| I am inquiring into the validity of the following analysis method:
| I determined parameters for realignment
| I resliced the images with these parameters
| I also entered the motion parameters as user specified regressors in the
| design matrix
| In looking at the results I used a t-contrast to look at the box-car and
| entered zeros for the motion parameters.
| Is this valid?
| My reason for this is due to the fact that the subjects moved at the
| beginning and end of the stimulus presentation.
This approach should discount false positives that are due to effects that
are linear with respect to the estimated motion parameters. However, it
probably also has the side effect of removing much of your true signal -
especially if the motion is correlated with the experimental tasks.
Currently, the image realignment uses a very simple six parameter rigid
body transformation model to register the images. However, this model
can not remove all motion artefacts. Even when the artifacts are only
very small, they can still appear very significant because of the
large number of degrees of freedom generally used in fMRI statistical
models. We are currently working on more complex registration models
that can remove more of the motion artifact than simple rigid body
transformations. So far, it appears to be working very well. We'll
keep you posted on progress.
Best regards,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|