Dear Mahinda,
> Will invwarp (as implemented in tbss_deproject) work as it is meant
> to with
> warps produced by convert warp. I have a group of subjects analysed
> using
> TBSS. Half have been normalised to the target FA image using a
> single warp,
> and the other half using a combination of 2 warps produced by
> convertwarp.
> When I apply tbss_deproject 2, the results are fine for those
> subjects who
> had a single warp applied but not the other half (results obviously
> in the
> wrong place).
I had a look at inverting one of your combined warps, and it looks
like it has worked fine. The command I used was
invwarp --warp=mahindas_combined_post2standard_warps.nii.gz --
ref=po_L_EP.dtifit_FA --out=mahindas_combined_post2standard_warps_inv
and then I did
applywarp --ref=po_L_EP.dtifit_FA --
warp=mahindas_combined_post2standard_warps_inv --in=$FSLDIR/data/
standard/FMRIB58_FA-skeleton_1mm.nii.gz --super --
out=skeleton_warped2standard_using_mahindas_combined_inv
In order to see if the resulting transform would map the skeleton onto
the "post-space". The results look very good (if I'm allowed to say so
myself ;-) ).
So, repeat my commands above and ensure that you get sensible results.
If you do, it means that the problem must reside in the way you apply
tbss_deproject. The obvious possible problem, as already pointed out
by Gwen, is if you haven't made sure that your combined warps are
named the way that tbss_deproject expects. The easiest way for you to
work out the naming convention is to look how TBSS has named the warps
in the case where you didn't use the combined warp strategy.
Good Luck Jesper
|