Delayed convergence between brain network structure and function in rolandic epilepsy

Abstract

Introduction: Rolandic epilepsy (RE) manifests during a critical phase of brain development, and has been associated with language impairments. Concordant abnormalities in structural and functional connectivity (SC and FC) have been described before. As SC and FC are under mutual influence, the current study investigates abnormalities in the SC-FC synergy in RE.

Methods: Twenty-two children with RE (age, mean ± SD: 11.3 ± 2.0 y) and 22 healthy controls (age 10.5 ± 1.6 y) underwent structural, diffusion weighted, and resting-state functional magnetic resonance imaging (MRI) at 3T. The probabilistic anatomical landmarks atlas was used to parcellate the (sub)cortical gray matter. Constrained spherical deconvolution tractography and correlation of time series were used to assess SC and FC, respectively. The SC-FC correlation was assessed as a function of age for the non-zero structural connections over a range of sparsity values (0.01-0.75). A modularity analysis was performed on the mean SC network of the controls to localize potential global effects to subnetworks. SC and FC were also assessed separately using graph analysis.

Results: The SC-FC correlation was significantly reduced in children with RE compared to healthy controls, especially for the youngest participants. This effect was most pronounced in a left and a right centro-temporal network, as well as in a medial parietal network. Graph analysis revealed no prominent abnormalities in SC or FC network organization.

Conclusion: Since SC and FC converge during normal maturation, our finding of reduced SC-FC correlation illustrates impaired synergy between brain structure and function. More specifically, since this effect was most pronounced in the youngest participants, RE may represent a developmental disorder of delayed brain network maturation. The observed effects seem especially attributable to medial parietal connections, which forms an intermediate between bilateral centro-temporal modules of epileptiform activity, and bear relevance for language function.

Keywords: brain maturation; functional connectivity; graph theory; rolandic epilepsy; structural connectivity; structure-function correlation.

Figure 1

Assessing the correlation between structural and functional connectivity (SC and FC, respectively). The most prominent structural connections (here at sparsity s = 0.04) are selected. The SC and FC values of these are appended in two vectors, which are subsequently correlated. For visualization purposes, the vectors are not visualized in full length. SC values were scaled to a Gaussian distribution (mean, SD: 0.5, 0.1); negative FC values were set to 0.

Figure 2

Sub-networks as derived from a modularity analysis of the mean structural connectivity (SC) network of the controls. Five sub-networks (i.e., modules) were found; a prefrontal cluster (blue), a medial parietal cluster (red), an inferior temporo-occipital cluster (green), and a left and a right centro-temporal cluster (yellow and magenta, respectively). For visualization purposes, only the within-module connections were visualized, at sparsity level 0.1.

Figure 3

At the whole-brain level, the correlation between structural and functional connectivity is reduced in patients vs. controls for the sparsity range 0.01–0.11 (A, top). Within this sparsity range, similar effects were found in all modules; effects of highest significance were found within the medial parietal module (A, module 2). For sparsity value 0.04, the connections investigated and the regression lines are given in (B) and (C) for the whole brain and the medial parietal module, respectively. Note that the reduction is SC-FC correlation is more pronounced in the medial parietal module. Furthermore, significant increases in SC-FC correlation with age were found in the patients only (solid lines).

Figure 4

Structural and functional connectivity (SC and FC, respectively) were investigated separately by assessing the normalized path length (λ) and clustering coefficient (CC). A decrease in SC clustering coefficient was found for sparsity value 0.04, and an increase in FC path length for sparsity value 0.08.