Skip to content

Requirements & setup

This page lists what you need before running SpinePrep. For the step-by-step walkthrough, see the Quickstart; for a fuller tour, see Install & use.

What SpinePrep needs

The full pipeline calls several neuroimaging tools — the Spinal Cord Toolbox (SCT), FSL, and ANTs. The supported way to get all of them in one place is the container.

You need Docker or, on HPC, Apptainer / Singularity. The container image bundles SCT, FSL, ANTs, PAM50 and SpinePrep itself, so nothing else has to be installed on the host.

SpinePrep ships as a build recipe (Dockerfile.spineprep), not a prebuilt image — see the Quickstart for the build command and why.

If you would rather not use a container, you need on your PATH:

  • SCT (Spinal Cord Toolbox)
  • FSL (for topup and PNM)
  • ANTs
  • Python 3.10–3.12

Then install the Python orchestration layer:

pip install spineprep
# or, from a clone:
git clone https://github.com/SpinePrep/SpinePrep.git
cd SpinePrep && pip install .

The Python package alone does not include SCT/FSL/ANTs; installing those is your responsibility in this mode.

Hardware

  • SpinePrep runs on a normal workstation; no GPU is required.
  • A GPU can speed up SCT's deep-learning segmentation (S2) substantially, but the pipeline runs correctly on CPU.
  • Disk: allow room for BIDS-Derivatives outputs and the QC reports alongside your input data.

Your data

SpinePrep expects a BIDS dataset. At minimum it needs the functional runs; it also uses an anatomical image for the cord reference, and will make use of fieldmaps and physiological recordings if they are present. S1 (Input Verify) checks the layout and reports what it found before any processing runs.

Next