Karmacharya S, Gagoski B, Ning L, Vyas R, Cheng HH, Soul J, Newberger JW, Shenton ME, Rathi Y, Grant PE
Neuroimage Clin 2018;19:360-373
PMID: 30013919
Abstract
Background: Elucidating developmental trajectories of white matter (WM) microstructure is critically important for understanding normal development and regional vulnerabilities in several brain disorders. Diffusion Weighted Imaging (DWI) is currently the method of choice for in-vivo white matter assessment. A majority of neonatal studies use the standard Diffusion Tensor Imaging (DTI) model although more advanced models such as the Neurite Orientation Dispersion and Density Imaging (NODDI) model and the Gaussian Mixture Model (GMM) have been used in adult population. In this study, we compare the ability of these three diffusion models to detect regional white matter maturation in typically developing control (TDC) neonates and regional abnormalities in neonates with congenital heart disease (CHD).
Methods: Multiple b-value diffusion Magnetic Resonance Imaging (dMRI) data were acquired from TDC neonates (N = 16) at 38 to 47 gestational weeks (GW) and CHD neonates (N = 19) aged 37 weeks to 41 weeks. Measures calculated from the diffusion signal included not only Mean Diffusivity (MD) and Fractional Anisotropy (FA) derived from the standard DTI model, but also three advanced diffusion measures, namely, the fiber Orientation Dispersion Index (ODI), the isotropic volume fraction (V), and the intracellular volume fraction (V) derived from the NODDI model. Further, we used two novel measures from a non-parametric GMM, namely the Return-to-Origin Probability (RTOP) and Return-to-Axis Probability (RTAP), which are sensitive to axonal/cellular volume and density respectively. Using atlas-based registration, 22 white matter regions (6 projection, 4 association, and 1 callosal pathways bilaterally in each hemisphere) were selected and the mean value of all 7 measures were calculated in each region. These values were used as dependent variables, with GW as the independent variable in a linear regression model. Finally, we compared CHD and TDC groups on these measures in each ROI after removing age-related trends from both the groups.
Results: Linear analysis in the TDC population revealed significant correlations with GW (age) in 12 projection pathways for MD, V, RTAP, and 11 pathways for RTOP. Several association pathways were also significantly correlated with GW for MD, V, RTAP, and RTOP. The right callosal pathway was significantly correlated with GW for V. Consistent with the pathophysiology of altered development in CHD, diffusion measures demonstrated differences in the association pathways involved in language systems, namely the Uncinate Fasciculus (UF), the Inferior Fronto-occipital Fasciculus (IFOF), and the Superior Longitudinal Fasciculus (SLF). Overall, the group comparison between CHD and TDC revealed lower FA, V RTAP, and RTOP for CHD bilaterally in the a) UF, b) Corpus Callosum (CC), and c) Superior Fronto-Occipital Fasciculus (SFOF). Moreover, FA was lower for CHD in the a) left SLF, b) bilateral Anterior Corona Radiata (ACR) and left Retrolenticular part of the Internal Capsule (RIC). V was also lower for CHD in the left Posterior Limb of the Internal Capsule (PLIC). ODI was higher for CHD in the left CC. RTAP was lower for CHD in the left IFOF, while RTOP was lower in CHD in the: a) left ACR, b) left IFOF and c) right Anterior Limb of the Internal Capsule (ALIC).
Conclusion: In this study, all three methods revealed the expected changes in the WM regions during the early postnatal weeks; however, GMM outperformed DTI and NODDI as it showed significantly larger effect sizes while detecting differences between the TDC and CHD neonates. Future studies based on a larger sample are needed to confirm these results and to explore clinical correlates.
Comments are closed.