Closed renelabounek closed 7 months ago
Thank you @renelabounek!
I successfully downloaded the re-defaced images for 49 subjects and saved them to ~/duke/temp/janvalosek/data-multi-subject-redefaced
.
@mguaypaq, could you please push the data to git-annex? Thanks!
In the most of the images in bids format, I have executed fslreorient2std. But not in all the cases. Thus, it may be worth to double check if orientation of all updated scans is the same as manually drawn masks, etc. in the database.
In the most of the images in bids format, I have executed fslreorient2std.
Thanks for this comment! This is important information!
I checked the orientation of both the original images (located under master branch) and the re-defaced ones (~/duke/temp/janvalosek/data-multi-subject-redefaced
):
# orig
data-multi-subject-redefaced$ for file in sub-*/anat/*nii.gz;do echo ${file}; sct_image -i ../data-multi-subject/${file} -header | grep -E qform_[xyz];echo "";done >> orientations_orig.txt
# re-defaced
data-multi-subject-redefaced$ for file in sub-*/anat/*nii.gz;do echo ${file}; sct_image -i ${file} -header | grep -E qform_[xyz];echo "";done >> orientations_redefaced.txt
orientations_orig.txt orientations_redefaced.txt
Indeed, in 18 subjects, orientation has been changed from LPI to RPI; see here.
Thus, it may be worth to double check if orientation of all updated scans is the same as manually drawn masks, etc. in the database.
Good point! The thing is that the label files under derivatives/labels
already have a different orientation (and sometimes even resolution!!!) relative to the original images. This happened because the images were reoriented to RPI and resampled to 1mm iso, see code here. See also issue https://github.com/spine-generic/data-multi-subject/issues/144.
A possible solution: reorient the re-defaced images back to LPI using sct_image -setorient LPI
to match the original images under master branch.
A possible solution: reorient the re-defaced images back to LPI using
sct_image -setorient LPI
to match the original images under master branch.
We agreed on this solution during the SCT dev meeting on 2023-08-10 --> I will do it.
Now, the orientation between the re-defaced and original images (located under master branch) is identical:
orientations_orig.txt orientations_redefaced.txt
See diff here.
I saved the data to ~/duke/temp/janvalosek/data-multi-subject-redefaced
. @mguaypaq, could you please push the data to git-annex? Thanks!
Following the #115, I have shared with @jcohenadad 49 re-defaced T1w scans. I have done visual inspection and confirm that no subject ID switch happenned and that re-defaced scans contains the whole brain volume without any cut-offs.
Six scans can not be re-defaced, because raw data are unavailable:
We should decide if bain cut offs are minimal in these 6 scans or if these IDs should be added into the brain_t1 category in the exclude.yml file.
I am still awaiting if another 4 scans of one center will be re-defaced or not.