Strong intercorrelations among global graph-theoretic indices of structural connectivity in the human brain

Area of Science:

• Neuroimaging
• Network Neuroscience
• Computational Psychiatry

Background:

• Graph-theoretic metrics from neuroimaging data are used to study brain organization.
• These metrics aim to reveal neural mechanisms underlying psychological traits and disorders.

Purpose of the Study:

• To assess the distinctiveness of 11 global graph-theoretic metrics in human structural connectomes.
• To determine the extent to which these metrics capture unique interindividual differences in brain organization.

Main Methods:

• Analysis of N=8,185 human structural connectomes from UK Biobank.
• Utilized unthresholded, FA-weighted networks.
• Employed sensitivity and simulation analyses to examine metric overlap and influence of network properties.

Main Results:

• Most graph-theoretic metrics (except Participation Coefficient) were highly intercorrelated (mean |r|=0.788).
• Metrics also strongly correlated with mean edge weight (mean |r|=0.873).
• Metric overlap was influenced by network sparseness and edge weight variation; specific metrics were nearly collinear with mean edge weight.

Conclusions:

• Global graph-theoretic measures may capture largely overlapping information about brain organization.
• Individual differences in these global metrics might be difficult to separate from mean edge weight.
• The utility of global graph metrics for indexing separable components of interindividual variation may be limited.