Hi,
It is normal to run BET on an image prior to using FLIRT.
If you ran BET on the T1_crop image then this is the image
you should use as the --ref for invwarp. If not, then just run
BET on T1_crop and then use the brain extracted image as
the --ref for invwarp - it will not be important if the original
registrations did not use this brain extracted image. However,
in general we do recommend brain extraction prior to using
FLIRT (and then using the brain extracted and non-brain-extracted
versions for FNIRT).
All the best,
Mark
On 9 Jun 2010, at 16:02, Chiara C. wrote:
> Dear FSL experts,
>
> here it is my problem:
> to extract the brain efficiently for quite a big group of subjects I
> follow this procedure:
> * I crop the FOV of the original T1 images (1x1x1): starting from
> 176x224x256 voxels I get 160x200x160
> * On the cropped images (T1_crop) I use FLIRT first and then FNIRT
> with the FLIRT matrix as initial guess (--aff) to register T1_crop
> to the template MNI152_T1_2mm
> * I apply the non linear transformation using applywarp
> * I bet the fnirted images asking BEt to output the brain mask (-m
> option)
> This process takes about 20 minutes for each subject.
>
> Now I would like to use invwarp on the brain mask to have it in the
> original (T1_crop's) space in order to mask each T1_crop with its
> mask and have only the brain in the original space.
>
> The instruction is:
> invwarp --ref=T1_crop --warp=T1_crop_coeff --out=T1_crop_coeff_real
>
> So I get this message:
> WARNING!! reference image has large non-zero ROI. Check that BET has
> run successfully on this
>
> Could anyone explain me what does it mean and why does it appear? Do
> I have to apply BET on T1_crop?
> Thanks,
> Chiara
>
|