While using 5ttgen hsvs, we noticed that the brainstem is assigned to volume 4 (pathological tissue) with the hsvs algorithm, cf screenshot 1.
Although the 5ttgen hsvs algorithm includes the option to assign the brainstem to white matter, it is confusing to see the brainstem in the pathological tissue class.
Shouldn’t the brainstem always result in the same tissue class, regardless of the 5tt algorithm used, and especially not in vol. 4 (pathological tissue class)? Cf. screenshot 2 where I have used the 5ttgen freesurfer algorithm which maps the brainstem in vol. 2 and only the lesion in vol. 4.
Could someone please explain to me if this behaviour is intentional or a bug? Probably @rsmith knows more about this?
Hi, it looks like this phenomenon is still occuring in the current version.
Could someone please clarify @jdtournier@rsmith ?
Edit: I assume the bs is being assigned to patholgy so there are no priors?
Thank you very much & best,
Lucius
Check out the option -white_stem for the hsvs algorithm.
Conversely, if the goal is to label the brainstem as the 5th tissue type using the FreeSurfer algorithm, consider the following steps:
Extract the brainstem segmentation from the aparc+aseg image (i.e., mrcalc <aparc+aseg> <brainstem_index> -eq <brainstem_segmentation>).
Add the brainstem to the 5TT image as “pathology” using 5ttedit <5TT_image> -path <brainstem_segmentation> <modified_5TT_image>.
Note: The brainstem_index can be found using a lookup table. In ${FREESURFER_HOME}/FreeSurferColorLUT.txt, it is indexed at 16.
Additionally, it might be necessary to regrid the brainstem segmentation to match the resolution of the 5TT image. In that case, try this command: mrgrid <brainstem_segmentation> -template <5TT_image> regrid <brainstem_segmentation_regridded>.
Thank you very much for your reply. However, I think the issue should be why the brain stem is attributed to different tissue types in different algorithms. In this case, 5ttge -hsvs attributes the brain stem to the ‘pathology’ tissue while the 5ttgen -freesurfer attributes it to the sub-cortical tissue type.
Thanks for the clarification I am unsure if this will be addressed anytime soon…In the meantime, the suggested approach could help achieve consistent brainstem segmentation labelling across the different 5TT algorithms, if required.
