Spatial normalisation will shrink the ventricles to make them match those of
the young healthy subjects that constitute the tissue probability maps.
Because only about 1000 parameters are used to define the deformations, it is
not able to accurately model the local deformations that would be required to
shrink the ventricles without shrinking the surrounding tissue. This means
that the surrounding grey matter is also shrunk.
Modulation corrects for volume changes such that the total amount of tissue is
preserved. This means that the shrunken grey matter regions will become more
intense.
Intersubject brain registration is still not the "done deal" that many people
assume. Evaluations have shown that in terms of aligning landmarks, SPM does
a fairly good job - but there is still room for improvement. These
inaccuracies can make findings more difficult to interpret.
Best regards,
-John
> after routine segmentation in spm5 for vbm we see (in some datasets)
> relatively strong gradients in intensity within the image, especially
> midcingulate portions and ACC become bright (see attached examples).
>
> It seems there is a tendency to get this result in elderly patients with
> larger ventricles.
>
> With regard to the position of the cortex band, the results are satisfying.
>
> To my understanding, the translation parts of affine transformations do not
> have an effect on modulation (for example when the stretched out cingulate
> cortex is shifted to its normal(ised) position).
>
> We just wonder if these datasets need to be taken out.
>
> thanks a lot for any hints,
> Philipp
|