Deprecated

This program is superseded by pl-emerald.

Fetal Brain Mask

Automatic brain image masking tool for fetal brain MRI using deep learning.

For each .nii or .nii.gz file in a directory, a _mask.nii is produced.

Usage

$ docker run --rm fnndsc/pl-fetal-brain-mask:latest fetal_brain_mask --help 2> /dev/null

usage: fetal_brain_mask [-h] [--json] [--savejson DIR] [--inputmeta INPUTMETA]
                        [--saveinputmeta] [--saveoutputmeta] [--version]
                        [--meta] [-v VERBOSITY] [--man] [-p INPUTPATHFILTER]
                        [-i COPY_INPUT]
                        [--overlay-destination OVERLAY_DESTINATION]
                        [--overlay-suffix OVERLAY_SUFFIX]
                        [--overlay-background OVERLAY_BACKGROUND] [-o SUFFIX]
                        [-s] [-l SKIPPED_LIST]
                        inputdir outputdir

Automatic masking of fetal brain images using deep learning

positional arguments:
  inputdir              directory containing the input files
  outputdir             directory containing the output files/folders

optional arguments:
  -h, --help            show this help message and exit
  --json                show json representation of app and exit (default:
                        False)
  --savejson DIR        save json representation file to DIR and exit
                        (default: None)
  --inputmeta INPUTMETA
                        meta data file containing the arguments passed to this
                        app (default: None)
  --saveinputmeta       save arguments to a JSON file (default: False)
  --saveoutputmeta      save output meta data to a JSON file (default: False)
  --version             print app version and exit (default: False)
  --meta                print app meta data and exit (default: False)
  -v VERBOSITY, --verbosity VERBOSITY
                        verbosity level for the app (default: 0)
  --man                 show the app's man page and exit (default: False)
  -p INPUTPATHFILTER, --inputPathFilter INPUTPATHFILTER
                        selection for which files to evaluate (default:
                        **.nii[,.gz])
  -i COPY_INPUT, --input-destination COPY_INPUT
                        copy input files into a subdirectory of the output
                        directory (default: )
  --overlay-destination OVERLAY_DESTINATION
                        create volumes in given directory where the mask is
                        overlayed on the source image, Useful for
                        visualization. (default: )
  --overlay-suffix OVERLAY_SUFFIX
                        file name suffix for overlays, if needed. (default:
                        _mask_overlay.nii)
  --overlay-background OVERLAY_BACKGROUND
                        intensity scale for masked-out portion in the optional
                        overlay. (default: 0.2)
  -o SUFFIX, --suffix SUFFIX
                        output filename suffix (default: _mask.nii)
  -s, --smooth          perform post-processing on images including
                        morphological closing and defragmentation (default:
                        True)
  -l SKIPPED_LIST, --skipped-list SKIPPED_LIST
                        produce an output file containing the names of files
                        which were skipped (default: )

Examples

Single sample, minimum example

mkdir in out
mv scan.nii.gz in

docker run --rm  \
    -v $PWD/in:/incoming:ro -v $PWD/out:/outgoing:rw  \
    fnndsc/pl-fetal-brain-mask:latest fetal_brain_mask \
    /incoming /outgoing

Multiple inputs are processed in parallel, using all the cores visible inside the container. For large datasets, you can limit the number of concurrent jobs to N jobs with docker run --cpuset-cpus 0-$((N-1))

mkdir in out
mv scan0001.nii.gz scan0002.nii.gz scan0003.nii.gz scan0004.nii.gz \
   scan0005.nii.gz scan0006.nii.gz scan0007.nii.gz scan0008.nii.gz in

# limit to 5 concurrent jobs
docker run --rm -u $(id -u):$(id -g)                   \
    -v /etc/localtime:/etc/localtime:ro                \
    --cpuset-cpus 0-4                                  \
    -v $PWD/in:/incoming:ro -v $PWD/out:/outgoing:rw   \
    fnndsc/pl-fetal-brain-mask:latest fetal_brain_mask  \
    --inputPathFilter 'scan????.nii.gz'                \
    --skipped-list skipped.txt                         \
    --verbosity 5                                      \
    /incoming /outgoing

See docker run --help for other throttling options.

Some (not all) unnessessary tensorflow output can be silcence

docker run --rm -e TF_CPP_MIN_LOG_LEVEL=3 fnndsc/pl-fetal-brain-mask:latest

Running on ChRIS

The option --input-destination input facilitates access to data by downstream plugins.

--overlay-destination overlay, followed by pl-pfdo_med2img is useful for visualization of masking results via ChRIS_ui.

Potential Future Features

• --overlay-background accepts multiple values so user can experiment with different values, and also directly output the masked brain with the value of 0.0

Links

Original source: https://github.com/chrisorozco1097/masking_tool