345 - Brain Diffusion MRI Biomarkers for Predicting Neurodevelopmental Disorders in High-Risk Preterm Infants
Friday, April 25, 2025
5:30pm – 7:45pm HST
Publication Number: 345.4410
Lillian Swanz, University of Cincinnati College of Medicine, Covington, KY, United States; Mitchell A. Batschelett, University of Cincinnati College of Medicine, Cincinnati, OH, United States; Armin Allahverdy, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States; Nehal A. Parikh, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States
Professor of Pediatrics Cincinnati Children's Hospital Medical Center Cincinnati, Ohio, United States
Background: 30-40% of very preterm birth survivors develop one or more neurodevelopmental disorders (NDD) such as cognitive impairments, sensorimotor abnormalities, and behavioral abnormalities. There are several known predictors and models of NDDs; however, none are sufficiently accurate to diagnose NDDs before definitive motor and cognitive testing at 2-3 years of age. Diffusion magnetic resonance imaging (dMRI) is a promising method for detecting microstructural brain abnormalities that are invisible on conventional MRI and for delineating white matter tracts in vivo. Objective: Our objectives were to 1) evaluate the value of fractional anisotropy (FA) of automatically segmented corpus callosum (CC) and corticospinal tracts (CST) to predict motor and cognitive outcomes and 2) perform manual tractography of those tracts to validate automatic segmentation. Design/Methods: This is a secondary study of preterm infants born at or before 32 weeks gestational age from the multisite, prospective Cincinnati Infant Neurodevelopmental Early Prediction Study (CINEPS) cohort. All infants were scanned between 40-44 weeks postmenstrual age (PMA) at Cincinnati Children’s Hospital Medical Center. We used TractSeg, a tractography software developed for adult brains, to automatically segment the CC and CST. Bayley-III motor assessment was done at 2 years corrected age (CA), and Differential Abilities Scales-II (DAS-II) cognitive testing was done at 3 years CA. 280 CINEPS infants with full datasets were used for multivariable linear regression analysis to evaluate the value of CC and CST FA to predict motor and cognitive scores. After training two raters, they randomly selected 40 infants and performed manual tractography using MRtrix3, then repeated segmentation after 2 weeks. We used intraclass correlation coefficients (ICC) to assess reliability of automated tractography and ICC and Dice Similarity Index to assess manual tractography. Results: Independent of other known predictors, FA of the right CST was predictive of cognitive but not motor outcomes at 2-3 years CA (Table 1). FA of the CC and left CST were not predictive of motor or cognitive outcomes. Manual tractography of the CC and CST exhibited high intrarater and interrater reliability (Table 2). Comparison of automated to manual tractography also demonstrated high ICCs (Table 3).
Conclusion(s): FA of the right CST, automatically determined from dMRI at term-equivalent age, was predictive of DAS-II cognitive score at 3 years CA in very preterm infants. TractSeg reliably segmented the CC and CST in neonates, but further studies are needed to validate this software.
Table 1. Results of multivariable linear regression for the R corticospinal tract (CST) fractional anisotropy (FA), the only track which showed any significant correlation with motor or cognitive outcomes. Table 1_CSTBayleyOutcomes.pdf
Table 2. Intra- and inter-rater reliability of manually segmented FDC values (a surrogate marker that incorporates fiber density and fiber cross-section) from the corpus callosum, and left & right corticospinal tracts. Table2_Reliability_ManualTractograph_LillyPAS'25.pdf
