Hi Mark,
Let me explain it. For each subject, we got a 2-sec full-brain epi, two runs
data for fMRI, and one t1 structure image. Our TR of 1.7 seconds for fMRI
data requires that the cerebellum is cut off. We learned that FSL does better
if we have a whole brain epi to register to before it registers to the T1 image.
We feed the full-brain epi (after BET) into the initial image for registration, and
the t1 structure image (after BET) for main image. I did not find the problem
for basic results of run1 and run2, however it may have happened. After I
cimbined the two runs with fixed effect, I found for some post-stats results,
some red activated areas were beyond the background brain image, which
were obvious in OFC and occipital areas. We think something must have been
wrong, and registration part should be the first part we need to concern.
BTW, I just ran Renderstats - GUI with the basic feat result for run1, found
the activated areas were beyond the background brain image too. So,
something was wrong for the first analysis. We need your help for this issue.
|