Hi,
I am using fnirt command to register to T1 IBSR images in a masked region. I
am using a command as follows:
flirt -in image1_brain.nii -ref image2_brain.nii -omat affine1.mat -out
image1_flirted.nii
fnirt --in=image1.nii --ref=image2.nii --iout=image1_fnirted.nii --cout=warp --
refmask=MNI152lin_T1_1mm_subbr_mask.nii --aff= affine1.mat --sumsample=
8,4,2,2
I expect that image1_fnirted.nii has better alignment specially in subcortical
region (as applying refmask) than image1_flirted.nii. But, the results do not
support this. After applying some processing to evaluate the accuracy,
image1_fnirted.nii has a volume around 25% more than image1_flirted.nii in
subcortical region and has not a good match to the reference image. I mean,
fnirt expands the volume and does not result in a good alignment in that area.
I checked it several time using IBSR datasets that I have the segmented
image as gold standard.
Is there any other settings in fnirt command that result in a robust dependable
results?
Siamak Yousefi
|