Dear Andreas,
The results from FAST can vary with the quality of your original images.
The main parameters to try varying are:
-H
-R
-O
and if you think the bias field estimate or restored image are not
optimal then try changing
-l
-I
(that is, lower-case "ell" and upper case "eye")
Although we have tried to set the default values of all of these
parameters to give good results, different acquisitions with
different resolutions and contrast-to-noise-ratios mean that
no one set of parameters is ideal for all cases. So try varying
the above (roughly in the order written) and hopefully you will
get a segmentation that you are happy with.
All the best,
Mark
On 2 Feb 2011, at 20:03, Andreas Glatz wrote:
> Hi,
>
> I also discovered recently that fast v417 behaves differently than fast v353. Although I don't have a lot of experience in manually segmenting MRI images, I think, based on the attached image collage, that the WM mask from fast v417 (2nd image from right) has rather a lot of holes in it. I compared it with the WM mask from fast v353 (4th image from right) but also with the WM mask after I removed the bias field with v353 and segmented the restored image with v417 (3rd image from right). Personally, I find that the 3rd image from right (bias field corr. v353, segm. v417) looks a bit better than the others... so I think that the bias field correction of v417 is problematic in my case. Can you give me a hit how I could improve it?
>
> Thank you.
>
> Andreas
> <FastComp.JPG>
|