Extremely preterm birth is a significant public health concern. Common sequelae of preterm birth include cognitive, learning and behavioural impairments in childhood that persist throughout adulthood[1]. These impairments have been linked to alteration in the white matter microstructure in the brain [2, 3], measured with diffusion MRI. In this study, we extend the study of diffusion-tensor parameters (for instance, Fractional Anisotropy [FA]) in the white matter to Neurite Orientation Dispersion and Density Imaging (NODDI) [4]. We investigate the known differences in FA in preterm subjects in terms of their fiber dispersion and intra-axonal volume fraction. We recruited 80 nineteen year-old subjects born at fewer than 26 weeks completed gestation, and 51 term controls. Three preterm subjects were excluded due to ventriculomegaly, because it left the white matter tracts unable to be registered to the template. We acquired diffusion MRI in three shells, with 6 at 300, 16 at 750, 32 at 2000, and 4 at b=0 s mm-2. We eddy-corrected the data via affine registration to the b=0 images, and corrected the data for susceptibility artifacts with acquired field maps. We used Tract-Based Spatial Statistics (TBSS)[5] on maps from the diffusion tensor model and NODDI, to identify white matter regions where there was a significant difference between subjects and controls. This technique registers FA maps to MNI space and projects the local maxima to the skeletonized tract. We used threshold-free cluster enhancement to find clusters of significant difference, implemented in FSL [6]. Finally, we correlated FA and NODDI parameter averages.In white matter tracts, the differences in FA between subjects and controls occurred predominantly in previously-noted[7] regions of difference – including the corpus callosum. We made maps of the NODDI parameters of intra-cellular volume fraction (Vi) and orientation dispersion index (ODI) (Figure 1). Regions of significant different in FA closely matched regions of altered ODI, but not V­i. We calculated the mean parameter values in each WM region and correlated them – each data point is one region of one subject (Figure 2). Mean FA was correlated with both ODI and Vi (p<2e-16), the R2 value was 0.71 for ODI and FA, and 0.25 between Vi and FA.

Previous work has attributed changes in the FA in the corpus callosum to changes in the packing of axons, including their density, rather than fiber structure or axonal injury [3]. However, our TBSS analysis suggested that changes in orientation dispersion are significant between the groups. Despite a lack of significant findings in the Vi between the groups, the FA and Vi are significantly correlated, when looking at all means across all subjects and every white matter region. Thus, we can’t rule out differences in Vi between the groups, although changes in ODI are more obvious.

It would be interesting to observe developmental, longitudinal changes of the white matter microstructure during childhood.

Microstructure imaging has the potential to observe, in-vivo, indicators of healthy or unhealthy development, with the view to being able to intervene earlier for those at risk of neurocognitive deficits. This study suggests that NODDI attributes changes in DTI parameters as being due to a mixture of changing axonal density and orientation dispersion. However, the differences between preterm and term-born adults are mainly in orientation dispersion within the white matter.