Abstracts

Rationale: To investigate brain functional topology in patients with epilepsy, graph analysis based on anatomical defined graph were applied to resting-state fMRI data. Small-world(SW) properties as well as measures of integration, and centrality were compared between subjects with epilepsy (EPI) and healthy controls (HC). Methods: Thirty-one EPI subjects (age 40±12 years, 19 female) and thirty-eight age and gender matched HC (age 39.8±16.2 years, 23 female) participated in the study. Eyes-closed resting-state fMRI scans were collected along with high resolution T1 weighted anatomical scans on GE MRI scanners (3.0T; 1.5 T for 13 patients). MRI Scanner was considered as a covariate factor in all statistical analysis for consistency of results. 106 anatomically defined regions were used to derive Pearson correlation coefficient matrices from each subject’s R-S fMRI using AFNI. By using minimum spanning tree (to guarantee connectedness) combined with proportional thresholding on these matrices, Binary undirected connection matrices for each subject was obtained and it then were used to study networks’ measurements . Clustering coefficient C and average shortest path length L were calculated and compared with a random network with the same number of nodes and degree with 500 re-wiring for each subject, resulting in a clustering ratio C/Cr and shortest path length ratio L/Lr[Maslov S, Sneppen K. Specificity and stability in topology of protein networks. Science. 2002;296:910–913.]. SW was then calculated via dividing C/Cr by L/Lr [Humphries MD, Gurney K. Network ‘small-world-ness’: a quantitative method for determining canonical network equivalence. PLoS ONE. 2008;3:e0002051.]. Local efficiency and eigenvector centrality were used to investigate network integration and hubs respectively at the individual regions level. SW was calculated in a range of 2% to 16% and local measures were calculated in 16% of graph density [Bullmore E, Sporns O (2012) The economy of brain network organization. Nature Reviews Neuroscience 13: 336–349.]. All statistical tests were corrected with Bonferroni correction. Results: At the global level, SW was higher in EPI compared to HC in all densities which indicates that brain network of EPI is closer to random network and is less segregated. Particularly SW in EPI was significantly higher at 16% network density.In the regional level, EPIs show lower local efficiency in frontal and temporal regions and higher local efficiency in subcortical and cerebellar regions. Particularly, in superior temporal, mid temporal and left supplementary motor cortex local efficiency were significantly lower for EPI than HC, while in bilateral putamen, right amygdala and cerebellar regions EPIs show higher local efficiency.Eigenvector centrality in most of subcortical and cerebellar regions were higher in EPI compared to HC suggesting that these regions communicate with more important regions of network in EPI compare to HCs’ network. Conclusions: EPIs’ brain network was significantly less segregated than controls, as an indication of less robust network. Higher SW shows that information circulates faster in the network ,causes it to be less specialized for functional processes. Higher local efficiency and centrality in cerebellum and subcortical regions in EPIs indicate higher interaction between theses regions and other parts of the brain in EPI compare to HC. Funding: NIH grants NINDS3R01NS044351-09S1, 1K23NS086852-01A1, 1R01 NS081926-01.