AUTHOR=Yan Chao-Gan , Craddock R. Cameron , He Yong , Milham Michael P. TITLE=Addressing head motion dependencies for small-world topologies in functional connectomics JOURNAL=Frontiers in Human Neuroscience VOLUME=7 YEAR=2013 URL=https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2013.00910 DOI=10.3389/fnhum.2013.00910 ISSN=1662-5161 ABSTRACT=

Graph theoretical explorations of functional interactions within the human connectome, are rapidly advancing our understanding of brain architecture. In particular, global and regional topological parameters are increasingly being employed to quantify and characterize inter-individual differences in human brain function. Head motion remains a significant concern in the accurate determination of resting-state fMRI based assessments of the connectome, including those based on graph theoretical analysis (e.g., motion can increase local efficiency, while decreasing global efficiency and small-worldness). This study provides a comprehensive examination of motion correction strategies on the relationship between motion and commonly used topological parameters. At the individual-level, we evaluated different models of head motion regression and scrubbing, as well as the potential benefits of using partial correlation (estimated via graphical lasso) instead of full correlation. At the group-level, we investigated the utility of regression of motion and mean intrinsic functional connectivity before topological parameters calculation and/or after. Consistent with prior findings, none of the explicit motion-correction approaches at individual-level were able to remove motion relationships for topological parameters. Global signal regression (GSR) emerged as an effective means of mitigating relationships between motion and topological parameters; though at the risk of altering the connectivity structure and topological hub distributions when higher density graphs are employed (e.g., >6%). Group-level analysis correction for motion was once again found to be a crucial step. Finally, similar to recent work, we found a constellation of findings suggestive of the possibility that some of the motion-relationships detected may reflect neural or trait signatures of motion, rather than simply motion-induced artifact.