Project Details

Advancing Neurodevelopmental Prognosis in Congenital Heart Disease: AI-Driven Methods Using Cortical Sulcal Patterns

Hyeokjin Kwon, PhD

Summary

BACKGROUND: Congenital heart disease (CHD), the leading birth defect, is associated with disrupted neurogenesis and lifelong neurodevelopmental disabilities (NDD), underscoring the need for early biomarkers for risk stratification. The spatial organization of early cortical sulcal folds, shaped by functional specialization and connectivity of cerebral cortex, and influenced by genetic predisposition, emerges before the third trimester, and remains stable with age, suggesting that it holds potential as a biologically meaningful biomarker for identifying NDD risk in CHD. GAP: While recent studies have revealed group-level associations between NDD and altered spatiotemporal patterns of cortical sulcal folds in CHD, there remains an unmet need to develop reliable and unbiased biomarkers that can predict individual-level risk for NDD. HYPOTHESIS: Our AI model will predict the risk for NDD in individual CHD patients with high accuracy (significantly high correlation with low mean relative absolute errors < 5%, between prediction and true measures) and identify prototypes of sulcal pattern abnormalities varying by cortical functional region corresponding to specific ND outcomes and CHD subtypes including single ventricle, transposition of the great arteries, and tetralogy of Fallot. METHODS: In this retrospective study, to capture the complex inter-relationships of primary sulcal folds, we will construct sulcal pattern graphs from individual MRI. Using these, we will develop an AI model to predict individual ND outcomes through prototype-based graph neural networks. RESULTS: Pending. IMPACT: Our AI model could provide an imaging biomarker for early prediction of NDD in CHD. It could enable timely, targeted intervention and enhance child ND outcomes.