Hi Erik,
Without seeing the images it is difficult to say what exactly is going on, but one thing that I can think of is the following: If you have a very tight brain mask in the second image, it might be that there are too few background voxels for FAST to consider it a class and it might get a better fit by splitting the CSF into two classes instead.
If that is not what is going on, I’m happy to have a look at the images!
Best,
Eelke
> On 22 Feb 2017, at 10:09, Erik Olsson <[log in to unmask]> wrote:
>
> Hi,
>
> I have used FAST on some acquisitions from focal coronal 2D T2 TSE sequence (0.35 x 0.35 x 2 mm, 512 x 512 x 46) images.
> The problem is that one acquisition gets a different classification than the others and I cant see any important difference with this MRI volume.
>
> The command I used on the bet skull stripped file was:
>
> fast -n 4 -t 2 -g -o "${OUTPUT_DIR}/tissueT2.nii.gz" "${T2}"
>
> The normal classification entails that in the tissueT2_pveseg.nii.gz file 1; CSF, 2; GM, 3; WM and 4; background are accurately classified. In the deviating case the classification results in 1; subarchnoid CSF, 2; GM, 3; WM and 4; ventricular CSF & background. (background is the remaining skull or meninges parts from the skull stripped file)
>
> I also have tried to use the --manualseg option without success (should the mean intensities be specified in a certain order with a newline as separator?).
>
> best regards
> Erik
|