Tips and FAQs
Leveraging precomputed results
Whether manual intervention is required, or you want to reduce processing time, NiBabies allow the use of certain pre-computed files during processing, which can skip part of the workflow. Support is limited to the following files:
Anatomical mask in T1w or T2w space
Antomical segmentation (aseg) in T1w or T2w space
To use pre-computed results, one or more BIDS Derivatives directories must be passed in to NiBabies using the --derivatives
flag.
Derivative directories must include a dataset_description.json
and the required fields.
Additionally, files must include the space-T1w
or space-T2w
key-value pair in the filenames, and a matching sidecar JSON file with the SpatialReference
field defined.
A sample layout of a derivatives directory can be found below:
my_precomputed/
├── dataset_description.json
└── sub-01
└── anat
├── sub-01_space-T2w_desc-aseg_dseg.json
├── sub-01_space-T2w_desc-aseg_dseg.nii.gz
├── sub-01_space-T2w_desc-brain_mask.json
└── sub-01_space-T2w_desc-brain_mask.nii.gz
and the contents of the JSON files:
{"SpatialReference": "sub-01/anat/sub-01_T2w.nii.gz"}
The SpatialReference
file will be used to ensure the raw data and the derivatives are aligned and in the same space.
Multi-atlas segmentation with joint label fusion
By default, NiBabies will run FSL FAST for tissue segmentation, and Infant FreeSurfer for segmentation labels.
However, you can instead use ANTs Joint Label Fusion to generate both, granted you provide multiple atlases with anatomicals / segmentations via the --segmentation-atlases-dir
flag.
When using this approach, there are a few assumptions being made:
The anatomicals are brain masked.
The labeled segmentations follow the FreeSurfer lookup table.
Here is an example layout of what the --segmentation-atlases-dir
flag expects:
$ tree JLF-templates
JLF-templates/
├── Template01
│ ├── Segmentation.nii.gz
│ ├── T1w_brain.nii.gz
│ └── T2w_brain.nii.gz
└── Template02
├── Segmentation.nii.gz
├── T1w_brain.nii.gz
└── T2w_brain.nii.gz
More context on releases
Like other NiPreps, NiBabies follows Calendar Versioning (CalVer), in format of YY.MINOR.MICRO
.
In short, here is a quick heuristic on how new releases should be looked at:
If the
YY
orMINOR
has changed, it is a feature release, with substantial changes to the workflow.If the
YY.MINOR
match the version you used, but theMICRO
has changed, it is a bug-fix release. Check the release notes - if the fixes do not pertain to your data, there is no need to upgrade.
For more in-depth information, refer to the NiPreps release documentation.