Kent,
We have observed such a feature of our data and have
often reported it in papers in the Methods section to justify
filtering of the Nyquist frequency. Your practice of including an
alternating covariate in your analyses is equivalent to this filtering.
(You might be interested to hear that it would not matter if you
did convolve this covariate with the hrf.)
Sincerely,
Eric
>
> Dear SPMers,
> We have recently run into an interesting 'feature' of our MR scanner
> that produces a small, but significant artifact in fMRI time series. We
> are curious to know if anyone else has observed this artifact or has
> come up with other ways to deal with it other than that listed below.
> It is my understanding that, by default (at least on GE scanners), rf is
> applied in an alternating fashion in EPI single shot acquisitions. This
> is known as rf-chop on our machine and it is typically used to increase
> SNR when combining acquisitions as in typical MR applications. However,
> we have observed that this feature produces stronger signal when rf is
> applied in one direction than when it is applied in the other direction.
> Moreover, these effects seem to be vary in different slices within
> acquisitions. Thus, there is a large amount of image intensity variance
> between acquisitions. The tell tale sign of this artifact is the
> presence of a small 'saw tooth' like (about .2 mm) movement estimate in
> the realignment output.
> We have recently been able to compensate for this feature by simply
> adding a covariate of no interest or 1 -1 1 -1 1 -1......etc for each
> image in the time series during statistical analyses. Of course, we do
> not convolve this with covariate of no interest with the hrf. This
> effectively removes the variance associated with this artifact from our
> data. Indeed, z-scores typically jump from the 5 to 5.5 range to over
> 6-7.0 range when controlling for this artifact.
>
> I would very much appreciate any comments that any other users,
> physicists in particular, might have on this issue.
>
> I discussed this issue with members from several other fMRI labs that
> have observed this artifact in their realignment time series at the
> recent Cognitive Neuroscience meeting in Washington. They had not been
> sure what to make of it. I hope that this note helps others as well.
>
> Best regards,
> Kent Kiehl
>
>
> :::::::::::::::::::::::::::::::::::::::::::
> Kent A. Kiehl
> 2255 Wesbrook Mall
> Departments of Psychiatry and Psychology
> University of British Columbia
> Vancouver, B.C., Canada V6T 2A1
> work (604)822-7128
> home (604)221-7809
> fax (604)822-7756
> email [log in to unmask]
> http://www.psychiatry.ubc.ca/sz/nilab/kiehl
> ::::::::::::::::::::::::::::::::::::::::::::
>
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|