Diffuse white matter abnormality in very preterm infants at term reflects reduced brain network efficiency

Abstract

Between 50 and 80% of very preterm infants (<32 weeks gestational age) exhibit increased white matter signal intensity on T2-weighted MRI at term-equivalent age, known as diffuse white matter abnormality (DWMA). A few studies have linked DWMA with microstructural abnormalities, but the exact relationship remains poorly understood. We related DWMA extent to graph theory measures of network efficiency at term in a representative cohort of 343 very preterm infants. We performed anatomic and diffusion MRI at term and quantified DWMA volume with our novel, semi-automated algorithm. From diffusion-weighted structural connectomes, we calculated the graph theory metrics local efficiency and clustering coefficient, which measure the ability of groups of nodes to perform specialized processing, and global efficiency, which assesses the ability of brain regions to efficiently combine information. We computed partial correlations between these measures and DWMA volume, adjusted for confounders. Increasing DWMA volume was associated with decreased global efficiency of the entire very preterm brain and decreased local efficiency and clustering coefficient in a variety of regions supporting cognitive, linguistic, and motor function. We show that DWMA is associated with widespread decreased brain network efficiency, suggesting that it is pathologic and likely has adverse developmental consequences.

Keywords: Diffuse white matter abnormality; Graph theory; Very preterm; diffusion MRI.

Copyright © 2021 The Authors. Published by Elsevier Inc. All rights reserved.

Figures

Fig. 1

Example Semi-automated DWMA Segmentation. The top row shows three slices from a T2-weighted MRI scan of the same male infant; gestational age of 25.4 weeks, postmenstrual age at MRI of 41.7 weeks. The bottom row displays the same slices with clusters of voxels designated as DWMA by our algorithm circled in yellow. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Heatmap of DWMA concentration in the very preterm brain. 311 DWMA segmentations created by our algorithm were aligned to the AAL infantneo template using linear and nonlinear warping, to produce a heatmap of its location in our cohort. Two different views of the same map (top and bottom row) are shown from three perspectives. Yellow means that DWMA existed in a particular voxel for more subjects. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

DWMA vs regional LE and CC for FA-weighted structural networks. Nodes in which DWMA volume was negatively correlated with local efficiency (top) or clustering coefficient (bottom) in the very preterm brain, after covariate correction and FDR. There were no significant positive correlations. The networks are shown from sagittal (left), axial (middle), and coronal (right) perspectives. Colors represent network membership: blue = cognitive, yellow = language, red = motor, green = cognitive/language, purple = cognitive/motor, orange = language/motor, and pink = other. The size of each sphere represents the total variance explained in the covariate-corrected model of normalized DWMA volume.

Fig. 4

DWMA vs regional LE and CC for inverse MD-weighted structural networks. Nodes in which DWMA volume was negatively correlated with local efficiency (top) or clustering coefficient (bottom) in the very preterm brain, after covariate correction and FDR. There were no significant positive correlations. The networks are shown from sagittal (left), axial (middle), and coronal (right) perspectives. Colors represent network membership: blue = cognitive, yellow = language, red = motor, green = cognitive/language, purple = cognitive/motor, orange = language/motor, and pink = other. The size of each sphere represents the total variance explained in the covariate-corrected model of normalized DWMA volume. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

DWMA vs regional LE and CC for inverse RD-weighted structural networks. Nodes in which DWMA volume was negatively correlated with local efficiency (top) or clustering coefficient (bottom) in the very preterm brain, after covariate correction and FDR. There were no significant positive correlations. The networks are shown from sagittal (left), axial (middle), and coronal (right) perspectives. Colors represent network membership: blue = cognitive, yellow = language, red = motor, green = cognitive/language, purple = cognitive/motor, orange = language/motor, and pink = other. The size of each sphere represents the total variance explained in the covariate-corrected model of normalized DWMA volume. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)