search

FMRIB-Software-Library / support

Public FSL issue tracker and support portal
4 stars 0 forks source link

fslmaths erosion question -- eroding too much? #6

Closed hanayik closed 3 years ago

hanayik commented 3 years ago

Hi experts,

I had a follow-up question to this post regarding fslmaths -ero options: https://www.jiscmail.ac.uk/cgi-bin/wa-jisc.exe?A2=ind2002&L=FSL&P=R6629

I am trying to erode a thalamus ROI in a subject's native space by 1 voxel. However, as the poster here described, it appears to be eroding by more than 1 voxel? It's reducing the # of voxels in my ROI from e.g., 587 voxels -> 157 voxels.

fslmaths $ROI -kernel boxv 3 -ero $subj_ROI_folder/$ROIname'ero'${SUB}.nii.gz

This seems to be incorrect if I only want to take off one voxel?? Or am I thinking about this entirely incorrectly?

hanayik commented 3 years ago

It can be hard to judge if “excess” erosion has occurred, due to the 3D nature of the kernel. If you upload your input image to:

https://oxfile.ox.ac.uk/oxfile/work/extBox?id=700839B35190E48BF

we will be able to double-check the output. If you have a specific “bad” voxel to check as a starting point, that would be great as well.

hanayik commented 3 years ago

I just wanted to follow up on this- I submitted my input images on 3/22 to the link you provided.

hanayik commented 3 years ago

I checked your files, and all voxels removed by the ero flag had at least one zero-valued voxel in the 3x3x3 kernel, so it looks like all is fine with the output image.