Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020;25:102195.
doi: 10.1016/j.nicl.2020.102195. Epub 2020 Jan 23.

Neonatal Morphometric Similarity Mapping for Predicting Brain Age and Characterizing Neuroanatomic Variation Associated With Preterm Birth

Affiliations
Free PMC article

Neonatal Morphometric Similarity Mapping for Predicting Brain Age and Characterizing Neuroanatomic Variation Associated With Preterm Birth

Paola Galdi et al. Neuroimage Clin. .
Free PMC article

Abstract

Multi-contrast MRI captures information about brain macro- and micro-structure which can be combined in an integrated model to obtain a detailed "fingerprint" of the anatomical properties of an individual's brain. Inter-regional similarities between features derived from structural and diffusion MRI, including regional volumes, diffusion tensor metrics, neurite orientation dispersion and density imaging measures, can be modelled as morphometric similarity networks (MSNs). Here, individual MSNs were derived from 105 neonates (59 preterm and 46 term) who were scanned between 38 and 45 weeks postmenstrual age (PMA). Inter-regional similarities were used as predictors in a regression model of age at the time of scanning and in a classification model to discriminate between preterm and term infant brains. When tested on unseen data, the regression model predicted PMA at scan with a mean absolute error of 0.70 ± 0.56 weeks, and the classification model achieved 92% accuracy. We conclude that MSNs predict chronological brain age accurately; and they provide a data-driven approach to identify networks that characterise typical maturation and those that contribute most to neuroanatomic variation associated with preterm birth.

Keywords: Brain age; Developing brain; MRI; Morphometric similarity networks; Multi-modal data; Preterm.

Conflict of interest statement

Declaration of Competing Interest Authors declare no conflict of interests.

Figures

None
Fig. 1
Fig. 1
Distribution of postmenstrual age at scan for all subjects. a) Age distribution for the for term (blue) and preterm (orange) groups. b) Age distribution for male (blue) and female (pink) participants. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 2
Fig. 2
Quality control results. a) Results for the overall population with two selected subjects, one from the top quartile of the SNR and CNR distributions (green star) and the other from the bottom quartile (red star). b) The SNR and CNR maps for the selected subjects. c) The b0 of both subjects before and after the processing pipeline. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 3
Fig. 3
a) Individual MSN construction. Different metrics are extracted from dMRI and sMRI data. The same parcellation is applied to all image types and the average metric values are computed for each ROI. A MSN (represented here as a connectivity matrix) is built by computing the Pearson correlation between the vectors of metrics of each pair of ROIs. b) Training of a predictive model (here for PMA at scan) from individual MSNs. The inter-regional correlations are used as predictor variables in a machine learning model. The performance of the model is evaluated on an independent test set.
Fig. 4
Fig. 4
Histograms of the performance of the 73 models compared in the backward feature selection scheme for the age prediction task (a) and for the classification task (b). Bars are grouped by the number of modalities included in the models.
Fig. 5
Fig. 5
a) Average MSN computed across all subjects using the combination of features selected through the backward feature selection scheme for the age prediction task (Volume, FA, MD, AD, MK, υiso, ODIP). b) Correlation between each connection weight (inter-regional similarity) shown in (a) and PMA at scan across subjects. c) Average MSN computed across all subjects using the combination of features selected through the backward feature selection scheme for the classification task (Volume, T1/T2, FA, MD, AD, RD, MK, υic, υiso, ODIP, ODITOT). d) Correlation between each connection weight (inter-regional similarity) shown in (c) and PMA at birth across subjects. Connections that were identified as predictive features by the models are highlighted in black. ROIs are ordered as in Supplementary Table S1.
Fig. 6
Fig. 6
Null distributions computed over 1000 random permutations of the target variable for the age prediction (a) and the classification tasks (b). The red dotted lines indicate the performances of our models. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 7
Fig. 7
Chord diagram showing MSN edges used for age prediction in at least 99% of regression models in the cross-validation folds. Connections shown in gray are inter-regional similarities that increase with chronological age, while connections in red are inter-regional similarities that decrease with chronological age. The edge width is proportional to the correlation between inter-regional similarities and PMA. The left side of the diagram corresponds to the left side of the brain. Abbreviations for ROI names are explained in Supplementary Table S1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 8
Fig. 8
MSN edges showing a divergent pattern of morphological properties in term and preterm infants in at least 99% of classification models in the cross-validation folds. Gray connections indicate inter-regional similarities that are greater in the preterm group, while red connections are greater in the term group. The edge width is proportional to the correlation between inter-regional similarities and prematurity. The left side of the diagram corresponds to the left side of the brain. Abbreviations for ROI names are explained in Supplementary Table S1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Similar articles

See all similar articles

Cited by 1 article

References

    1. Alexander B., Kelly C.E., Adamson C., Beare R., Zannino D., Chen J., Murray A.L., Loh W.Y., Matthews L.G., Warfield S.K., Anderson P.J., Doyle L.W., Seal M.L., Spittle A.J., Cheong J.L., Thompson D.K. Changes in neonatal regional brain volume associated with preterm birth and perinatal factors. NeuroImage. 2018 - PubMed
    1. Alexander-Bloch A., Giedd J.N., Bullmore E. Imaging structural co-variance between human brain regions. Nat. Rev. Neurosci. 2013;14:322–336. - PMC - PubMed
    1. Anblagan D., Pataky R., Evans M.J., Telford E.J., Serag A., Sparrow S., Piyasena C., Semple S.I., Wilkinson A.G., Bastin M.E., Boardman J.P. Association between preterm brain injury and exposure to chorioamnionitis during fetal life. Sci. Rep. 2016;6:37932. - PMC - PubMed
    1. Anderson P.J. Neuropsychological outcomes of children born very preterm. Seminars Fetal Neonatal Med. 2014;19:90–96. - PubMed
    1. Andersson J.L., Graham M.S., Drobnjak I., Zhang H., Filippini N., Bastiani M. Towards a comprehensive framework for movement and distortion correction of diffusion MR images: within volume movement. NeuroImage. 2017;152:450–466. - PMC - PubMed

Publication types

LinkOut - more resources

Feedback