Images are missing in dcm-oklahoma

[NathanMolinier]

Description

I noticed that some spinal cord segmentation did not have their corresponding raw image:

ValueError: missing files:
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM050/anat/sub-CSM050_flip-1_mt-off_MTS.nii.gz
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM129/anat/sub-CSM129_flip-1_mt-off_MTS.nii.gz
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM130/anat/sub-CSM130_flip-1_mt-off_MTS.nii.gz
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM135/anat/sub-CSM135_flip-1_mt-off_MTS.nii.gz
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM137/anat/sub-CSM137_flip-1_mt-off_MTS.nii.gz
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM138/anat/sub-CSM138_flip-1_mt-off_MTS.nii.gz
/home/GRAMES.POLYMTL.CA/p118739/data/datasets/dcm-oklahoma/sub-CSM142/anat/sub-CSM142_flip-1_mt-off_MTS.nii.gz

Do you have any information about that @valosekj ?

[valosekj]

Sorry for the late response. Unfortunately, I do not have any additional details regarding these missing MT files.

[NathanMolinier]

No problem ! Should we remove them from the dataset ?

[valosekj]

What about keeping them? We might get the raw images in the future.

[mguaypaq]

I guess there are two options, here are some pros and cons:

• Delete the extra segmentations
  • Pro: avoid errors in processing scripts which don't expect this situation.
  • Pro: if we get raw images, we can recover the segmentations from the git history.
  • Con: if we get raw images, we might forget that the segmentations exist.
• Keep the extra segmentations
  • Pro: if we get raw images, we won't accidentally forget that we have these segmentations.
  • Con: the segmentations might not match the eventual raw images? We might forget to check this.
  • Con: the situation is unusual.

[NathanMolinier]

Personally I would prefer the first option, deleting these images would be cleaner to avoid mistakes. I created this issue based on an error that I got from one of my scripts.

Also as mentioned by @mguaypaq, if we can use git to recover them I guess we don't lose any work.

[mguaypaq]

Also, if we want a better chance of remembering and finding the "deleted" masks, we can keep a branch or a tag with a hint in the name.