Auto-afids uses random forest models to automatically locate 32 anatomical fiducials, which were originally described in the Afids project.
Auto-afids consists of two workflows: auto-afids-train, which trains a set of random forest models from an input BIDS dataset (which is first registered to MNI space), and auto-afids, which automatically locates the 32 fiducials for each T1 weighted image in the input dataset.
Clone the git repository. auto-afids dependencies are managed with Poetry, which will need to be installed. You can find the instructions on the poetry website. Once installed, the development environment can setup with the following commands:
poetry install
poetry run poe setup
poethepoet is used as a task runner. You can see what commands are available by running
poetry run poe
Additionally, pre-commit hooks (installed via the poe setup command) is used to lint and format code (we use black, isort, pylint, and flake8).
auto-afids-train is run with Snakebids through auto-afids-train/run.py. It is formatted as a BIDS app, so it can be run with:
python3 auto-afids-train/run.py <input BIDS dataset> <output directory> <participant or group> <snakemake arguments>
The output models will be available in <output dir>/derivatives/models.
auto-afids is also run with Snakebids through auto-afids/run.py. The CLI is:
python3 auto-afids/run.py <input BIDS dataset> <output directory> <participant or group> --model_dir <root model directory> <snakemake arguments>
The output should be one text file for each of the 32 AFIDs. This workflow is a WIP, and properly formatted FCSV or JSON files should be produced for each subject in the future.