An MRI Atlas of the Human Fetal Brain: Reference and Segmentation Tools for Fetal Brain MRI Analysis

Abstract

Characterizing in-utero brain development is essential for understanding typical and atypical neurodevelopment. Building on prior spatiotemporal fetal brain MRI atlases, we present the CRL-2025 fetal brain atlas, a spatiotemporal (4D) atlas of the developing fetal brain between 21 and 37 gestational weeks. This atlas is constructed from MRI scans of 159 fetuses with typically developing brains using a diffeomorphic deformable registration framework integrated with kernel regression on age. CRL-2025 uniquely includes detailed tissue segmentations, transient white matter compartments, and parcellation into 126 anatomical regions. It offers significantly enhanced anatomical details over the CRL-2017 atlas and is presented along with a re-release of the CRL diffusion MRI atlas featuring newly created tissue segmentation and labels. We release de-identified, processed subject-level fetal MRI datasets used to generate CRL-2025, providing input-output transparency and reproducibility. We also provide FetalSEG, a deep learning-based multiclass segmentation tool to facilitate automatic fetal brain MRI segmentation. The CRL-2025 atlas and its tools enable scalable fetal brain MRI segmentation, analysis, and neurodevelopmental research for the broader community.

Figures (6)

• Figure 1: Age distribution of samples used in CRL2025 atlas construction. We used Gaussian kernel regression on age with a normalized $\pm$1 week kernel width.
• Figure 2: Overview of fetal MRI processing for spatiotemporal atlas generation. (a) Spatiotemporal anatomical (T2-weighted) atlas generation process based on fast T2-weighted MRI scans. (b) Spatiotemporal diffusion MRI (DTI) atlas generation process, which is a separate, modality-specific pipeline including 1) motion-tracking–based slice-to-volume registration for robust diffusion tensor image reconstruction marami2017temporal, 2) diffusion tensor atlas construction khan2019fetal, and 3) diffusion tensor atlas labeling calixto2025detailed. These two workflows are independent (T2w for structural atlas, DTI for diffusion atlas) rather than alternative preprocessing paths. Data, atlas construction, labeling procedures, automatic segmentation methods, and validations are discussed in this article.
• Figure 3: Comparison of spatiotemporal fetal brain MRI atlases ( CRL-2025 vs. CRL-2017) at six representative gestational ages: 22, 25, 28, 31, 34, and 37 weeks. Axial, coronal, and sagittal views are presented for each atlas at each age.
• Figure 4: Tissue and regional segmentations and structural labels overlaid on axial views of the CRL-2025 spatiotemporal fetal brain MRI atlas at six representative gestational age (GA) weeks. All label schemes have subcortical structures including lentiform and caudate nuclei, internal capsules, thalami, and hippocampi separately on each hemisphere. Tissue segmentation labels (middle row) delineate the cortical plate-white matter boundary, CSF, and subcortical structures. In the top row, the white matter of the tissue segmentation is divided into white matter compartments (WMC) including the ventricular and intermediate zones and the subplate. These transient WMCs gradually disappear and were not clearly observable on the atlases beyond 31 weeks. The bottom row displays regional segmentations which are useful for regional and connectivity analyses.
• Figure 5: The spatiotemporal fetal brain diffusion MRI atlas shown at representative GA weeks 23-35. Top row: Color fractional anisotropy (FA). Middle row: Tissue segmentation. Bottom row: Regional segmentation.