Dear Experts,
I am trying to analyse a group of patients who have all had similar
surgery, and been scanned pre and postoperatively. I want to
compare the post-op diffusion data to the pre-op data using TBSS. I
decided to do this by trying to co-register the postop to the preop
data before then running tbss_2_reg for the preop images only, and
applying the derived warps to the postop data which should be in the
same space as the preop data. My question mainly relates to step
4/convertwarp, but any methodological comments/criticisms would be
appreciated. My processing steps are as follows:
1) Apply tbss_1_preproc to all pre- and postop data files
2) Apply tbss_2_reg with only preop data to derive the non-
linear warp and affine transformation for preop to target ----- get
<pre_to_target_warp>
3) Co-register postop to preop (affine/rigid body registrations
cannot cope well with the brainshift present which is non-linear)
- flirt –in <postop FA> -ref <preop FA> -out <filename> -omat
<matrix> -bins 256 –cost normcorr –searchrx -90 90 –searchry -90 90 –
searchrz -90 90 –dof 12 –inweight <surgicalmask> -interp trilinear
- fnirt –in <postop FA> -ref <preop FA> -config <FA_2_FMRIB58_1mm> -inmask
<surgicalmask> -applyinmask <1,1,1,1> -aff <matrix> -cout
<post_to_pre warp> -iout <post_to_pre image>
4) Combine the post-op to pre-op transformation with the pre-op
to target transformation so that one warp can be applied to postop
images to get them into target space
- convertwarp –ref <target> -warp1 < post_to_pre warp> -warp2
< pre_to_target_warp> -rel –out <post_to_target_warp> -relout
In this step applying the individual warps gives good registration between
the target and post-op images, (and each warp itself gives great
registration to respective target) but when I apply the product of
convertwarp though the target and post-op images are reasonably well
aligned, there is some brain shift visible (similar to what I see if I just
apply an affine transformation between pre- and post- images), which
definitely affects the stats analysis as all my significant voxels are where
the brain has shifted from preoperatively to postoperatively. So if anyone
could point me in theirght direction here - that would be appreciated.
5) Go back to the FA directory and place the
<post_to_target_warp> in the directory using a similar naming
convention to the <pre_to_target_warp> images. Also copy and
rename the pre_to_target_affine transforms as appropriate for postop images
(postop images are registered with the preop images and should therefore
need the same affine transformations to standard space as the preop images -
although my understanding looking at the tbss scripts is that because I am
using FMRIB FA as target these affine transformations are not needed ?)
6) Apply tbss_3_postreg and tbss_4_prestats as normal
Thanks.
Mahinda
|