Language impairment is a common comorbidity in childhood epilepsy and may reflect disruption of distributed brain networks that support language. In this study, investigators used high-density EEG to measure functional connectivity among bilateral frontal, temporal, occipital, and motor regions. Multivariate pattern analysis revealed that connectivity within frontal and occipitotemporal networks, particularly in the left hemisphere, predicted language ability across both children with epilepsy and age-matched controls. These findings indicate that language performance is associated with spatially specific connectivity patterns rather than only global measures of network function.
Although some connectivity patterns predicted language ability across groups, the study identified disease-specific alterations in children with self-limited epilepsy with centrotemporal spikes (SeLECTS). Connectivity involving the motor network emerged as the dominant SeLECTS-specific predictor of language dysfunction. This result links the epileptogenic network characteristic of SeLECTS to impaired language function and highlights the motor network as a potential focal point for understanding how epilepsy disrupts language-related circuitry.
The cohort comprised 27 children with SeLECTS and 29 age-matched control participants. High-density EEG was recorded during a verb-generation task and during resting state. Functional connectivity was quantified across multiple frequency bands and among anatomically defined bilateral regions (frontal, temporal, occipital, and motor). The use of both task-based and resting-state recordings allowed the authors to compare which context provided stronger predictive information about language ability.
Connectivity measures across frequency bands and interregional connections served as input for multivariate pattern analysis. This analytic approach identified patterns that predicted individual language ability and allowed separation of connectivity patterns that were shared by both groups from those specific to SeLECTS. The multivariate framework enabled assessment of whether spatially specific interregional connections contributed information beyond aggregated measures.
The authors tested whether spatially specific connectivity provided added value compared with whole-brain or hemispheric averaged connectivity and with conventional clinical variables. Connectivity between specific regions outperformed averaged connectivity measures in predicting language ability. Moreover, these region-pair connectivity patterns predicted language beyond epilepsy diagnosis and antiseizure medication use, indicating that focal network measures carry unique explanatory power for language outcomes.
Task-based connectivity recorded during the verb-generation task provided stronger predictive power for language ability than resting-state connectivity. This suggests that measures obtained during active language engagement better capture the functional dynamics relevant to language performance and may reveal disease-specific disruptions that are less apparent at rest.
The findings show that language ability in children is associated with distributed yet spatially specific patterns of brain connectivity, while epilepsy introduces distinct alterations centered on the epileptogenic network. By identifying disease-specific network patterns—most notably motor-network connectivity in SeLECTS—this work provides a mechanistic basis for language dysfunction in pediatric epilepsy and supports the concept of spatially targeted neuromodulation as a rational therapeutic strategy. The results suggest that targeting connectivity alterations at specific regions or interregional connections, rather than applying nonspecific or whole-hemisphere approaches, could be a more precise way to remediate language deficits in epilepsy.
This report is a preprint and has not been peer reviewed. The source provides cohort sizes (27 children with SeLECTS and 29 controls), recording contexts (verb generation task and rest), the general analytic approach (multivariate pattern analysis of functional connectivity across bands and regions), and the principal findings described above. Details that were not reported in the source text provided here include precise frequency bands analyzed, specific statistical values or effect sizes, the exact regional pairs that were most predictive, parameter settings for multivariate models, and participant demographic breakdown beyond age matching. These details may be available in the full preprint PDF or supplementary materials but were not stated in the source excerpt presented.