Diffusion wavelets on connectome: Localizing the sources of diffusion mediating structure-function mapping using...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Graph Theory

Background:

• The relationship between structural connectivity (SC) and functional connectivity (FC) in the brain is complex.
• Existing models using graph diffusion to predict FC from SC lack region-specific scale optimization.
• This limits the applicability and interpretability of graph diffusion methods in neuroscience.

Purpose of the Study:

• To develop a novel graph diffusion wavelet approach to learn optimal diffusion scales for each region of interest (RoI).
• To accurately estimate the structural connectivity-functional connectivity (SC-FC) mapping using these learned scales.
• To investigate the relationship between diffusion scales and brain network organization.

Main Methods:

• Utilized a graph diffusion wavelet framework to learn diffusion scales tailored to individual brain regions.
• Applied the method to the Human Connectome Project dataset for SC-FC mapping.
• Employed a linear and computationally efficient architecture.

Main Results:

• Achieved a high average Pearson's correlation of 0.833 for predicting FC, outperforming existing methods.
• Demonstrated the power-law distribution of diffusion scales across the brain.
• Identified that the bilateral frontal pole, with its large diffusion scale, forms a significant community structure.

Conclusions:

• The graph diffusion wavelet framework offers a powerful and efficient method for SC-FC mapping.
• The findings support the role of the frontal pole in resting-state networks, as indicated by its community structure.
• This approach enhances our understanding of how brain structure dictates functional connectivity.