search

spine-generic / data-multi-subject

Multi-subject data for the Spine Generic project
Creative Commons Attribution 4.0 International
21 stars 15 forks source link

Two T1w labels for some subjects #147

Closed valosekj closed 3 weeks ago

valosekj commented 11 months ago

@NathanMolinier and I noticed that some subjects have duplicated T1w labels:

data-multi-subject/derivatives/labels/sub-amu02/anat$ ls -1 *T1w*labels*nii.gz
sub-amu02_T1w_RPI_r_labels-manual.nii.gz
sub-amu02_T1w_labels-disc-manual.nii.gz

T1w_RPI_r_labels-manual.nii.gz

These labels were created in the context of the spine-generic paper on the reoriented and resampled images (see lines here). These files contain only two labels at C3 and C5 mid-vertebral levels (because we used only two labels for the registration to the template), see lines here. These labels are available only for 31 subjects (those where the automatic labeling did not work).

T1w_labels-disc-manual.nii.gz

These labels were created manually by @NathanMolinier using the manual-correction.py script in the original images (i.e., no reorientation and resampling were applied). These files contain disc labels for all visible discs. These labels are available for all the subjects.

@NathanMolinier pointed out that it might be confusing to have two "labels" files.

Potential solutions:

Related to https://github.com/spine-generic/data-multi-subject/issues/95

mguaypaq commented 11 months ago

It sounds like the two kinds of label files should be in two separate derivatives, each with their own README (which can just be your two paragraphs from this issue description!). Maybe derivatives/vertebral-labels/ and derivatives/disc-labels/?

valosekj commented 11 months ago

It sounds like the two kinds of label files should be in two separate derivatives, each with their own README (which can just be your two paragraphs from this issue description!).

Yeah, this would make sense.

Maybe derivatives/vertebral-labels/ and derivatives/disc-labels/?

I initially thought that we could keep the T1w_RPI_r_labels-manual.nii.gz files under the current derivatives/labels/ folder to keep them together with other files created using the same analysis script process_data.sh and move the T1w_labels-disc-manual.nii.gz files created by @NathanMolinier to a new derivatives/disc-labels/ folder.

But the tricky thing is that Nathan just updated/corrected the T1w_labels-disc-manual.nii.gz files which already had been living beside T1w_RPI_r_labels-manual.nii.gz under derivatives/labels/ for a long time.

Sidenote: I will also upload soon spinal nerve rootlets segmentations created in the context of the model-spinal-rootlets project. I guess a good strategy would be to save them under a new folder like derivatives/spinal-rootlets-labels/.