Structural connectome with high angular resolution diffusion imaging MRI: assessing the impact of diffusion weighting and sampling on graph-theoretic measures.

PURPOSE: Advances in computational network analysis have enabled the characterization of topological properties of human brain networks (connectomics) from high angular resolution diffusion imaging (HARDI) MRI structural measurements. In this study, the effect of changing the diffusion weighting (b value) and sampling (number of gradient directions) was investigated in ten healthy volunteers, with specific focus on graph theoretical network metrics used to characterize the human connectome.

METHODS: Probabilistic tractography based on the Q-ball reconstruction of HARDI MRI measurements was performed and structural connections between all pairs of regions from the automated anatomical labeling (AAL) atlas were estimated, to compare two HARDI schemes: low b value (b = 1000) and low direction number (n = 32) (LBLD); high b value (b = 3000) and high number (n = 54) of directions (HBHD).

RESULTS: LBLD and HBHD data sets produced connectome images with highly overlapping hub structure. Overall, the HBHD scheme yielded significantly higher connection probabilities between cortical and subcortical sites and allowed detecting more connections. Small worldness and modularity were reduced in HBHD data. The clustering coefficient was significantly higher in HBHD data indicating a higher level of segregation in the resulting connectome for the HBHD scheme.

CONCLUSION: Our results demonstrate that the HARDI scheme as an impact on structural connectome measures which is not automatically implied by the tractography outcome. As the number of gradient directions and b values applied may introduce a bias in the assessment of network properties, the choice of a given HARDI protocol must be carefully considered when comparing results across connectomic studies.