated parcellations.

Conclusions:

• Structural connectivity, analyzed through graph harmonic modes, provides a robust basis for human cortical parcellation.
• This novel approach offers a data-driven alternative to traditional methods, utilizing only diffusion MRI.
• The developed technique facilitates the generation of personalized and hierarchical brain atlases.