Dear AFNI experts,
I have previously run a dataset through the afni_proc.py pipeline, with affine TT spatial normalization.
For a new collaboration project based on these data, I need to convert all the pre-GLM data to MNI space (EPI+T1). Could anyone point me to the quickest/best way to do it (I have 45 subjects, with 6 EPI runs per subject)?
More specifically, would you advise to Qwarp the original data from native space to MNI, or could I just go from the TT-space data to MNI (perhaps getting better results this way, as the images are already quite closely aligned)?
As for the T1 image (1mm-isotropic), I have the following files:
- t1+orig: original
- t1_al_keep+orig: skull-stripped, aligned to EPI data
- t1_al_keep+tlrc: skull-stripped, aligned to EPI data, and affine-warped to TT
- anat_final+tlrc: same as t1_al_keep+tlrc
Plus various transformation matrices:
- t1_al_keep_mat.aff12.1D: from T1 to EPI
- t1_al_keep_e2a_only_mat.aff12.1D: from EPI to T1??
- t1_al_keep.Xaff12.1D: from t1_al to TT??
Note that I don’t have a non-skullstripped version of either the EPI-aligned T1, or the EPI-aligned TT-warped T1 (but I can easily make them by applying the appropriate transformation matrices)
For the EPI data, I have the following files:
- pb01.subj.r.tshift+orig: slice-timing corrected EPI data (6 runs)
- mat.r0.vr.aff12.1D: volreg matrices for each EPI run
- pb02.subj.r0.volreg: TT-transformed EPI data
Just to get my feet wet, I tried to launch an MNI non-linear warping of a non-skullstripped version of a T1 (aligned to EPI) volume, with the command:
auto_warp.py -base MNI152_T1_2009c+tlrc -input t1_aligned+orig. -skull_strip_input yes
but that seems to be taking a really long time (1 hour and counting).
**** UPDATE ****
The Qwarp has finished and the results look really good. I also tried to apply the warp field to the slice-timing corrected EPI runs with a command like the following:
3dNwarpApply -source pb01.subj.r01.tshift+orig. -nwarp ‘anat.un.aff.qw_WARP.nii anat.un.aff.Xat.1D mat.r01.vr.aff12.1D’ -master anat.un.aff.qw_WARP.nii -newgrid 2.0 -prefix ep_r1_Qwarped
which works well too (it took me a while, though, to understand that I needed to sandwich the “anat.un.aff.Xat.1D” matrix between the qwarp and the volreg matrices: in the 3dQwarp help file, it says that if you use the -allineate option, " then you do NOT catenate the affine and nonlinear warps in 3dNwarpApply, since the output nonlinear warp will ALREADY have be catenated with the affine warp; maybe the help of auto_warp.py should be updated to reflect the correct concatenation).
But am I on the right track here, generally speaking? Could I shave off some time by using the already skullstripped T1 (from old @auto_tlrc) as input to auto_warp.py?
Alternatively, is there a quicker way to go from TT to MNI that trades off some of the quality of the Qwarp results with a much quicker computation?
Thanks in advance for any suggestion!