Hi,
> I'm wondering whether we can use FIRST vertex analysis on manually
> traced hippocampi, which are saved as mask nifti files.
It used to. A while back, I had recoded FIRST to be more efficient and
removed the option. At the time, it was virtually unused. The models could
be fit based on overlap of the filled surface with the mask (could also
look for gradient at the surface). It wouldn't take very long to add it
back into the FIRST code. Could potentially add this in the next week or
two if your interested.
> Could we make the
> segmentation detect our manually traced hippocampus by setting the
> hippocampal region in the T1 to be extreme, or at least homogeneous?
This wouldn't necessarily work since it is not just looking for a strong
gradient, it model the intensity distribution based on the training set.
This would mean that an artificial gradient where it doesn't expect it,
could mess up the fit.
> Alternatively, I read on the list that a model could be created on a set
> of segmented images, so is there software available to do that?
Indeed, this can be done. Currently the software is not released. As far
as I am aware, there's no immediate plans for releasing/supporting it
(Steve can correct me on this if I'm wrong). How many manual segmentations
do you have?
cheers,
Brian
> Hi all,
>
> I'm wondering whether we can use FIRST vertex analysis on manually
> traced hippocampi, which are saved as mask nifti files. I see that the
> segmentation has an option -loadBvars that uses a previous estimate as a
> starting point, but the input is not a mask file. Could we make the
> segmentation detect our manually traced hippocampus by setting the
> hippocampal region in the T1 to be extreme, or at least homogeneous?
> Alternatively, I read on the list that a model could be created on a set
> of segmented images, so is there software available to do that?
>
> Any thoughts appreciated!
>
> Best wishes,
> Paul
>
>
|