Cortical folding of the preterm brain: a longitudinal analysis of extremely preterm born neonates using spectral matching

Abstract

Introduction: Infants born extremely preterm (<28 weeks of gestation) are at risk of significant neurodevelopmental sequelae. In these infants birth coincides with a period of rapid brain growth and development, when the brain is also vulnerable to a range of insults. Mapping these changes is crucial for identifying potential biomarkers to predict early impairment.

Methods: In this study we use surface-based spectral matching techniques to find an intrasubject longitudinal surface correspondence between the white-grey matter boundary at 30 and 40 weeks equivalent gestational age in nine extremely preterm born infants.

Results: Using the resulting surface correspondence, we identified regions that undergo more cortical folding of the white-grey matter boundary during the preterm period by looking at changes in well-known curvature measures. We performed Hotelling T(2) statistics to evaluate the significance of our findings.

Discussion: The prefrontal and temporal lobes exhibit most development during the preterm period, especially in the left hemisphere. Such correspondences are a promising result as longitudinal measurements of change in cortical folding could provide insightful information about the mechanical properties of the underlying tissue and may be useful in inferring changes during growth and development in this vulnerable period.

Keywords: Cortex; development; registration; shape analysis; spectra.

Figure 1

Infant brain segmentation into cortical grey matter (GM), white matter (WM), cerebrospinal fluid (CSF), deep grey matter (dGM), cerebellum, and brainstem of infant c for two different time points: 31 and 42 weeks EGA, respectively.

Figure 2

The spectral matching algorithm. After preterm‐specific tissue segmentation and labeling the different regions, meshes of the grey–white matter boundary are defined for the 31‐ and 42‐week EGA time points of infant c. After a rigid Coherent Point Drift initialization (blue and red correspond to the two now overlapping meshes), a point‐wise correspondence is defined by Joint‐Spectral Matching (spectral matching color‐coded anterior‐posterior).

Figure 3

The first five spectral modes of infant c for two different time points: 31 and 42 weeks EGA, respectively. Although the meshes are quite different in cartesian space, they have similar representations in the spectral domain.

Figure 4

Example maps of mismatching variation between the whole‐brain matching and regional spectral matching correspondences (units are in mm). Vertices that are not represented in both individual and whole‐brain meshes were attributed a value of −1 (represented with dark blue).

Figure 5

Maps of mean curvature of whole‐brain white–grey matter boundary (left) for infant c scanned at 31 weeks (A) and 42 weeks (B) EGA. Positive values (red/yellow) represent gyri (convex structures) and negative values (blue) represent sulci (concave structures). The Joint‐Spectral Matching correspondence allows us to map mean curvatures from 42‐week to 31‐week space (C) and compute the changes in mean curvature between these two time points in 31 weeks space (D).

Figure 6

Maps of Gaussian curvature of the grey–white matter boundary of the prefrontal cortex (left) for infant c scanned at 31 weeks (A) and 42 weeks (B) EGA. It can be seen that the Gaussian curvature is more sensitive to geometric errors of the meshes and noisier than the mean curvature maps.

Figure 7

Maps of mean curvature of the grey–white matter boundary of the prefrontal cortex (left) for infant c scanned at 31 weeks (A) and 42 weeks (B) EGA. Positive values (red/yellow) represent gyri (convex structures) and negative values (blue) represent sulci (concave structures). The Joint‐Spectral Matching correspondence allows us to map mean curvatures from 42 week to 31 week space (C) and compute the changes in mean curvature between these two time points in 31 weeks space (D). The difference map of the mean curvature over the preterm period indicates the further development of several primary gyri and sulci, like the middle frontal gyrus (indicated by the black arrow), as well as regions where secondary and tertiary sulci will emerge (black square).

Figure 8

Total curvature of the white–grey matter boundary of the prefrontal cortex (left) shown for infant c at 31 weeks (A) and 42 weeks (B) EGA, local surface area change between the two time points (C), and computed bending energy (D). Positive values (red) in the bending energy represent regions of gyrification, and negative values (blue) represent regions of sulcation.

Figure 9

Maps of mean curvature of the grey–white matter boundary of the temporal lobe (left) for infant c scanned at 31 weeks (A) and 42 weeks (B) EGA. Positive values (red/yellow) represent gyri (convex structures) and negative values (blue) represent sulci (concave structures). The Joint‐Spectral Matching correspondence allows us to map mean curvatures from 42‐week to 31‐week space (C) and compute the changes in mean curvature between these two time points in 31 weeks space (D). The black arrow indicates the transverse temporal gyrus, which becomes definite around 31 weeks of gestation and therefore all other secondary and tertiary gyri surrounding it will emerge during the preterm period.

Figure 10

Total curvature of the white–grey matter boundary of the temporal lobe (left) shown for infant c at 31 weeks (A) and 42 weeks (B) EGA, local surface area change between the two time points (C), and computed bending energy (D). Positive values (red) in the bending energy represent regions of gyrification, and negative values (blue) represent regions of sulcation.

Figure 11

Maps of mean curvature of the grey–white matter boundary of the left occipital lobe (upper row) and left parietal lobe (bottom row) for infant c scanned at 31 weeks (A and D) and 42 weeks (B and E) EGA. Positive values (red/yellow) represent gyri (convex structures) and negative values (blue) represent sulci (concave structures). The Joint‐Spectral Matching correspondence allows us to map mean curvatures from 42‐week to 31‐week space and compute the changes in mean curvature between these two time points in 31 weeks space (C and F). It can be noticed that most differences (c and f) appear on already existing gyri and sulci as most of them are already formed by the time of the first scan.

Figure 12

Changes in the intrasubject mean curvature mapped into the earlier scan using Joint‐Spectral Matching for all nine infants. For each infant (A–I) we show the mean curvature changes in the regions we looked at: right hemisphere (first column), left hemisphere (second column), prefrontal cortex right hemisphere (third column), prefrontal cortex left hemisphere (fourth column), temporal lobe right hemisphere (fifth column), and temporal lobe left hemisphere (sixth column). Positive values (red/yellow) represent gyri (convex structures) and negative values (blue) represent sulci (concave structures).

Figure 13

Changes in the intrasubject mean curvature mapped into the earlier scan using Joint‐Spectral Matching for all nine infants. For each infant (A–I) we show the mean curvature changes in the regions we looked at: right parietal lobe (first column), left parietal lobe (second column), right occipital lobe (third column), and left occipital lobe (fourth column). Positive values (red/yellow) represent gyri (convex structures) and negative values (blue) represent sulci (concave structures).

Figure 14

Statistical maps of longitudinal differences during the preterm period in nine infants of the left and right hemisphere. The P‐values are described by the color bar.