Hi,
Just to be sure I tested this by running run_first_all on an
image and the same image after fslswapdim to induce a
large transformation. The coordinates I got in the vtk
files were very different (mainly swapped) while the image
outputs looked appropriately registered to the respective
starting images. So I'm absolutely convinced that the
coordinates are in the native space of the original image
and not in MNI space. I guess yours just look aligned
because the original images are fairly well aligned.
All the best,
Mark
j janssen wrote:
> sorry, i now understand that even in native space, the number of
> vertices may be identical between meshes. however, i have plotted
> three .vtk's from three different subjects (after run_first). they
> look aligned to me, or not?
>
> thanks,
> -joost
>
>
> On Mon, Mar 16, 2009 at 2:19 PM, j janssen <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> all the meshes that are generated after run_first for a given
> structure have the same number of vertices. does that mean that
> they are saved in native space but thereafter aligned?
>
> thanks,
> -joost
>
>
>
> On Mon, Mar 16, 2009 at 1:49 PM, Mark Jenkinson
> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>
> Sorry - my mistake before.
> The meshes are actually saved in the native space, and are
> what would be used by the ReconNative option. The model
> is transformed into the native space at the beginning of the
> segmentation.
>
> All the best,
> Mark
>
>
>
> On 16 Mar 2009, at 11:12, Mark Jenkinson wrote:
>
> Hi,
>
> Going to and from "volumes" where these are just a
> collection of the
> vertices in a 1D order and applying FDR on this is fine,
> since FDR
> does not use any spatial information when calculating its
> correction.
> This is why it is the easiest to implement in this case,
> since the volume
> representation does lose all spatial information, but for
> FDR this does
> not matter.
>
> The mesh coordinates are all saved in MNI space (the space
> of the
> model) for the vtk files. When you perform vertex
> analysis in native
> space it transforms these coordinates before doing the
> vertex analysis.
> Writing transformed meshes can be done using the --doMeshReg
> option with the -f option to specify the flirt matrix (the
> option usage
> message for this is currently not correct). However, this
> won't give quite
> the same space as that used for the useReconNative and
> useRigidAlign
> options, as these alignments are calculated internally to
> the vertex
> analysis.
>
> It is probably worth waiting for a few days, if you can,
> since we are going
> to release a patch with several changes in FIRST. In
> particular, we have
> substantially improved the statistical sensitivity of the
> vertex analysis. We also
> will have easier options for performing many of the steps
> for vertex analysis
> and FDR correction.
>
> All the best,
> Mark
>
>
>
> On 16 Mar 2009, at 09:18, j janssen wrote:
>
> Dear all,
>
> 1st point: some time ago i got code to do FDR
> correction after vertex-wise analysis with FIRST.
> actually, the FDR was applied to images, not surfaces,
> as FDR was not implemented yet in run_mesh_utils.
>
> - is the FDR correction, i.e. going from surfaces to
> volumes, applying FDR to volumes, and going back to
> surfaces valid?
>
> 2nd point: run_first for a given structure and
> multiple subjects generates .vtk's for the given
> structure. i noticed that the meshes are not in native
> space but aligned.
>
> - in which space are the meshes saved by run_first?
>
> - i used first_utils with -vertexAnalysis
> --useReconNative --useRigidAlign, works fine. however,
> i would like to write out the individual meshes, in
> native space and rigid aligned (--useReconNative
> --useRigidAlign). how would i go about this?
>
> thanks,
> -joost
>
>
>
>
>
>
>
>
>
>
>
> ------------------------------------------------------------------------
>
|