Scale-Dependent Variability and Quantitative Regimes in Graph-Theoretic Representations of Human Cortical Networks

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• Brain connectivity studies are vital for understanding neurological health and disease.
• Graph-theoretic approaches to brain networks are sensitive to spatial scale and model definitions (vertices/edges).
• Previous macroscale neuroimaging studies used fewer network nodes than the current study.

Purpose of the Study:

• To investigate how spatial scale and model parameters influence graph-theoretic network analysis of brain circuitry.
• To identify distinct regimes of network model behavior as a function of spatial scale.
• To provide insights for comparing macro- to mesoscale brain network studies.

Main Methods:

• Acquired magnetic resonance and diffusion tensor images from 136 healthy adults.
• Parceled each subject's cortex into up to 50,000 regions, represented as network nodes.
• Inferred interregional connectivity using deterministic tractography and explored network behavior with varying node numbers and edge weights.

Main Results:

• Identified three distinct quantitative behavior regimes in network models related to spatial scale.
• Observed that network model properties significantly vary with vertex assignment and edge weighing schemes.
• Found that the spatial folding scale of the cortex may modulate these network behaviors.

Conclusions:

• Graph-theoretic analysis results are scale-dependent and influenced by model topology.
• Direct comparison of network analysis results across different spatial scales requires careful consideration of these modulations.
• Choosing appropriate network-theoretic parameters is critical for accurate interpretation of brain connectivity studies.