Motion corrected MRI differentiates male and female human brain growth trajectories from mid-gestation

It is of considerable scientific, medical, and societal interest to understand the developmental origins of differences between male and female brains. Here we report the use of advances in MR imaging and analysis to accurately measure global, lobe and millimetre scale growth trajectory patterns over 18 gestational weeks in normal pregnancies with repeated measures. Statistical modelling of absolute growth trajectories revealed underlying differences in many measures, potentially reflecting overall body size differences. However, models of relative growth accounting for global measures revealed a complex temporal form, with strikingly similar cortical development in males and females at lobe scales. In contrast, local cortical growth patterns and larger scale white matter volume and surface measures differed significantly between male and female. Many proportional differences were maintained during neurogenesis and over 18 weeks of growth. These indicate sex related sculpting of neuroanatomy begins early in development, before cortical folding, potentially influencing postnatal development.

For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above.

Software and code
Policy information about availability of computer code Data collection

Data analysis
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Colin Studholme
May 8, 2020 Data collection was carried out using a 1.5T Phillips Achieva/ dStream MRI system, with software release level of 5.1+ during the study period The data analyses in the paper made use of both generic open source code using R version 3.6.0 (2019-04-26) (https://www.rproject.org/) with the lmerTest package (https://cran.r-project.org/web/packages/lmerTest/index.html), C and C++ code written using the gnu g++ (Version 4.8.5) environment (https://gcc.gnu.org) and scripts using csh and bash (https://www.gnu.org/software/bash/ manual) shells and make (https://www.gnu.org/software/make/), installed as part of standard linux environments. Custom code components written in gnu C and C++ in our Biomedical Image Computing Group (BICG) source code control library collectively implement algorithms described in previous work we referenced in the Methods section. These code components may be partially reliant on separate binary libraries or are specific to MRI file formats acquired on our scanner system (such as slice timing and ordering information not fully supported in DICOM image formats) and may therefore require significant changes to use with other image data. Custom code segments will be made available on reasonable request to the first author, that are permitted within the rules and protocols governing intellectual property at the University of Washington and the University of Washington Medical Center. We are also happy to share expertise and collaborate with researchers developing their own implementations of our processing pipelines using other computing environments and MRI data formats.  Tables 1, 2 and 3 and Figures 2 and 3 are stored in CSV format for analysis in the R package. This data can be made available upon reasonable request to the first author given the constraints imposed by institutional regulations at the University of Washington. Primary imaging data files, originally collected at the University of Washington Medical Center may contain components governed by regulations on anonymity, or the intellectual property rules of the University of Washington, and will be shared in part on reasonable request to the first author, within the limits imposed by the institutional regulations and protocols. We are also happy to share expertise and collaborate with researchers aiming to collect their own data on different imaging systems.
The sample size was not known prior to the study. The data analyzed was collected as part of normal cohorts in two separate NIH studies into methods for fetal brain imaging. This data was pooled for the purposes of the analysis in this paper. Statistical modeling accounting for multiple covariates in the population is used to ascertain the significance of the resulting contrasts that are reported.
The approach to data exclusion is described in the methods section of the paper. Exclusion begins with cohort selection using per-established criteria, where we did not include cases of known or suspected clinical abnormality. The next stage of data exclusion is based on scan quality using measures of noise and image artifact as the repeated scans are being combined to form a 3D image. Finally, data is excluded after automated image segmentation based on metrics for image consistency in labeled tissues and outlier analysis of the measures which are then confirmed as tissue segmentation errors by visual inspection using per-established procedures.
We have used all the data we acquired for the modeling. We do not have the resources to carry out a replication study.
No randomization of the cohort data was used. All collected data was used for the modeling and statistical analysis using mixed effects linear modeling using a maximum likelihood framework to account for multiple covariates.
Measurement methods were automated and therefore users did not need to be blinded.

nature research | reporting summary
October 2018

Human research participants
Policy information about studies involving human research participants Population characteristics

Recruitment
Ethics oversight Note that full information on the approval of the study protocol must also be provided in the manuscript.  (See Eklund et al. 2016) The subjects were volunteers from the Puget Sound area with some wider recruits from more remote Washington state locations. The study population therefore represents the socioeconomic mix of that area. We list the key characteristics of the pregnant cohort that was studied in Table 4. The results may potentially not be applicable to other populations.

Magnetic resonance imaging
Recruitment of pregnant volunteers was carried out as described in the methods of the paper from the Puget sound area clinics. All subjects were volunteers who therefore chose to enter the study out of there own interest in the development of their child. This, as with many studies of normal brain anatomy and function, may potentially may bias the cohort.
All protocols for imaging and data handling were approved by the University of Washington Institutional review board Committee on Human Subjects Research,. All pregnant subjects provided written consent before entering the study. Details are given in the methods section of the paper.

T2W multi slice structural imaging
Single shot fast spin echo half Fourier acquisitions (SSFSE) planned in three approximately orthogonal planes These were not collected yet as not all fetal subjects have reached an age where testing can be applied.

1.5T
Multi slice HASTE imaging parameters were as described in the methods section of the paper.

fetal head
Image preprocessing for motion correction is described in the methods section using published and referenced methods.
All data were spatially normalized to a common average anatomical space using a non-linear unbiased groupwize alignment described in the methods section. (an implementation of the symmetric demons approach) Unbiased Template Free Normalization was used for Tensor Based Morphometry All 3D images were reconstructed using iterative methods described in the methods section.
Large field of view 2D multi slice imaging was used.
As described in the methods section, Mixed effects models were used as implemented in the lmerTEST package in R. REML was used for fitting.
All models tested the difference in growth trajectories fitted to measures from male and female brains.
Tissue classes globally and parcellated by the main lobes of the fetal brain as listed in the online methods section of the paper.
As described in the methods section, Non Parametric Permutation testing was used to correct T statistics maps of morphometric differences at a voxel level. No cluster assumptions were used.