Update T1w scans with re-defaced images

[mguaypaq]

Images supplied by @renelabounek (thanks!)

Fixes #146.

[valosekj]

Thanks for the push @mguaypaq!

I visually checked a few subjects in FSLeyes, and I also checked the orientation. The re-defaced images correspond with the original images 👍🏻

However, I noticed that some of the re-defaced images have changed intensity range:

$ cd data-multi-subject

# master branch
$ fslstats sub-fslAchieva02/anat/sub-fslAchieva02_T1w.nii.gz -R
0.000000 4022.933105

# mgp/redefaced branch
$ git fetch
$ git checkout mgp/redefaced
$ git pull && git annex get .
$ fslstats sub-fslAchieva02/anat/sub-fslAchieva02_T1w.nii.gz -R
0.000000 3219834.250000

The data type is the same for both:

# master branch
$ sct_image -i sub-fslAchieva02/anat/sub-fslAchieva02_T1w.nii.gz -header | grep data_type                                              
data_type	FLOAT64                                                                                                                                                                                         

# mgp/redefaced branch
$ sct_image -i sub-fslAchieva02/anat/sub-fslAchieva02_T1w.nii.gz -header | grep data_type                                      
data_type	FLOAT64

And both images look identical in FSLeyes. So I wonder if this is a problem.

[mguaypaq]

Ultimately, I don't know whether the difference in intensity range is a problem. But, I did look at the old and new versions of sub-fslAchieva02/anat/sub-fslAchieva02_T1w.nii.gz, and this is what I noticed:

• The affines (both the sform and the qform) between the two versions are very slightly different (probably not a problem?)
• The descrip header field is different, but probably not useful anyway (old: TE=3.6;Time=154907.020, looks like metadata; new: 6.0.4:ddd0a010, looks like a software version).
• After rescaling the new image from the range [0, 3219834] to the range [0, 4023], the absolute difference outside of the re-faced region seems to be less than 40, that is, less than 1% of the max intensity.

[valosekj]

Hey @mguaypaq! Very sorry, this PR slipped through the cracks...

I can see on the 2023-09-07 SCT dev meeting agenda that we concluded that rescaling is not necessary, but should be documented in changelog (no dtype change, intensities rescaled for some images).