Hi,
It is quite likely that this is due to the downsampling.
However, if you look at the floating point values you
should see values *between* 0 and 1 (e.g. 0.3) and
you can decide how to threshold them yourself. If you
are concerned that the mask is becoming too small,
then choose a low threshold and see if that helps.
To make a binary mask from the floating point values
and a threshold just do:
fslmaths floating_mask -thr threshold_value -bin binary_mask
Hopefully this will solve the problem.
Note that if you do not get floating point values in
your output, you might need to specify "-d float" in
the applywarp call.
All the best,
Mark
On 21 Oct 2009, at 18:16, Anastasia A Ford wrote:
> Hi,
>
> I registered my native FA images to FMRIB58 template and the
> alignment was
> very good. Next, I drew the masks on the template and I would like to
> register them back to the native FA space. I used the following
> invwarp command:
>
> invwarp --ref=../dti_FA.nii.gz
> --warp=FA2_FMRIB58_inweight_stroke_nonlinear_transf.nii.gz
> --out=FMRIB58_2_FA_inweight_stroke_nonlinear_transf.nii.gz
>
> This only took about 30 min. When I use apply warp to regiter my
> FMRIB58
> masks to the native space the masks come out very small, only a few
> voxels.
> I tried using all three available interpolation methods and all three
> produce fairly similar results. Could it be the issue of resolution?
> My
> original FA is 2x2x2mm
>
> Sincerely,
>
> Anastasia
>
|