Stephen,
I'm not sure there's any consensus about motion correction but I
would keep in mind that the degree to which the motion parameters are
correlated with the task/stimulus paradigm is equally, perhaps even
more, important than the amplitude of the motion. In a phantom study
(Field et al. AJNR Am J Neuroradiol 21:1388-1396, Sep 2000) we found
false activations resulting from very small movements (<1mm/1deg) when
the paradigm-motion correlation was >0.5 or so. Unfortunately, the
higher the correlation, the more dicey it becomes to include the motion
parameters as covariates. Therefore, we check the motion-paradigm
correlation and discard the data if r>0.5 regardless of the motion
amplitude.
Regards,
Aaron S. Field, MD, PhD
Department of Radiology
Wake Forest University School of Medicine
Winston-Salem, NC
> Subject: motion correction rules of thumb?
> Date: Mon, 5 Feb 2001 09:46:35 -0500
> From: Stephen Fromm <[log in to unmask]>
> To: [log in to unmask]
>
> Is there any (rough) consensus in the community about motion correction
> rules of thumb? Something like
> "if translation along any axis is > a, or rotation along any axis is > b,
> then include the movement parameters as confounding covariates. And if
> these numbers are > a' or b', respectively, then the subject data should be
> discarded."
>
> Best wishes,
>
> Stephen Fromm, PhD
> NIDCD/NIH
|