> although the problem of removing movement related confounds was discussed
> intensively during the last weeks I do not completely understand the
> effect of this procedure on my data:
> A subject was scanned for about 12 min. During this time the head
> performed a slow drift (1-2 mm translation and 5 deg yaw). Because
> TR was 6 sec, I would expect that this movement should have almost no
> influence on local image intensity after reslicing.
>
> So I compared "no adjustment" to "2nd order adjustment" and I got
> almost completely different results:
> The results from the "no adjustment" data look reasonable, while the other
> analysis shows huge regions of activation and an area of deactivation
> which is unfortunately located in the ventricular system. So there
> might be a small correlation of movement and stimulation. Is it
> possible that this correlation is overcompensated by the "2nd order
> adjustment"?
My own personal feelings are that movement effects should probably be
modelled as some type of covariate of no interest in the statistics part
of SPM. I think what you are probably getting is the effect of incorrectly
removing confounding signal from your data. The assumption that allows
the adjustment step to be performed seperately from the rest of the statistics,
is that the estimated motion parameters are in no way related to the paradigm.
This is clearly not the case in many fMRI experiments.
Perhaps someone else would like to comment.....
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|