General factors of white matter microstructure from DTI and NODDI in the developing brain

Abstract

Preterm birth is closely associated with diffuse white matter dysmaturation inferred from diffusion MRI and neurocognitive impairment in childhood. Diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI) are distinct dMRI modalities, yet metrics derived from these two methods share variance across tracts. This raises the hypothesis that dimensionality reduction approaches may provide efficient whole-brain estimates of white matter microstructure that capture (dys)maturational processes. To investigate the optimal model for accurate classification of generalised white matter dysmaturation in preterm infants we assessed variation in DTI and NODDI metrics across 16 major white matter tracts using principal component analysis and structural equation modelling, in 79 term and 141 preterm infants at term equivalent age. We used logistic regression models to evaluate performances of single-metric and multimodality general factor frameworks for efficient classification of preterm infants based on variation in white matter microstructure. Single-metric general factors from DTI and NODDI capture substantial shared variance (41.8-72.5%) across 16 white matter tracts, and two multimodality factors captured 93.9% of variance shared between DTI and NODDI metrics themselves. General factors associate with preterm birth and a single model that includes all seven DTI and NODDI metrics provides the most accurate prediction of microstructural variations associated with preterm birth. This suggests that despite global covariance of dMRI metrics in neonates, each metric represents information about specific (and additive) aspects of the underlying microstructure that differ in preterm compared to term subjects.HighlightsWe measured variation of 7 DTI and NODDI metrics across 16 major tractsGeneral factors for DTI and NODDI capture substantial shared variance across tractsGeneral factors also capture substantial shared variance between DTI and NODDISingle-metric and multimodality factors associate with gestational age at birthThe best preterm prediction model contains all 7 single-metric g-factors

Feature similarity gradients detect alterations in the neonatal cortex associated with preterm birth

“…In this work, we demonstrate that the VB index can be used to explore graded changes in the microstructural properties of the neonatal cortex. Consistent with previous descriptions of the dysmaturation following preterm birth as a substantially whole-brain phenomenon ( Telford et al, 2017 ; Blesa et al, 2020 ; Vaher et al, 2022 ), across two independent datasets we find that a significant proportion of cortical microstructure is consistently altered in association with preterm birth, in a spatially distributed rather than localised fashion.…”

“…Imaging of the preterm brain at term-equivalent age has characterised dysmaturation following preterm birth as a global phenomenon (i.e. encompassing the whole brain), with generalised dysconnectivity and atypical topology of developing neural networks, increased water diffusivity, altered white matter microstructure and reduced brain volume and cortical surface area ( Telford et al, 2017 ; Batalle et al, 2018 ; Boardman and Counsell, 2019 ; Blesa et al, 2021 ; Vaher et al, 2022 ). Motivated by the search of neural antecedents of cognitive impairment observed in some preterm born individuals, studies focusing on the neonatal cortex have demonstrated that prematurity impacts cortical growth and microstructural development in a dose-dependent fashion ( Kapellou et al, 2006 ; Ball et al, 2013b ) and that preterm birth alters cortical folding ( Shimony et al, 2016 ; Neil and Smyser, 2018 ; Dubois et al, 2019 ), thalamo-cortical connectivity ( Boardman et al, 2006 ; Ball et al, 2013a ), and different regional MRI metrics with varying spatial distribution, including water diffusion measures, markers of myelination and cortical morphology metrics ( Bouyssi-Kobar et al, 2018 ; Ouyang et al, 2019 ; Ball et al, 2020 ; Dimitrova et al, 2021 ).…”

A Role for the Fornix in Temporal Sequence Memory

“…This resulted in six microstructure metrics for three tracts for all 51 participants. These diffusion measures were reduced through PCA into microstructurally informative features (Chamberland et al, 2019; Geeraert et al, 2020; see also Gagnon et al, 2022; Guberman et al, 2022; Vaher et al, 2022). The tract microstructure data were combined in a single table.…”

Immuno-epigenetic signature derived in saliva associates with the encephalopathy of prematurity and perinatal inflammatory disorders

“…As above, details of individual white matter tract segmentation and subsequent extract of tractaveraged dMRI metrics have been outlined previously from infants from this study sample (79).…”

“…dMRI measures of white matter appear to be highly correlated (e.g. high FA in an individual tract such as the arcuate fasciculus is often accompanied by high FA across all other white matter tracts in that individual), a property that persists from early infancy through to older age (79,86,87). As a result of this, it is common to derive general factors (g-factors) of white matter microstructure to characterise global white matter microstructure.…”