5TT.mif and nodes.mif have differences in scale

image

I try to create the labelsgmfix.mif using apar+aseg.nii and T1w_acpc_dc_restore_brain.nii.gz files. I have created the aparc+aseg.nii using freesurfer following my own parcellation process. The T1w_acpc_dc_restore_brain.nii.gz has been downloaded from HCP server. I am not getting clear idea why the image is coming up like this. I have used the following command:

labelsgmfix nodes.mif T1w_acpc_dc_restore_brain.nii.gz fs_default.txt nodes_fixSGM.mif -premasked

I also overlay the nodes.mif image (it was generated from aparc+aseg.nii) onto 5TT.mif image.
5TT.mif image was generated using fsl algorithm as given in the tutorial. I can see the below image.

file:///home/sd4n15/Pictures/5ttoverlaynodes.png.
It looks like both images have some scaling difference.
But I am not be able to understand how to reduce the scaling difference.

If you give me some idea why my image is coming up like this will be very appreciated.
Thanks
Sarbani

Hi Sarbani,

I wouldn’t refer to this as a “scaling” difference; that word would typically refer to either the brain in one image being enlarged / shrunken with respect to the other, or the intensities within the images being larger or smaller compared to one another. This is simply a misalignment, most notably a translation.

labelsgmfix performs segmentation on the input T1 image, but maps those segmentations onto the input parcellation image. Therefore, if those two input images are not aligned with one another (in scanner space), then information drawn from one image cannot be faithfully mapped to the other. Your second screenshot suggests that this is indeed the case: I suspect that if you were to overlay aparc+aseg.nii on top of T1w_acpc_dc_restore_brain.nii.gz, you would see the same misalignment, entirely independently of any MRtrix3 commands. Perhaps by applying your own segmentation process, the image header transformation is being set to default, or some requisite transformation is not being applied?

Rob