Dear Vitaly,
|> SPM extends this a bit, by using one of the
|> spare fields in the .hdr file to 'scale' the image (the scale factor).
|
|What is the purpose of doing that? Is it just to increase the dynamic
|range of the image?
The scale factor is useful if the voxel values in the individual images
have meaning either relative to other images, or in some physiological
sense. For example, my water activation PET images have voxel values
in counts per voxel per second, and binding maps calculated for ligands
may be physiological values. These numbers are unlikely to
range from 0-255 (in the 8 bit case), but can be rescaled to this range
by using the scale factor. An alternative would be floating point images,
but these are a) larger and b) not displayable in the earlier (pre AVW)
versions of Analyze.
|That is, SPM itself does not scale the images? Then, the scale factor is
|taken into account when SPM does global normalization?
The routines that SPM uses to get values from images, do multiply by
the scale factor before returning image values, and this is
also true in the calculation of mean voxel values for global
normalisation. When writing out realigned images etc it merely
divides the new image values by the original scale factor, and
writes the original scale factor into the new header.
This is why the scale factor remains the same after realignment
and spatial normalisation. The exception to this is writing of the
mean image, where the scale factor is calculated from the image
maximum to give the optimum dynamic range.
|Now, to our specific problem. Our images are all 16-bit, and we use the
|full dynamic range, so the maximum value is 32767. The field Funused of
|the Analyze header is always 1. This field is not changed during any of
|SPM preprocessing steps, so I assume that in my case, no scaling is
|involved, right?
Well yes, in the sense that the scaling is one, and so the
scaling has no effect on the image values.
Hope that helps,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|