When an image is resliced, it is only possible to write out those parts of the
resliced image for which there is data in the original image (ie, the voxel
in the resliced image should map to some point within the field of view of
the original). Even very small movements can lead to loss of data at the
edge of the FOV.
In voxels where there is missing data, some value needs to be used to fill in
these unknown values. Typically these values are set to zero. In fact, when
an fMRI time series is resliced, if a particular voxel position has an
unknown value at any point in its time series, then the entire time series
for that voxel is set to zero. The reason for this is that because each
volume of the time series moves in different ways, then there are likely to
be different patterns of zeros in the resliced images. Prior to doing the
stats, it is typical to smooth the data, and this would have the effect of
introducing movement effects into the data, which would arise because some
voxels have data and others have zeros - depending on the subject motion.
Best regards,
-John
On Tuesday 24 June 2008 00:30, Vanessa Carmean wrote:
> Hi SPM'ers,
> I am processing data using SPM5, and on a couple of subjects, the raw
> data looks fine, but after the realignment step, the images appear as if it
> they have been cut, removing the lower portion of the image. The motion
> in these subjects doesn't seem to be significantly different from any of
> the other subjects, and nothing seems different between the steps we are
> doing for these subjects and the subjects who are going through the process
> correctly. Is anyone familiear with what might be causing this?
>
>
> Thank you,
> Vanessa
>
> Vanessa Carmean, B.S.
> M.I.N.D. Institute
> University of California, Davis
> 2825 50th Street
> Sacramento, CA 95817
> (916) 703-0438
|