Dear Lukas,
> I wonder, if any preprocessing (as linear trend correction or smoothing) should
> be performed prior to motion correction of an fmri run.
Though I never really tried it the other way around, I'll try to answer
your question in SPM terms...
In SPM, preprocessing of fMRI data (conditions measured in epochs) means
usually realignment (motion correction), normalization (optional) and
spatial smoothing of the functional data.
The linear trend correction can be described in SPM terms as part of the
high pass filter which is usually applied to the data during the
statistics part. The high-pass filter removes low-frequency confounds
('low' in relation to the known paradigm frequency). The reason, why
this high-pass filter is not applied to the data prior to realignment is
that it is the time series induced by a potential underlying neuronal
signal generator located in a fixed brain location which one wants to
analyze.
Smoothing prior to realignment:
I guess that the result of the realignment would be a function of the
width of the smoothing filter applied. Intuitively, I would assume that
one should give maximum information about the data to the realignment
routine. Since smoothing means loss of information, better do not
smooth. (However, if any degree of smoothing is good for the realignment
of the data, this decision should be left to the realignment routine....
)
Best wishes, Stefan
--
Stefan J. Kiebel
Dept. of Neurology
Friedrich Schiller University
Philosophenweg 3
07740 JENA
Germany
Tel. : +49 3641/9-35285
FAX : -35399
email: [log in to unmask]
WWW : www.neuro.uni-jena.de
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|