Dear Stefano,
There is an easy way to take a 2D registration transform and apply
it to all slices.
You are right in that a reference _volume_ is required.
The reference image is used to determine the size (number of voxels
and voxel dimensions) of the output image. Hence if you want a
3D volume output, you need a 3D volume reference of exactly the
right size. In your case this should be a 3D volume with the
same slice dimensions as your 2D reference slices, but with the
correct number of voxels in the z-dimension (and with the correct
slice thickness set - same as the input in this case).
So, if you have a volume like this, then just use it as the reference.
Otherwise, you can make one using MEDx or avwcreatehd.
(Also, if your input is in the form of slices rather than a volume
you can make a volume using a command line call to "avwmerge -z")
All the best,
Mark
"Marenco, Stefano (NIMH)" wrote:
>
> Thanks, that's a good suggestion.
>
> Here is another problem. I want to do a 2D registration on single slices. No
> problem there. Then I want to use the resulting transformation matrix to
> correct the position of a volume in plane (i.e. I want to correct each slice
> in the volume by the same parameters I used for the 2D registration). When I
> used the GUI, only the first slice of the volume was corrected. Is there a
> way around this? Stefano
>
> -----Original Message-----
> From: Stephen Smith [mailto:[log in to unmask]]
> Sent: Wednesday, January 16, 2002 5:23 PM
> To: [log in to unmask]
> Subject: Re: [FSL] 2 channel segmentation
>
> Hi.
>
> On Wed, 16 Jan 2002, Marenco, Stefano (NIMH) wrote:
>
> > I do a segmentation from 2 channels (a proton density weighted and a t2
> > weighted image) and it works really nicely.
> > Then I take the input images and I register them to a third image. The
> > registration works fine.
> > I repeat the segmentation with the same parameters that I used initially
> and
> > I get a completely different result.
> > Can anyone explain this to me?
>
> I assume that the 2 channels are registered to each other before
> registration to the third image. Are you applying exactly the same
> transformation to both? If not, their registrations to the third may be
> subtly different, which may be causing this. Otherise, maybe it's
> intepolation effects which may be causing slight blurring. You could try
> using sinc interpolation in FLIRT instead....
>
> thanks :)
|