I would strongly suggest that you include smoothing in the the
generation of the spatially normalised images. Without smoothing, there
can be aliasing effects (which is what you are seeing).
The actual warping of the images is done slightly differently, with the
aim that as much of the original signal is preserved as possible. This
essentially involves pushing each voxel from its position in the
original image, into the appropriate location in the new image - keeping
a count of the number of voxels pushed into each new position. The
procedure is to scan through the original image, and push each voxel in
turn. The alternative (older way) was to scan through the spatially
normalised image, filling in values from the original image (pulling the
values from the original). A minor disadvantage of this approach is
that it can introduce aliasing artifacts (think stripy shirt on TV
screen) if the original image is at a similar - or lower - resolution to
the warped version. Usually, these effects are masked by the smoothing.
Best regards,
-John
On Thu, 2009-11-12 at 13:18 +0000, Christoph Berger wrote:
> Hello, I tried to follow the previously posted instructions (by
> Christian Büchel, 02.09.2009) to use the dartel toolbox for
> normalisation of fmri images. I get also strange results in the
> individual masks after normalize them to mni space with dartel. Please
> see the attached picture. On the upper side you can see a individual
> mask after first level stats for this subject. Below is the result of
> the normalize to mni space dartel job. Are these effects also caused
> by image intensities below threshold for the mask? But the undeformed
> mask looks fine to me. What could be the reason for this? kind regards
> Christoph Berger
--
John Ashburner <[log in to unmask]>
|