Uncovering Cortical Units of Processing From Multi-Layered Connectomes

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Network Science

Background:

• Diffusion MRI (dMRI) and functional MRI (fMRI) provide insights into brain connectivity.
• Discrepancies between dMRI and fMRI connectivity profiles hinder understanding of the structural and functional connectome relationship.
• Current methods often focus on connection-level correspondence, overlooking modular organization.

Purpose of the Study:

• To investigate the correspondence between structural and functional brain networks based on modular organization.
• To develop a data-driven approach for identifying shared canonical processing units across modalities.
• To assess the benefits of multi-modal integration for characterizing brain connectomes.

Main Methods:

• Utilized a stochastic block-model (SBM) for data-driven clustering of whole-brain connectivity networks.
• Employed a joint model assuming shared parcellation but independent connectivity structures across modalities.
• Quantified clustering performance using prediction accuracy within each modality.

Main Results:

• A joint SBM model achieved a consensus representation that effectively described both functional and structural connectomes.
• The integrated model provided improved functional connectivity representations compared to using functional data alone.
• Removing anatomical correspondence between modalities significantly reduced predictive performance, confirming its importance.

Conclusions:

• Multi-modal integration, using shared processing units, yields consensus representations that characterize individual modalities despite inherent biases.
• The findings highlight the importance of multi-layered connectomes in revealing supplementary information about the brain's canonical processing units.
• Anatomical correspondence between structural and functional units is crucial for accurate brain network modeling.