Hi,
The problem here is that your images can been preprocessed in
such a way that most of the white matter voxels are set to an
intensity of *exactly* 110 (this looks like FreeSurfer pre-processing).
As a consequence, if FAST selects only voxels with 110 as the
white matter (which is somewhat likely) then there is zero variance
in this set which causes problems. Such a problem would never
arise in an image taken straight from the scanner.
You have three options as I see it:
1 - If you have a version of these images without the pre-processing
then simply use those images (after brain extraction) and it
should work fine.
2 - If you've only got processed images, through FreeSurfer, then
use the segmentation from FreeSurfer
3 - If you used FreeSurfer to do brain extraction only, then take the
current images and generate a binary mask from them and
apply this mask to your original images and then run these
new images through FAST.
All the best,
Mark
On 27 Oct 2010, at 03:36, Zheng Hui wrote:
> Hi Mark,
>
> Thanks for the reply. I have uploaded the *_struc_brain for two
> subjects. Subject SS001B is a problematic subject while subject
> SS007B is an ok subject. The session refeence is 206071.
>
> I can upload more subject/data if you need them.
>
> Regards,
> Zheng Hui
>
>
|