Tree representations of brain structural connectivity via persistent homology

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Graph Theory

Background:

• The brain's structural connectome, mapped by diffusion MRI (dMRI) tractography, represents neural pathways.
• Current adjacency matrix (AM) representations of connectomes are high-dimensional and computationally intensive.
• Relating individual brain connectomes to traits requires efficient and interpretable representations.

Purpose of the Study:

• To propose a novel, lower-dimensional tree representation for the brain structural connectome.
• To leverage computational topology and cortical surface information for improved connectome analysis.
• To enhance statistical and computational efficiency in relating brain structure to traits.

Main Methods:

• Applied tractography to dMRI data to generate white matter fiber bundles.
• Developed a tree-based representation of the brain connectome, incorporating topological and cortical surface data.
• Utilized data from the Human Connectome Project (HCP) for validation.

Main Results:

• The proposed tree representation effectively captures essential information and interpretability of the brain connectome.
• This novel representation significantly reduces dimensionality compared to traditional adjacency matrices.
• Improved statistical and computational efficiency was demonstrated for connectome-trait analyses.

Conclusions:

• A simpler, dimension-reduced tree representation offers a more efficient alternative to adjacency matrices for brain connectome analysis.
• This approach facilitates more effective statistical modeling of brain structure-trait relationships.
• The method provides a valuable tool for neuroscience research, with reproducible code available.