Brain Ventricular and Microstructural Correlates of Executive Dysfunction in Congenital Heart Disease Using Explainable Machine Learning

This study examined cerebrospinal fluid volumes as neuroimaging features and their role in predicting specific executive function impairments among adolescents with congenital heart disease using explainable machine learning models. The findings showed CSF volumes were among the most important predictors of executive function inhibition domain with 3 CSF volumes ranked amongst the top 20, and 4 more CSF volumes among the top 20% of all features. Selective increased lateral ventricular volume in the frontal regions in CHD patients may be secondary to loss of white matter integrity in the uncinate fasciculus (emotional regulation) and subsequently lead to inhibitory dysfunction.