philadelphia-pediatric - swapped R-L axis for rec-composed images?

[valosekj]

When fixing qform and sfrom matrices for the newly shared data (details: #334 (comment)), I noticed that the existing rec-composed_T2w images might also have a swapped R-L axis compared to the acq-top_run-1_T2w and acq-bottom_run-1_T2w images.

Let's check, for example, sub-123 by the following commands:

cd philadelphia-pediatric/sub-123/anat
# reorient the composed image (which is AIL) to LPI to match the orientation of acq-top and acq-bottom images to allow easy visual comparison
sct_image -i sub-123_rec-composed_T2w.nii.gz  -setorient LPI -o sub-123_rec-composed_T2w_LPI.nii.gz
# Open three separate FSLeyes
ff sub-123_rec-composed_T2w_LPI.nii.gz  & ff sub-123_acq-top_run-1_T2w.nii.gz & ff sub-123_acq-bottom_run-1_T2w.nii.gz

Now, you locate the subtle scoliosis in all images and compare the R-L axis. If I am not mistaken, I believe that the R-L labels in the composed image should actually be L-R.

GIF comparing rec-composed and acq-top

Notice the different directions of the spinal cord centerline between images while the R-L labels are still the same.

Do we have more details on how the rec-composed_T2w images were created?

Relevant comment to see: #334 (comment)