Whole-brain structural connectivity in temporal lobe epilepsy: a Diffusion Spectrum Imaging study
Introduction: Medial Temporal Lobe Epilepsy (mTLE) with hippocampal sclerosis is the most frequent cause of drug-resistant focal epilepsy in adults. These patients suffer from widespread subtle white matter abnormalities and abnormal functional connectivity extending beyond the affected lobe, as revealed by Diffusion Tensor Imaging and functional MRI studies. Diffusion spectrum imaging (DSI) is a new diffusion imaging technique with high angular resolution for improving the mapping of white matter tracts . In this study, we used DSI to investigate the whole brain effect of TLE as reflected by topological measures of the structural connectivity network. Methods: Eight patients with right-sided mTLE and hippocampal sclerosis and 25 controls underwent our DSI protocol at 3T using a 32 channel coil: DSI (voxel size 2.2x2.2x3mm; 44 slices, 257 diffusion directions; max b-value = 6400 s/mm2), 3D T1-weighted MPRAGE and 3D T2-weighted images. The analysis was performed using the Connectome Mapper. The cortical and subcortical grey matters were parcellated into 86 Regions of Interest (ROI) with anatomical landmarks using Freesurfer 5.0. The ROIs were coregistered to the diffusion data by using a nonlinear registration between T1 and T2, then T2 to the diffusion space. The Diffusion Tool Kit was used for the reconstruction of the diffusion Orientation Distribution Function in each voxel. Streamline fibre-tracking was then performed from each voxel in the white matter areas using an in-house streamline-based algorithm. Two scalar maps, the Fractional Anisotropy and the Generalized Fractional Anisotropy were calculated. For each patient, the connectivity between every region pair was estimated was used to construct a connectivity matrix (the adjacency matrix of the structural network). Finally we used topological measures in order to reduce the data and obtain whole-brain characteristics of the network : characteristic Path Length (L), Clustering Coefficient (C) and Small Worldness (S) where S=(C/Crand)/(L/ Lrand) with Crand and Lrand are computed for a random network. Significant differences between the mTLE and control groups were assessed using a Wilcoxon rank-sum test. Results: In patients, we found a higher characteristic path length (0.0288), with a lower clustering coefficient (0.0079) and reduced small-worldness (p=0.000058) compared to controls. Network nodes that contributed significantly to these alterations were located within as well as outside the temporal lobe in both hemispheres, although with a predominance for the ipsilateral hemisphere and bilateral limbic structures. The pattern of significant nodes showed concordance with brain regions known to be involved in epileptic networks in patients with mTLE. Conclusions: Our results show a reduced efficiency of structural brain networks and altered connectivity patterns that are concordant with the mapping of functional epileptic networks  and altered functional connectivity  in patients with mTLE. Therefore, the network in mTLE is less segregated than in controls with reduced global interactions between the nodes and reduced specialised communities. Further studies are needed to establish the relevance of these findings with respect to the propagation of epileptic activity, cognitive deficits in mTLE and outcome of epilepsy surgery in individual patients.