Hello,
this may be a function of the analysis threshold - you can easily check this
by having a look at the mask. After smoothing and with a loose proportional
and absolute threshold mask edges are usually larger than the MNI glass
contours. So you have to decide which areas you intend to include in your
analysis. As these out of brain areas are usually not normally distributed
with regards to the residuals results here tend to be falsely positive.
You could also produce masks from unsmoothed images and use the explicit
masking option if smoothing has caused the extension of your mask.
Best regards,
Philipp
|