Imaging structural and functional brain networks in temporal lobe epilepsy

Abstract

Early imaging studies in temporal lobe epilepsy (TLE) focused on the search for mesial temporal sclerosis, as its surgical removal results in clinically meaningful improvement in about 70% of patients. Nevertheless, a considerable subgroup of patients continues to suffer from post-operative seizures. Although the reasons for surgical failure are not fully understood, electrophysiological and imaging data suggest that anomalies extending beyond the temporal lobe may have negative impact on outcome. This hypothesis has revived the concept of human epilepsy as a disorder of distributed brain networks. Recent methodological advances in non-invasive neuroimaging have led to quantify structural and functional networks in vivo. While structural networks can be inferred from diffusion MRI tractography and inter-regional covariance patterns of structural measures such as cortical thickness, functional connectivity is generally computed based on statistical dependencies of neurophysiological time-series, measured through functional MRI or electroencephalographic techniques. This review considers the application of advanced analytical methods in structural and functional connectivity analyses in TLE. We will specifically highlight findings from graph-theoretical analysis that allow assessing the topological organization of brain networks. These studies have provided compelling evidence that TLE is a system disorder with profound alterations in local and distributed networks. In addition, there is emerging evidence for the utility of network properties as clinical diagnostic markers. Nowadays, a network perspective is considered to be essential to the understanding of the development, progression, and management of epilepsy.

Keywords: MRI; TLE; connectivity; connectome; graph-theory.

Figure 1

Schematic illustration of gray matter structural anomalies in temporal lobe epilepsy. (top) Results from MRI-based cortical thickness analysis, showing cortical thinning in left temporal lobe epilepsy (TLE) patients with hippocampal atrophy relative to healthy controls in mesial and lateral temporal as well as fronto-central neocortices (Bernhardt et al., 2010). (lower left) Patterns of atrophy in TLE patients relative to controls in CA1 subregions of the ipsilateral hippocampus (Kim et al., 2008) and (lower right) mediodorsal segments of the thalamus (Bernhardt et al., 2012), both generated using spherical harmonic surface-shape modeling techniques of manual MRI segmentations. The shown analyses have been generated using the SurfStat toolbox for Matlab (Worsley et al., 2009). Further details on the statistical procedures can be found in the original publications.

Figure 2

Assessment of inter-regional connectivity. (A) Diffusion tractography. Left: Illustration of diffusion tensor directions superimposed on a fractional anisotropy map derived from diffusion MRI. Right: Seed-based deterministic tractography of the uncinate fasciculus. (B) Altered mean diffusivity along the uncinate fasciculus tract in a group of patients with temporal lobe epilepsy (TLE) relative to controls (Concha et al., 2012). Prior to analysis, the tract was subdivided into bins with respect to the anatomical distance to the temporal and frontal lobes. (C) Structural covariance analysis. Shown is the cortical thickness correlation map of the left medial orbital cortex seed with the remaining cortical mantle in a group of healthy controls. High positive correlations are interpreted as connections, low correlations as absence of connections. (D) Structural covariance alterations in TLE patients relative to controls between an entorhinal cortex seed and target regions in medial orbitofrontal cortices (Bernhardt et al., 2008). (E) Functional connectivity between a left medial orbitofrontal cortex seed and the rest of the cortical mantle in healthy controls. Insets show exemplary time courses of the seed region with selected cortical target regions with high and low correlations, respectively. (F) Voxel-wise functional connectivity abnormalities in TLE, highlighting target regions with altered time-series correlation to a spatial component that closely matches the “default mode” network (Voets et al., 2012).

Figure 3

Graph-theoretical analysis. (A) Association matrix quantifying the degree of connectivity (derived from techniques such as diffusion MRI tractography, structural covariance, or functional connectivity, see Figure 2). (B) Matrices commonly undergo a thresholding procedure to remove spurious edges. In this example, the highest 25% of associations are preserved, leading to a binary adjacency matrix. (C) Each binary matrix is equivalent to an undirected graph. (D) Topological parameters such as the clustering coefficient and path length can then be measured; networks can be partitioned in modules based on groupings of connectivity among nodes; hubs can be identified, for example, as nodes with high degree centrality (i.e., a high number of connections). (E) Cortical degree centrality map, based on resting-state functional connectivity data from a single healthy control subject.