> Dear SPMlist members,I have realigned SPECT images using the coregister
> only option. Having done it twice on the same set and by the same
> algorithm, I thought that the second run would use the information in the
> .mat-files of the previous run. Thus, I expected the estimated parameters
> for the second run as displayed in the output graphics to be considerably
> lower than in the first run. However, they are identical. Does it mean that
> in the second run the entire matrix of the .mat-Files (including the
> information of the first run) gets displayed, i.e. that the second run
> didn4t gain anything (since the difference to the first run amounts to
> zero)?
That's it. In SPM96, the difference between the successive realignments
would be shown, but in SPM99 it plots the relative positions between the
images. It is a good sign that the plot after the first run is the same as
the plot after the second.
> Do people have any suggestions for "progessive" realignments? I have
> started with MI-coregistering (coreg only) and then used realignment at its
> highest quality. Should I use resliced images or does that introduce
> additional errors? I am also wondering about the option for weighing target
> images but didn4t find anything in the spm_realign documenation. I have
> used weighing because the image intensities did somehow differ between the
> SPECT runs. Therefore, I picked the individuals images of higher
> intensities / better contrast as the weighed target. Is that kosher?
MI-coregistering followed by coregistering by minimising the sum of squared
difference (as in realignment) should hopefully just give you the same answer
as if you had just used the realignment.
Ideally, I wouldn't suggest using resliced images, as these contain regions
of zeros that can look like edges, which bias the estimates.
There is some documentation on spatial weighting in spm_realign.m.
Basically, you need to pass a PW field with the flags structure, where PW is
the filename of a weighting image. When you say the image intensities
differ, do you mean that the overall scaling is different? If this is the
case, then the realignment should take care of this as a scaling factor is
included in the model. As the second pass involves aligning all images to an
average, do you mean that you want to create some kind of a weighted everage
of all the images? This can't be done directly, but if you really wanted to,
you could tweek the image scalefactors so that the images were weighted
appropriately beforehand.
I'm not sure about the weighted target thing. PET/SPECT realignment is a two
pass procedure that realigns all images to the first, then creates a mean of
these realigned images to which all images are aligned to again.
Best regards,
-John
--
Dr John Ashburner.
Wellcome Department of Cognitive Neurology.
12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420
http://www.fil.ion.ucl.ac.uk/~john
mail: [log in to unmask]
|