The concept of a ChRIS Pipeline is a still evolving idea. At its core, a Pipeline is a structured description of a program (or several programs) and all parameters (and meta data) pertinent to using this program (or programs).
While the word pipeline might suggest a collection (i.e. N > 1) of compute elements (programs, or in the case of ChRIS plugins), a pipeline of just one plugin can still be a valid ChRIS pipeline. Why might a single plugin pipeline be useful? Simply because it can act as an alias or describe fully one manner in which to use (and re-use) the plugin in an extremely simple fashion.
This directory contains a collection of JSON specified ChRIS pipelines that can be uploaded to a ChRIS instance to perform some useful functions.
See the source directory.
A pipeline that accepts as input a single CT image and ultimately produces a report predicting the image to either
- show COVID infection
- show pneumonia
- normal
A pipeline that accepts as input several NIfTI volumes of maternal scans containing a fetus and produces as output a single image cropped to and only showing the fetal brain.
A pipeline that accepts DICOM images and produces as output copies of those images with meta header data appropriately anonymized.
The shell script, pipeline_upload.sh, can be used to upload a pipeline JSON file to a ChRIS instance. In its simplest case, to upload mypipeline.json to a ChRIS instance running on localhost, simply do
pipeline_upload.sh --pipelineFile mypipeline.json --usually, however, the ChRIS instance will not be on localhost but accessible from some IP address. In that case:
pipeline_upload.sh -a some.address.com --pipelineFile mypipeline.json --Please see pipeline_upload.sh -v for more inline options.
The script relies on some command line tools that might not be normally installed. These are jq for JSON handling from the shell, and the npm nodejs tool chripon.
-30-