
Contact the VUNO team anytime.
International Child Neurology Congress
Background: Neurofibromatosis type 1 (NF1) is frequently associated with neurocognitive dysfunction (ND). While brain morphometry identifies structural biomarkers, conventional methods are time-consuming and often lack pediatric-specific templates. We evaluated a deep learning–based segmentation (DLS) tool and characterized structural brain signatures in pediatric NF1 patients.
Methods: Brain MRIs from 28 pediatric NF1 patients (3–18 years) and 56 age-matched controls were analyzed using DLS and a surface–based morphometry tool. We assessed agreement using intraclass correlation coefficients (ICC) and compared structural differences among ND, non-ND, and control groups.
Results: DLS showed excellent concordance with the surface–based morphometry tool for whole-brain and subcortical volumes (ICC > 0.90), though cortical thickness demonstrated method-dependent variability (ICC < 0.50). NF1 patients exhibited significantly enlarged total intracranial (+10.9%), cerebral white matter (+20.5%), and subcortical volumes (e.g., thalamus, hippocampus) compared to controls (p < 0.05). While no global differences were observed between ND and non-ND groups, the DLS pipeline identified significantly larger left rostral anterior cingulate volumes in the ND group.
Conclusions: Automated DLS provides a rapid and reliable alternative for capturing macrostructural alterations in pediatric NF1, facilitating clinical monitoring and large-scale research. However, the divergent sensitivities in regional measures suggest that multimodal approaches remain essential to fully elucidate the neural basis of neurocognitive outcomes in NF1.