A nipype implementation of an anatomical MR and PET defacing pipeline for BIDS datasets. This is a working prototype, in active development denoted by the 0.x.x version number. However, it is functional and can be used to deface PET and MR data as well as co-register the two modalities. Use is encouraged and feedback via Github issues or email to openneuropet@gmail.com is more than welcome. As is often the case, this medical research software is constrained to testing on data that its developers have access to.
This software can be installed via source or via pip from PyPi with pip install petdeface
CI | Status |
---|---|
docker build . -t petdeface |
|
docker push |
|
Python 3.14 >= 3.10 |
|
packaging |
|
Docs |
- FreeSurfer and MiDeFAce >= 7.3.2
- nipype >= 1.6.0
- pybids
for a full list of dependencies see the pyproject.toml in this repo
NOTE: This project is currently in beta release, some features listed below may not be available for version numbers < 1.0.0
usage: petdeface.py [-h] [--anat_only] [--participant_label PARTICIPANT_LABEL [PARTICIPANT_LABEL ...]] [--docker] [--singularity] [--n_procs N_PROCS]
[--skip_bids_validator] [--version] [--placement PLACEMENT] [--remove_existing]
[--participant_label_exclude participant_label_exclude [participant_label_exclude ...]] [--session_label SESSION [SESSION ...]]
[--session_label_exclude session_label_exclude [session_label_exclude ...]] [--open_browser]
bids_dir [output_dir] [analysis_level]
PetDeface
positional arguments:
bids_dir The directory with the input dataset
output_dir The directory where the output files should be stored, if not supplied will default to <bids_dir>/derivatives/petdeface
analysis_level This BIDS app always operates at the participant level, if this argument is changed it will be ignored and run as a participant level
analysis
options:
-h, --help show this help message and exit
--anat_only, -a Only deface anatomical images
--participant_label PARTICIPANT_LABEL [PARTICIPANT_LABEL ...], -pl PARTICIPANT_LABEL [PARTICIPANT_LABEL ...]
The label(s) of the participant/subject to be processed. When specifying multiple subjects separate them with spaces.
--docker, -d Run in docker container
--singularity, -si Run in singularity container
--n_procs N_PROCS Number of processors to use when running the workflow
--skip_bids_validator
--version, -v show program's version number and exit
--placement PLACEMENT, -p PLACEMENT
Where to place the defaced images. Options are 'adjacent': next to the bids_dir (default) in a folder appended with _defaced'inplace':
defaces the dataset in place, e.g. replaces faced PET and T1w images w/ defaced at bids_dir'derivatives': does all of the defacing within
the derivatives folder in bids_dir.
--remove_existing, -r Remove existing output files in output_dir.
--preview_pics Create preview pictures of defacing, defaults to false for docker (only works if FreeView is installed on machine)
--participant_label_exclude participant_label_exclude [participant_label_exclude ...]
Exclude a subject(s) from the defacing workflow. e.g. --participant_label_exclude sub-01 sub-02
--session_label SESSION [SESSION ...]
Select only a specific session(s) to include in the defacing workflow
--session_label_exclude session_label_exclude [session_label_exclude ...]
Select a specific session(s) to exclude from the defacing workflow
--use_template_anat Use template anatomical image when no T1w is available for PET scans.
Options: 't1' (included T1w template), 'mni' (MNI template), or 'pet'
(averaged PET image).
--open_browser Open browser to show QA reports after completion
Working example usage:
petdeface /inputfolder /outputfolder --n_procs 16 --skip_bids_validator --placement adjacent
When PET scans lack corresponding T1w anatomical images, PETdeface can use template anatomical images for registration and defacing. Three options are available:
--use_template_anat t1
: Uses a T1w template included with the PETdeface library--use_template_anat mni
: Uses the MNI standard brain template--use_template_anat pet
: Creates a template by averaging the PET data across time
Important: When using template anatomical images, it's crucial to validate the defacing quality. Inspect the output using the generated HTML report (with --open_browser
) or a NIfTI viewer to ensure the defacing is valid for your data.
Recommended workflow for subjects missing T1w images:
- First, exclude subjects missing T1w using
--participant_label_exclude
- Run defacing on subjects with T1w images
- Then run defacing separately on subjects missing T1w using
--participant_label
and test different templates (t1
,mni
,pet
) to determine which works best for your data
Example usage with template anatomical:
petdeface /inputfolder /outputfolder --use_template_anat t1 --n_procs 16
Requirements:
- Docker must be installed and access to
docker run
must be available to the current user openneuropet/petdeface
must be present or reachable at dockerhub from the machine the cli is installed at, e.g.docker pull openneuropet/petdeface
must work- if one is unable to pull the image on can build locally with
make dockerbuild
NOTE: The docker image for petdeface is not intended to be used by itself, but instead accessed via the petdeface
command line written in Python.
Appending the --docker
after including all of the required arguments for petdeface will
automatically launch the dockerized version of this application, no additional input after
that is required.
Running directly with Docker, no Python, no installation:
If you run without using the CLI you will need to:
- bind the input and output volumes to the container
- bind a freesurfer license to the container at
/opt/freesurfer/license.txt
- provide all of the arguments you would normally need to provide to the Python CLI
- provide $UID and $GID if running on linux so that your output isn't written as root, you may disregard this if you're handy.
