Hi - a couple of comments that might help on this:
Are you _sure_ that all the values in the _seg image are all 1, or is
it possible just that the initial display range in FSLView is 0:1 and
you need to increase the Max of the display range?
Also, we don't generally recommend running segmentation on resampled
images, it generally should work better when run on the brain-
extracted native-resolution image.
Cheers, Steve.
On 7 Sep 2007, at 09:05, Peter Koopmans wrote:
> Hi everybody
>
> I'm trying to use fast4 on a dataset and get 6 output files:
> - name_pve_1, name_pve_2 and name_pve_3
> - name_mixeltype
> - name_pveseg
> - And finally, name_seg
>
> The first five all contain results which look fine. (pveseg for
> instance,
> contains a somewhat crude segmentation result)
>
> The problem however is in the name_seg file. All voxels have a
> value of 1.
> So while the segmentation algorithm at some point knew
> approximately how to
> divide the image somewhere along the line it lost it again.
>
> To check if it was a hyperparameter problem (maybe the algorithm
> tries to
> obtain a very smooth result by turning everything into a single
> class) I
> once ran it using the default value (0.3) and once letting it
> decide by
> itself (-H -1). Didn't work though.
>
> This is how I call fast4:
>
> fast4 -t 0 -v -n 3 -H -1 -o output_name input_name
> ----
> Starting Single Image Segmentation
> T1-weighted image
> Imagesize : 370 x 300 x 250
> Pixelsize : 0.15 x 0.15 x 0.15
> ----
>
> This is an interpolated dataset so it is a bit large but that
> should not be
> a problem should it?
>
> Has anyone encountered this before or does anybody know how to
> solve it?
>
> Many thanks,
>
> Peter
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|