It is possible that your original images contained negative values.
For PET images reconstructed with filtered back projection, there is
nothing to stop voxel values from being negative. Reslicing images
with sinc or B-spline interpolation can also introduce negative
values. I would suggest not truncating these values at zero, as this
is likely to introduce bias into your data.
Best regards,
-John
On 22 August 2011 18:04, Clifford Workman <[log in to unmask]> wrote:
> Hi SPM users,
>
> I'm working with functionalized FDG images acquired in 6 subjects over 3 time points on an HRRT. I used realign (MRs were not available for co-registration) to align the images to a mean image of the three conditions per subject. I then used the mean images to derive normalization parameters. I noticed that, following realignment, negative signal intensities appeared in the CSF. I tried manually removing the negative values to see if the problem reappeared after normalization, and it did. I’ve tried playing with the options for both SPM processes, but to no avail.
>
> I am concerned that these negative signal intensities might introduce error -- for example, when smoothing the normalized images, we may smooth negative values into smaller subcortical structures adjacent to the ventricles. Has anyone else encountered this problem or have insight as to why it might be occurring? I did see something similar posted, though I am not sure how applicable it is (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;9216a369.1106). Thanks.
>
|