Dear FSL experts,
I attempted to register original T1 image to b0 DTI image, create ROIs at T1 space, and then apply registration matrix T12b0.mat to generate ROI masks in b0 space. T1 orients are qform_xorient RL, qform_yorient SI and qform_zorient PA, which are the results of recon-all in freesurfer. T12b0 image looks fine on GM masks that exclude meninges beautifully, but the frontal horn of ventricles shift a bit backward compared to b0 image that results in overlapping between white matter mask and ventricles.
Command lines:
flirt -dof 6 -in T1.nii.gz -ref b0.nii.gz -out T12b0.nii.gz -omat T12b0.mat
flirt -in WMmask.nii.gz -ref b0.nii.gz -out WMmask2b0.nii.gz -applyxfm -init T12b0.mat
PS, I use 1.5T image, so there is noise around skull in b0 image. I extracted brains from T1 and b0 (BET), apply the same command lines on extracted brains, but the problem remains. I also tried to register b02T1, convert b02T1.mat to T12b0.mat, and then apply T12b0.mat on ROI masks, similarly, the problem remains. I wonder if I did anything wrong? or How can I improve registration on this bit? Any suggestions are appreciated!!!
Best Wishes,
Joyce
|