Make clear if reorientation was done before training MP2RAGE model

[valosekj]

It's not clear from hc-leipzig-7t-mp2rage/convert_bids_to_nnunetv2.py if reorientation was applied before training nnUNet models. I assume so, as the release info says:

The orientation of the input image is expected to be RPI (right-posterior-inferior).

It'd be nice to clarify this and make the reorientation reproducible, e.g., include in the convert_bids_to_nnunetv2.py script or mention it in a README file, which could be added to https://github.com/ivadomed/model-spinal-rootlets/tree/main/training/hc-leipzig-7t-mp2rage -- currently the hc-leipzig-7t-mp2rage folder contains a bunch of files with any explanation.

[valosekj]

Also, on a similar note, in the paper we say that nnunet preprocessing involved reorientation to RPI:

But, wasn't the reorientation done manually before by sct_image -setorient? If so, we should clarify it in the paper.

TODOs:

• confirm how and when reorientation was done (sct_image vs nnunet)
• clarify it in this repo (please see my first comment) and in the paper
• When clarifying reorientation, I think it would be nice to clarify resampling, i.e., state in this repo that the median resolution of x.x mm was automatically determined by nnunet and used for training and inference. To make explicit that we didn't resample manually before (as done by Thomas in Do we have to resample when performing inference ? model-canal-seg#33).

[KaterinaKrejci231054]

Reorientation to RPI was done using sct_image and it was added to the convert_bids_to_nnunetv2.py script in #101

model-spinal-rootlets/training/hc-leipzig-7t-mp2rage/convert_bids_to_nnunetv2.py

Line 92 in e50093b

subprocess.run(["sct_image", "-i", dst_file, "-setorient", "RPI"], check=True)