Integration of manual-correction functionalities in SlicerCART

[NathanMolinier]

https://github.com/spinalcordtoolbox/manual-correction is a related software tool that has been developed at NeuroPoly. This purpose of this software is to orchestrate the correction and/or generation of segmentation/labels for BIDS MRI data.

There has been discussions about merging the two projects, see notably:

• https://docs.google.com/document/d/1EuzMMVCfzFHEJ-878vb1IVLGhueFsFpJAqOmKnZO3cQ/edit
• 3DSlicer exploration - interpolation in 3DSlicer spinalcordtoolbox/manual-correction#97
• 3DSlicer exploration - versioning strategy spinalcordtoolbox/manual-correction#96
• 3DSlicer exploration - adding JSON file feature from manual_correction spinalcordtoolbox/manual-correction#95
• 3D Slicer basic exploration spinalcordtoolbox/manual-correction#90

Below are the specific features that are inherent to manual-correction, to see whether it makes sense to try to add them to SlicerCART:

• Pointwise labeling (ie: single voxel with specific integer value)
• Related to the previous point: manual-correction was relying on sct_label_utils which has a very convenient interface to quickly create pointwise labels in the spinal cord.
• Reads BIDS-structured datasets, and for example, parses through the file names for annotating specific contrasts
• Output BIDS JSON sidecar
• Script package_for_correction to zip-package a subset of the dataset to copy them somewhere else (eg: copy from the server to local computer for manual correction, then copy them back to the server).
• Relying on YAML files that include custom keywords that define the type of segmentation/labels to be performed on the images.
  • Related to this, one strategy would to implement a strategy to be able to switch between various annotation scenarios (as described in Make it possible to create various segmentation tasks and switch between them #42), and then we will be able to port the manual-correction specificities to SlicerCART