CRL-2025-MAS

Multi-atlas segmentation using the CRL2025 Atlas

This repository contains scripts and extra CRKIT tools used for fetal T2W reconstructed image segmentation. By default, this pipeline uses the CRL2025 T2W Atlas¹ as reference images to perform multi-atlas segmentation (MAS).
ANTs² is used to perform non-rigid registrations of template to target images before segmentation.
Segmentation³ is performed using Probabilistic GMM STAPLE available in the Computational Radiology Lab Toolkit, CRKIT.

Dependencies

• CRKIT: https://www.nitrc.org/projects/staple
• ANTs²

Pipeline script usage

• First rigidly register T2-weighted reconstructions to CRL atlas space with your registration tool of choice.

Command: sh MAS-pipeline.sh [Imagelist] [OutputDir] [MaxThreads]

• Image list is a path list of atlas-space T2-weighted reconstructions and their gestational ages (GA, rounded to whole number weeks), for example:

/workdir/CASE001_t2w.nii.gz 34
/workdir/CASE002_t2w.nii.gz 22
/workdir/CASE003_t2w.nii.gz 29
/workdir/CASE004_t2w.nii.gz 36

• Default settings will generate both tissue and regional segmentations
• Runs partial volume correction (PVC) on the tissue segmentation (--noPVC argument to disable)

Output directory organization:

OutputDir/CASE001_t2w
template_rT: Temp files; non-rigid registrations of atlas images to the target image (and warped segmentations)
log: Records the command and input files for each segmentation
seg: Output segmentations
calc: If available, crosses tissue and regional segmentation to attempt a parcellated tissue segmentation

Modifying atlas images

You can swap or add atlas images to the atlas directory specified in config.sh, just make sure the filename of each file ends in _atlas.nii.gz.
The script matches each _atlas file with corresponding segmentations, by default these are named tissue, tissueWMZ and regional.
Specify a custom label scheme like -l YourLabelSuffix
You can change the output naming of the segmentation files with -p YourOutputPrefix

CRL Toolkit (CRKit) Download

Download CRKit, including STAPLE and other image maniuplation binaries utilized in these scripts, from NITRC: https://www.nitrc.org/projects/staple

There's also a Docker container available with CRKit installed: https://github.com/sergeicu/crkit-docker Your mileage may vary; in its current state not all relevant binaries compile properly

License/Data Use Agreement

These files are published under CC BY 4.0: https://creativecommons.org/licenses/by/4.0/

Files in or referenced in this repository were developed for research purposes and are not intended for medical or diagnostic use and have no warranty. The authors and distributors do not make any guarantees regarding the accuracy or usefulness of results generated from these tools or their derivatives, and are not liable for any damages resulting from their use.

When making use of this work, based on the data use agreement you are required to cite the noted publication with its associated DOI link.¹
If you utilize Probabilistic GMM STAPLE, please cite CRKit and Akhondi-Asl et al³.
Please cite ANTs if the ANTs toolkit is used for image registration².
3D rendering created using ITK-SNAP⁴.

Footnotes

1. Bagheri, M., Velasco-Annis, C., Wang, J., Faghihpirayesh, R., Khan, S., Calixto, C., Jaimes, C., Vasung, L., Ouaalam, A., Afacan, O., Warfield, S.K., Rollins, C.K., Gholipour, A., 2025. An MRI Atlas of the Human Fetal Brain: Reference and Segmentation Tools for Fetal Brain MRI Analysis. arXiv preprint arXiv:2508.15034. https://doi.org/10.7910/DVN/QOO75G ↩ ↩²

2. Avants, B.B., Epstein, C.L., Grossman, M. and Gee, J.C., Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal (2008). https://github.com/ANTsX/ANTs ↩ ↩² ↩³

3. Akhondi-Asl, A. and Warfield, S.K., 2013. Simultaneous truth and performance level estimation through fusion of probabilistic segmentations. IEEE transactions on medical imaging, 32(10), pp.1840-1852. ↩ ↩²

4. Yushkevich, P.A., Gao, Y. and Gerig, G., 2016, August. ITK-SNAP: An interactive tool for semi-automatic segmentation of multi-modality biomedical images. In 2016 38th annual international conference of the IEEE engineering in medicine and biology society (EMBC) (pp. 3342-3345). IEEE. https://www.itksnap.org/pmwiki/pmwiki.php ↩