1

i'm working on brain images (nifti). I segmented the brain with SPM12. So the segmentation went fine but for some (a lot) of the subjects, i get this:
grey matter segmentation.nii

It's a 3D image (T1 MRI) from "ADNI" database.

Like you can see there are still pieces of the skull. I need to remove them. can you tell me the steps to accomplish that? A code/algorithm using matlab or R?

Thank you!

learners
  • 195
  • 1
  • 12
  • Do you have only one image per patient, or multiple "slices" per patient? – Leander Moesinger Nov 06 '17 at 11:16
  • Just as an idea, you can get rid of at least some of the skull by 1) calculating the connected components, 2) draw a convex hull around the largest component 3) delete everything outside the convex hull. In the example image this would work well in the lower image but not remove everything in the upper one. – Leander Moesinger Nov 06 '17 at 11:17
  • 2
    can you upload an image without the arrows and the blue lines? – ibezito Nov 06 '17 at 11:18
  • Hello, thank you for your time @LeanderMoesinger i've multiple slices per patient. Upper and lower images are the same it's just a different plan (saggital, axial) but same slice! – learners Nov 06 '17 at 11:59
  • Is this the output of the gray matter from SPM? I usually get it cleaner. Anyway, SPM gives the gray matter as probability map, I assume that the skull areas have lower values than the real gray matter. – Adiel Nov 06 '17 at 13:30
  • the skull areas don't necessarily have lower values than the real gray matter :( if it was the case I would have imposed a threshold. This problem didn't happen for all of my patient – learners Nov 06 '17 at 14:02
  • I would try adjusting the segmentation parameters in SPM and run again Default options balance efficiency and efficacy. For example, you might try adjusting downward the distance between sampled points. Also, read section 5.3.1-2 of the SPM12 manual, wrt the clean up procedure. Also, might want to adjust the number of gaussians used to model the intensity distribution of each tissue class. – Jacob Lee Jan 30 '18 at 15:58

0 Answers0