Altered white matter structural network in self-limited epilepsy with centrotemporal spikes

The changes in white matter (WM) structural networks associated with self-limited epilepsy with centrotemporal spikes (SeLECTS) remain poorly understood, despite the established role of WM network changes in elucidating pathological mechanisms underlying epileptic disorders. We aimed to explore the structural network alterations in SeLECTS, and analyze the relationship between structural networks and disease status.

This study enrolled 64 participants, comprising 32 children with SeLECTS and 32 age- and sex-matched typically developing (TD) children. Graph-theoretic methods were applied to characterize the topological organization of WM structural networks derived from diffusion tensor imaging (DTI). Intergroup comparisons of global and nodal network metrics were performed using non-parametric tests with false discovery rate (FDR) correction. Correlation analyses were conducted to examine relationships between network alterations and clinical variables, including seizure duration and intelligence quotients (IQs). Additionally, a support vector machine (SVM) classifier was trained using network metrics to distinguish patients from controls, with model performance evaluated on an independent test set using accuracy (ACC), sensitivity (SEN), specificity (SPE), and area under the curve (AUC).

Compared with controls, children with SeLECTS exhibited disrupted global network organization, characterized by significantly reduced global efficiency (z=-5.05, P<0.001, FDR-corrected) and increased shortest path length (z=4.58, P<0.001, FDR-corrected). At the nodal level, decreased nodal efficiency (E_nodal ) was observed in the left precentral gyrus (PreCG.L), opercular part of left inferior frontal gyrus (IFGoperc.L), left Rolandic operculum (ROL.L), and left superior temporal gyrus (STG.L). Seizure duration was negatively correlated with E_nodal in ROL.L (r=-0.57, P<0.001) and STG.L (r=-0.52, P<0.001). E_nodal in STG.L showed positive correlations with full-scale (r=0.53, P<0.001) and verbal IQs (r=0.64, P<0.001). The SVM classifier achieved 78.13% ACC, 75.00% SEN, 81.25% SPE, and an AUC of 0.82 in distinguishing patients from TDs.

SeLECTS presents with topological changes, which were associated with duration of epilepsy and neuropsychological score, and WM structural network might help clinical classification for children with SeLECTS. The STG.L may might be a suitable brain area for monitoring the disease condition.