I am running some preprocessing whereby i do some motion correction, functional to volume registration and some talairach transformation, in each case, saving the transformation matrices. I perform it on a single brik template first so as to speed up the process, check visually registration, etc.

Once I have a this single brik template with a good registration on my talairached structural I'll try to combine all my transformatoins into one step for the actual data.

like so

cat_matvec motion.aff12.1D func2struc.aff12.1D anat_at+tlrc::WARPDRIVE_MATVEC_FOR_000000 > mot.func2struc.tlrc.aff12.1D

3dAllineate -1Dmatrix_apply mot.func2struc.tlrc.aff12.1D -input rawdata+orig
...

However, if i split up the steps, like so

cat_matvec motion.aff12.1D func2struc.aff12.1D > mot.func2struc.aff12.1D

3dAllineate -1Dmatrix_apply mot.func2struc.tlrc.aff12.1D -input rawdata+orig -prefix rawdata.mot.func2struc

@auto_tlrc -apar anat_at+tlrc -input rawdata.mot.func2struc+orig


The results are often noticeably different. Often the former method does not seem as good as the original registration with a single brik template


Is this just an issue of accumulation of errors? Or is there maybe some difference between running @auto_tlrc -apar and 3dAllineate?

Thanks
Drew