Dear Maite,
I think you misunderstood Jesper.
The suggestion is to get a transform from your original structural images
(not segmented) to the MNI space (using FLIRT and FNIRT). You then
also segment the image with FAST in the native space and only then
do you take the GM output from FAST and apply the non-linear transformation
that you got previously. There is no need to choose thresholds as it will
appropriately transform all the partial volume quantities, although if you want
them to represent the original absolute volume measures then you will need
to modulate by the Jacobian.
I hope this helps.
All the best,
Mark
On 13 Dec 2011, at 11:37, Maite Termenon wrote:
> Dear Jesper,
>
> thank you very much for your quick response. For sure, I will try both options you suggest. If apply Fast at the end of the pipeline, I obtain a probability of segmented GM. To obtain a final image, I guess I have to choose a threshold, create a mask and apply it to non linear registered final image. But, how to know which is an addecuate threshold value?
>
> Thanks,
>
> Maite
>
|