AFNIans,

What is the best way to estimate activation in the presence of possible motion effects?

Specifically, I have data from a block-design fMRI task. For the subject in question there appears to be strong activation-related signal change in regions of the brain I would expect. However, I noticed that the GLM coefficient associated with the block task is slightly above average in the medial prefrontal cortex. The GLM coefficient map for the "pitch" motion correction parameter (produced by 3dvolreg) has a relatively strong negative magnitude in a region that is suspiciously similar in shape and location to the above-average MPFC region. However, the pitch motion correction parameter itself does not have a strong correlation with either the task hemodynamic response estimate regressor in the GLM matrix produced by 3ddeconvolve or a time-course file representing the absolute "on" and "off" times for the task presentation (a file with one row per TR; "0" between blocks, "1" during blocks).

I would rather not exclude this subject from further analysis but I also don't want to include activation estimates that are thrown off by motion effects.

While I've brought up a case for one single subject I'm curious about the question in general: is there another method to get a feel for how motion might be affecting the task GLM coefficient beyond looking at the motion regressor coefficient maps and the correlation between the motion parameter time-courses and the task time-course?

Thanks,
-John