search

neuropoly / data-management

Repo that deals with datalad aspects for internal use
4 stars 0 forks source link

Data arrays of `grenobleOfsep` subjects have top/bottom halves stacked side-by-side in LR axis #144

Open joshuacwnewton opened 3 years ago

joshuacwnewton commented 3 years ago

Notice how the cursor changes in LR axis.

Top ==> right half:

original-data-1

Bottom ==> left half:

original-data-2

From https://github.com/spinalcordtoolbox/spinalcordtoolbox/issues/3556#issuecomment-956349410:

Ah, this is how all of the grenobleOfsep subjects are formatted! Is this some sort of special convention for larger scans, then?

oh gosh! no, this is definitely an error in the image formatting (i don't know who did what to the raw images after their conversion from DICOM to NIfTI). OK, until this is resolved we need to ignore these images.

jcohenadad commented 3 years ago

One way to fix this is by digital surgery:

  1. sct_crop_image --> produce two NIfTI files (one for the top part, one for the bottom part)
  2. sct_image -concat --> concatenate the two NIfTI files

One thing to look out for also is the output qform/sform: we want it to be consistent with the other images in the series (if there are some?); one test is to overlay two images (eg T1w and T2w) and see if they are approximately overlapping