Hi all,
Just to clarify the situation, and the query,
Jeff Browndyke, Ph.D. wrote:
> I posted a couple of days ago about apparent image intensity changes
> following reslicing with Ged's modified resize_img script. Discussion with
> Ged backchannel revealed that the script was detecting negative intensity
> values in the data, which were then being scaled to zero and resulting in
> high intensities in non-negative voxels.
The reslicing isn't really detecting and zeroing negative voxels; it
was just that they were isolated negative voxels surrounded by
reasonably strongly positive ones, so when interpolating for the
reslice, they happened to end up positive.
The concern was partly just a display issue, and partly due to padding
around the resliced image. The presence of negative voxels in the
originals makes the image display a little brighter (e.g. -26 maps to
black, rather than 0 to black, and other intensities get shifted up a
little too). In terms of means/medians etc. the issue here was that
the resliced image includes more zeros around the edge (the world
bounding box is fixed, but in voxel space, the reorientation means the
resliced image requires a bigger voxel BB). The mean/median/etc of the
*non-zero* voxels in the original and resliced images match very well
--- the reslicing isn't doing anything peculiar.
> I went back to our original image
> data and found that the original files did not contain negative intensity
> values. These only appeared after undergoing the SPM2 Realign procedure.
This however, sounds very peculiar to me... Jeff assures me that only
realign is being used, not reslice, in which case, my understanding is
that only a new .mat is created --- intensities shouldn't be altered
at all. Can anyone see anything obvious I'm missing?
Thanks,
Ged.
|