Authors :
Presenting Author: Asmaa Mhanna, MD – University of Iowa
Joseph C. Griffis, PhD – University of Iowa
Joel Bruss, BA – University of Iowa
Christine K. Fox, MD – University of California, San Francisco
Aaron D. Boes, MD, PhD – University of Iowa
Rationale:
Pediatric ischemic stroke is a rare but serious condition with long-term neurological consequences, including a risk of developing post-stroke epilepsy (PSE). Recent work implicates lesion location and brain network disruptions in PSE in adults—specifically, stroke lesions that are anticorrelated to the globus pallidus and cerebellar vermis have been associated with increased epilepsy risk1—these associations remain unexplored in pediatric stroke populations. To address this gap, we aimed to determine whether lesion connectivity to PSE networks identified in adult stroke patients is associated with PSE in pediatric stroke patientsMethods:
: We analyzed neuroimaging data from 333 children enrolled in the Vascular Effects of Infection in Pediatric Stroke (VIPS I) study2. Lesion network mapping was used to evaluate each lesion’s connectivity to adult-derived structural and functional networks previously associated with PSE. We used independent samples unequal variance t-tests to compare mean intra-lesion network connectivity scores between children who developed epilepsy (n = 43) and those who did not (n = 290), and we used logistic regression to evaluate whether lesion-connectivity scores provide unique predictive power when accounting for other known risk factors for PSE: age, lesion size, and the presence of acute seizures.
Results:
: Lesion-network connectivity scores significantly differed between children who developed epilepsy and children who did not for both the structural (t[60.18]=4.19, p< 0.001) and functional (t[65.11]=5.29, p< 0.001) adult-derived post-stroke epilepsy networks. The fit of logistic regression models that included lesion volume, age, and presence of acute seizures as predictors was significantly improved when functional (p=0.01), structural (p=0.04), or structural and functional (p=0.01) connectivity scores were added to the models, and the addition of connectivity scores also improved classification performance in cross-validation analyses.