hi Michael
> I wonder if it is necessary to include motion regressors for the analysis of spatially normalized BOLD images.
> Specifically, I have used a preprocessing batch to estimate the necessary realignment and normalization parameters of BOLD images based on the corresponding segmented anatomical images. When checking several of the resulting imges with the "SWS-" prefix for registration with a normalized EPI template I see very good registration results.
>
> So do I then have to incude the realignment parameters in the model. Because my SPM's look different when I include them.
>
> I am thinking that I should not include the motion parameters in that case because the original image headers have already been overwritten to reflect the realignment and normalization to standard space.
Including the motion regressors doesn't really have to do with the
spatial normalization; rather, it is intended to remove "activations"
that may be a result of head movement (which can result in
fluctuations in the signal you record). These tend to occur near
tissue class boundaries where motion would induce the largest contrast
differences. There is an example of this in Chapter 29 of the SPM
manual (the face data set; Figure 29.15 I think). The idea is that if
you DON'T include movement parameters in your model, then there is a
chance that task-related "activation" is really just due to movement
effects. So I would say a conservative rule of thumb is to include
them.
On the other hand, especially if for some reason you have
task-correlated motion, then you may also be removing effects of
interest. I'm fairly sure there are a number of previous posts on the
list on alternative methods for taking into account movement
parameters (as opposed to just including them as columns in your
design matrix)...
Hope this helps!
Jonathan
|