kamran kazemi wrote:
> Dear SPMers
> i want to ask that if SPM5 do the intensity normalization during the
> normalization process?
Hi Kamran,
If you mean correction of the slowly varying "bias field" intensity
variation, due to field non-uniformities, then yes, it's part of the
Unified Segmentation process - described in
Ashburner, J. & Friston, K.J.
Unified segmentation.
Neuroimage, 2005, 26, 839-851
If you mean between scan and/or between subject intensity variation
due to sequence changes, different scanners, or other variability,
then no. Since this is very dependent on modality and sequence it
doesn't sound to me like it should be in the unified model, but I
could well be wrong about that.
This is something I've been looking at, the main difficulty is
deciding what you really want to do, since there are quite a range of
options from simple multiplicative re-scaling, through affine
intensity rescaling, to higher-order corrections, and a range of
things one might be interested in basing the correction on, and also a
decision of when to perform the re-scaling. For example, one approach
would be to use SPM to segment the images, then scale the intensities
of each original image so that the average intensity within a region
defined by the SPM5 WM segmentation had a fixed value. Using an affine
re-scale you could set CSF and WM to have specified values. The
problem with this approach is that I've heard reports that the SPM
segmentations themselves are affected by intensity differences.
One option is to manually delineate one or more ROIs in each image (or
in template space if the images have been accurately warped to it)
and to determine the average intensities within these regions. This is
a lot more time-consuming (you should check the roi overlay on each
image, to be sure that there are no unusual artefacts or pathology
within the region) but I think would be the best method, if you are
concerned about absolute intensity variations.
Hope that helps,
Ged.
|