An example of the command generated from the Python cli to run the docker based version can be seen below:
docker run --user=$UID:$GID -a stderr -a stdout --rm \
-v /Data/faced_pet_data/:/input \
-v /Data/defaced_pet_data/:/output \
-v /home/freesurfer/license.txt:/opt/freesurfer/license.txt \
--platform linux/amd64 \
petdeface:latest /input /output --n_procs 16 --skip_bids_validator --placement adjacent --user=$UID:$GID system_platform=Linux
Requirements:
- Singularity must be installed
openneuropet/petdeface
must be present or reachable at dockerhub
One can execute petdeface in singularity either directly via:
singularity exec -e --bind license.txt:/opt/freesurfer/license.txt docker://openneuropet/petdeface:0.1.1 petdeface
Input and Output directories don't need to be bound, but one does need to bind a freesurfer license to the image before they can proceed with defacing.
Otherwise, one can run and execute petdeface with the same syntax as calling it from the command line, the only difference being that petdeface is prepended
with singularity exec -e
singularity exec -e --bind license.txt:/opt/freesurfer/license.txt docker://openneuropet/petdeface:0.1.1 petdeface /input /output --n_procs 10
NOTE: Testing with singularity has been limited to version singularity-ce 4.2.0, please let us know in the issues section of this repo if you have trouble running this container in singularity/apptainer.
This project supports both UV and standard Python (pip + build) workflows for development and packaging.
git clone https://github.com/openneuropet/petdeface.git
cd petdeface
uv build
pip install dist/petdeface-<X.X.X>-py3-none-any.whl # where X.X.X is the version number of the generated file
git clone https://github.com/openneuropet/petdeface.git
cd petdeface
pip install --upgrade pip
pip install .[dev]
# To build a wheel or sdist:
pip install build
python -m build
pip install dist/petdeface-<X.X.X>-py3-none-any.whl # where X.X.X is the version number of the generated file
Then install the tar or wheel file created in dist
:
pip install petdeface-<X.X.X>-py3-none-any.whl # where X.X.X is the version number of the generated file
- Dale A, Fischl B, Sereno MI. Cortical Surface-Based Analysis: I. Segmentation and Surface Reconstruction. Neuroimage. 1999;9(2):179–94. doi:10.1006/nimg.1998.0395.
- Fischl B. FreeSurfer. Neuroimage. 2012 Aug 15;62(2):774-81. doi: 10.1016/j.neuroimage.2012.01.021. Epub 2012 Jan 10. PMID: 22248573; PMCID: PMC3685476.
- Stefano Cerri, Douglas N. Greve, Andrew Hoopes, Henrik Lundell, Hartwig R. Siebner, Mark Mühlau, Koen Van Leemput, An open-source tool for longitudinal whole-brain and white matter lesion segmentation, NeuroImage: Clinical, Volume 38, 2023, 103354, ISSN 2213-1582, https://doi.org/10.1016/j.nicl.2023.103354. (https://www.sciencedirect.com/science/article/pii/S2213158223000438)
- Gorgolewski, Krzysztof J. ; Esteban, Oscar ; Burns, Christopher ; Ziegler, Erik ; Pinsard, Basile ; Madison, Cindee ; Waskom, Michael ; Ellis, David Gage ; Clark, Dav ; Dayan, Michael ; Manhães-Savio, Alexandre ; Notter, Michael Philipp ; Johnson, Hans ; Dewey, Blake E ; Halchenko, Yaroslav O. ; Hamalainen, Carlo ; Keshavan, Anisha ; Clark, Daniel ; Huntenburg, Julia M. ; Hanke, Michael ; Nichols, B. Nolan ; Wassermann , Demian ; Eshaghi, Arman ; Markiewicz, Christopher ; Varoquaux, Gael ; Acland, Benjamin ; Forbes, Jessica ; Rokem, Ariel ; Kong, Xiang-Zhen ; Gramfort, Alexandre ; Kleesiek, Jens ; Schaefer, Alexander ; Sikka, Sharad ; Perez-Guevara, Martin Felipe ; Glatard, Tristan ; Iqbal, Shariq ; Liu, Siqi ; Welch, David ; Sharp, Paul ; Warner, Joshua ; Kastman, Erik ; Lampe, Leonie ; Perkins, L. Nathan ; Craddock, R. Cameron ; Küttner, René ; Bielievtsov, Dmytro ; Geisler, Daniel ; Gerhard, Stephan ; Liem, Franziskus ; Linkersdörfer, Janosch ; Margulies, Daniel S. ; Andberg, Sami Kristian ; Stadler, Jörg ; Steele, Christopher John ; Broderick, William ; Cooper, Gavin ; Floren, Andrew ; Huang, Lijie ; Gonzalez, Ivan ; McNamee, Daniel ; Papadopoulos Orfanos, Dimitri ; Pellman, John ; Triplett, William ; Ghosh, Satrajit (2016). Nipype: a flexible, lightweight and extensible neuroimaging data processing framework in Python. 0.12.0-rc1. Zenodo. 10.5281/zenodo.50186