Dear Dr. Taylor,
I inherited this code from ALICE toolbox. I ran the registration without step 2 (resampling), and it does not work (step 3 align_epi_anat.py runs forever to converge) and the result on visual check in afni is not ok. The only way the code works is as is, including step 2 (using the resampling at the native T1 resolution, no other resolution -dxyz works, see also the thread 'align_epi_anat.py registration variability' about a mo ago). If you can take a look using the actual data (t1, ct), I can upload those to you (I would need to know how). I am not sure to what extent the need for step 2 is because our CTs are anisotropic (0.43, 0.43, 0.8) vs isotropic MRI (1,1,1,)
If using step 2 as is, the registration is good. Then, I would need to know how to move a ribbon.nii (mask generated in freesurfer, aligns with T1) to the ct space, I have tried:
cat_matvec -MATRIX -ONELINE ct_shft.1D -I > inv_ct_shft.1D #inverting forward matrix from step 1 of the registration
cat_matvec -MATRIX -ONELINE ct_res_shft_al_mat.aff12.1D -I > inv_ct_res_shft_al_mat.aff12.1D #inverting forward matrix from step 3
3dAllineate -source ribbon.nii -base ct_res_shft.nii -prefix ribbon_shft.nii -1Dmatrix_apply inv_ct_res_shft_al_mat.aff12.1D
3dresample -input ribbon_shft.nii -prefix ribbon_res_shft.nii -master $ct -rmode NN
3dAllineate -source ribbon_res_shft.nii -base $ct -prefix ribbon_ct.nii -1Dmatrix_apply inv_ct_shft.1D
For some reason, the output ribbon_ct.nii looks aligned to the CT, but appears truncated, in afni.
Thank you very much, Octavian
Edited 1 time(s). Last edit at 01/02/2021 04:59PM by octavian lie.
