Stacy Suskauer wrote:
> In processing fMRI data with SPM2, I have noticed that the time point chosen
> for spatial normalization impacts the final contrast map. That is, if
> I normalize the same data using time point 1, 60, or 116, I get varying
> contrast maps which appear to be the same pattern of activation at
> different thresholds (even though each contrast is thresholded with the
> same parameters). Has anyone else had this experience? Why is this
> happening? What is the best solution to issue (is there a "best" time point
> to choose for spatial normalization?)?
>
> Thank you for your assistance,
> Stacy
>
>
Yes, I have certainly noticed this. The take home point point,
I think, is that the normalization process can have a huge impact on the
outcome. Therefore you should have a mechanism in place that makes
very sure you are doing it the very best you can.
I know some disagree, but I personally like to co-register to a structural
image, and use the structural image for the normalization process. By
the way, this sometimes leads to regions with less (worse) significance
than normalizing
the functional images.
Its hard to determine when one normalization is better than another. What
we do as a quality control step is after the mri's have been normalized
we check all of them using check reg to make sure all of the image
features line up. You
should be happy at this point before you move on to the statistics section,
or you should go back and repeat the normalization process. There are
a lot of parameters that can be set in the normalization process; you
should try to
become familiar them, and choose settings that work well with your data.
david wack
university at buffalo
|