Abstracts

Rationale: Topological disorganization of brain connectivity in TLE patients has been revealed by network analyses of functional and anatomical connections. The analysis of cortical thickness correlation provides limited network information, as the model prevents the inclusion of subcortical structures (i.e. hippocampi). To investigate the topological organization of a comprehensive anatomical network (with cortical/subcortical structures) we used graph theoretical analysis of the brain volumetric network constructed from regional volumes in TLE patients with unilateral hippocampal sclerosis. Methods: Whole brain T1 images were acquired from 86 left TLE (48±10yrs, 35 males) and 70 right TLE (37±12yrs, 25 males) patients and 116 controls (46±12yrs, 39 males) on a 3T Philips scanner. Using Freesurfer5.3, native images were processed/parcellated into 80 cortical/subcortical regions (Table 1), automatically providing volumes of those regions. For each group, we calculated the Pearson correlation coefficient between any pairs of regions in volume across individuals and obtained cross-correlation matrix (80x80) after removing the effects of age, gender and total brain volume by regression. The absolute of the matrices were then thresholded into binarized connectivity matrices for all groups. We computed the global efficiency (gE), local efficiency (locE) and clustering coefficient (CC) over a wide range of network sparsity (5-40%). Last, we identified hub regions for all networks and examined epilepsy-related changes in the inter-regional connectivity and regional centrality in both TLE groups. Results: Compared with controls, both LTLE and RTLE groups demonstrated significantly decreased gE, with LTLE decreasing over a much wider sparsity range (Fig. 1A). RTLE demonstrated significant increases in both locE and CC while LTLE only showed a trend in the CC (Fig. 1B/C). Controls presented hub nodes distributed evenly over hemispheres. In contrast, TLE groups concentrated the hub nodes primarily in the paralimbic/limbic and temporal association cortices. Compared with controls, both TLE groups (Table 1) presented reduced cortical/subcortical connectivity including bilateral hippocampi (Fig. 1D). Interestingly, most significant inter-regional correlation increases occurred within the limbic system and contralateral hemisphere in LTLE and RTLE groups, respectively. Regions with increased centrality are mostly located in occipital lobes for LTLE patients and contralateral limbic/temporal associated areas for RTLE patients. Conclusions: Both TLE groups presented decreased global network efficiency of the brain volumetric network; however, only RTLE patients displayed significant increase in the local network efficiency, suggesting a better underlying compensation mechanism. Pathological distribution of hub nodes and alterations of inter-regional/regional connectivity involves mainly paralimbic/limbic and temporal association areas, pointing to a structural reorganization of the temporolimbic network. Our results provide first evidence of disrupted topological organization of the brain volumetric network in TLE patients.