Hi,
With any global affine registration it is hard to match all areas well
unless
the images are distortion-free images from the same subject. However, using
cost function weighting should make the fit care more about the areas of
higher weight. The question is - what weights did you use?
Typically if you were want a better fit around the ventricles, for
example, I would
put a weight value of around 5 or 10 over the ventricles and just
outside (to cover their
boundary) and then a weight of 1 everywhere else. Do not use 0
elsewhere unless you
totally want to disregard all other aspects of the brain fit - which
will make the whole
thing less robust.
Of course, if the fit you want (say the ventricles) cannot be well
approximated by
an affine transformation then no amount of weighting etc will do the trick.
Could you let us know exactly what regions you are trying to register
and the
details of your weighting function?
All the best,
Mark
Edward Vessel wrote:
>Hi -
>
>I have noticed that while the 'global' normalization for many of my subjects
>to the MNI standard brain are good, that particular regions show very poor
>local alignment. I had thought that perhaps by including a -refweight volume
>biased towards those regions I cared most about, I could improve this issue.
>However, including a volume of weights on the reference image doesn't seem to
>be changing my normalization much at all. Do these weights only apply to the
>initial search algorithm, or to the final affine transformation as well? Any
>other suggestions on biasing the normalization for certain regions/voxels?
>
>Please respond directly to me so I don't have to wait to receive the digest to
>get your comments =)
>
>thanks,
>Ed
>
>--
>Ed Vessel
>U. of Southern California [log in to unmask]
>Dept. of Neuroscience
>HNB, 3641 Watt Way http://geon.usc.edu/~vessel
>Los Angeles, CA 90089-2520
>(213) 740-6102
>
>
|