fetal_brain_mri_atlas

Multi-channel spatio-temporal MRI atlas of normal fetal brain development (dHCP project)

---

This repository contains multi-channel spatio-temporal MRI atlas of normal fetal brain development cretated as a part of the Developing Human Connectome Project (dHCP project).

• 16 timepoints consistent in space covering 21 to 36 weeks GA range:

• structural (T2w, T1w) and diffusion (FA, MD, RD, average DWI, ODF) MRI channels and tissue parcellation:

---

Content

• structural: structural (T2w, T1w) images for each of the timepoints

• diffusion: diffusion (FA, MD, RD, average DWI b=1000 shell, ODF) images for each of the timepoints

• parcellations: tissue label parcellation files for each of the timepoints

• transformations: affine transformations to the common space for each of the timepoints

• info: parcellation label description

---

Fetal MRI datasets

• the fetal MRI datasets used for the atlas were acquired at St. Thomas’s Hospital, London as a part of the dHCP project including structural (T2w, T1w) and diffusion MRI (Price et al., 2019); the datasets were motion-corrected and reconstructed using SVR methods (Cordero-Grande et al., 2019; Kuklisova-Murgasova et al., 2012; Christiaens et al., 2021) to 3D isotropic images

• the selected cohort includes 187 fetuses without reported anomalies, from 21 to 37 weeks GA range and with good MRI image quality

---

Methods

• the spatio-temportal atlas was constructed using MIRTK toolbox similarlty to the dHCP neonatal atlas based on (Schuh et al., 2018) with multi-channel (T2w+cortex) guided registration

• the atlas contains 16 timepoint from [21; 36] weeks GA range with 5 channels (T2w, T1w, FA, MD, RD, average DWI b=1000 shell, ODF) and 0.5mm output isotropic resolution

• the atlas timepoints are consistent in space (1 week termoral window in MIRTK atlas construction method)

• the atlas timepoints were inspected and the correct anatomy was confirmed by clinicians with more than 10 years experience in fetal MRI

• the multi-label brain tissue parcellation protocol was formalised based on both research relevance and visibility of individual structures in T2w images; it has 19 ROIs with separate R/L structures defined based on the fetal brain histology atlases (Bayer et al., 2003; Bayer et al., 2005); the parcellation map was created using a combination of semi-manual refinement of the optimised dHCP DRAW-Em neonatal pipeline segmentations (Makropoulos et al., 2018) and registration between the atlas time-points

---

Comments

• to transform the atlas timepoints to the world space (all timepoints are currently affinely scaled to the same space) - please use the following MIRTK command using affine transforms from the "transformations folder":

ga=24
mirtk transform-image t2-t${ga}.00.nii.gz transformed-t2-t${ga}.00.nii.gz -dofin affine-${ga}.dof -interp BSpline -target t2-t36.00.nii.gz

• to visualise atlas channels including ODF - please use the following MRtrix3 command:

ga=24
mrview t2-t${ga}.00.nii.gz t1-t${ga}.00.nii.gz fa-t${ga}.00.nii.gz -odf.load_sh odf-t${ga}.00.nii.gz

---

License

This brain atlas is distributed under the terms of the Creative Commons CC0 1.0 Universal license.

---

Acknowledgements

The work resulting in this brain atlas was done in the Centre for the Developing Brain at King's College London as part of the dHCP, which received funding from the European Research Council under the European Union's Seventh Framework Programme (FP/2007-2013) [grant no. 319456]. We are grateful to the families who generously supported this trial. The work was supported by the NIHR Biomedical Research Centers at Guy's and St Thomas' NHS Trust.

---

References

In case you found this resource useful please give appropriate credit to the atlases:

• dHCP fetal brain MRI atlas: https://gin.g-node.org/kcl_cdb/fetal_brain_mri_atlas

• dCHP project: http://www.developingconnectome.org

• the bioRxiv preprint is coming soon: "Multi-channel spatio-temporal MRI atlas of the normal fetal brain development from the developing Human Connectome Project"