sct-pipeline / binder-example Goto Github PK

Specs:

• Input: NIFTI
  • DICOM: Pros: "raw", CONS: binary dependency (dcm2niix), difficult to get cross-vendor
    compatibility
  • NIFTI: Pros: standard, most people know how to get it. Cons: not really...
• One subject: pros: faster
• Jupyter commands:
  • We opt for a unix system call for now because (i) more intuitive, (ii) more compatible if people want to replace commands with other UNIX-only (i.e. non-Python) commands such as dcm2niix, or FSL commands.
  • In the future, we could explore Python-only pipeline.
• OUTPUT:
  • SCT: csv (e.g., MTsat within various white matter tracts, averaged within C2-C4 vertebral levels)
  • Jupyter display: Figure, Left panel: WM tracts, Right panel: plot with MTsat values across tracts.
• Requirements:
  • nilearn (Mathieu will check)

use:

git clone -depth=1 <repo>

In its present form, this pipeline uses sct_example_data, which has NIFTI files. Instead, we would like to directly download Dicom files, convert them to NIFTI and do the rest of the processing.

For integration with Binder, batch_processing.sh should be converted into a Jupyter notebook. A few things to keep in mind while doing that:

• dcm2niix will be to be called as system UNIX call
• All SCT functions should be callable via the main(). More info here.