Hi,
There are many reasons that the segmentation might not be ideal,
and several things that you can change, not just using the priors.
It therefore might be good to explain what the main problem is and
see if modifying the MRF beta parameter, the bias field smoothing,
or the number of iterations might be more appropriate than using
priors. It would also be good to describe how you are making the
judgement that things are not so good.
If you really want to use priors (which mainly changes the segmentation
of subcortical structures) then you can use the tissue priors in:
$FSLDIR/data/standard/tissuepriors/
rather than defining your own.
All the best,
Mark
On 6 Jan 2011, at 09:06, Ha Ho wrote:
> Hi,
>
> Thanks a lot for the prompt reply. The results after the segmentation are not very good so I thought it could be improved using the probability maps.
>
> Would be great if you could help me here.
>
> Thanks a lot and best regards,
>
> J
>
> On Thu, Jan 6, 2011 at 4:51 PM, Mark Jenkinson <[log in to unmask]> wrote:
> Dear June,
>
> If you don't already have prior probability maps to hand, why do you want to use them?
> We generally do not use them when running FAST.
>
> All the best,
> Mark
>
>
> On 6 Jan 2011, at 06:53, June L wrote:
>
> Hi,
>
> I tried to create the 3 images of Grey, White Matters and CSF to feed in the -A option. However, I feel that I misunderstood the probability maps. What I have is the image, for example Grey Matter has value in the Grey Matter region, and 0 elsewhere. But it is not the probability map right?
>
> I would really really appreciate it if you could give me some guide here.
>
> Thanks a lot!
>
> J
>
>
|