Dear SPMers,
I've been noticing some fairly extensive ring-like
white-matter deactivations in a fast-rate event-related fMRI paradigm. In the
study, there are 7 trial types, which are presented in a randomized
fashion. So far, I have 15 subjects and am analyzing the data with
random effects analyses on the con* images.
When I contrast events in which a target stimulus appears to
events in which no target appears, I find activations in expected areas
such as the visual and motor cortices. However, ring-like deactivations
in the whitematter (and deactivations in other areas as well) occur when
I contrast non-target events to target events.
The extensive ring-like deactivations in the white matter for
non-target events led me to test whether they might be an artifact of
global proportional scaling which I had performed during estimation of
the general linear model for each individual subject. As others have
noted, when a particular volume contains highly activated brain regions,
global proportional scaling reduces activation in all brain areas within
that volume to make the volume's mean activation equal to that of all
other volumes in the series. Unfortunately, this procedure can sometimes
render inactive regions within a highly active volume (e.g., the white
matter) deactivated.
Thinking along these lines, I re-estimated the model for each
individual subject, this time selecting no global scaling during the
procedure. However, the ring-like deactivations remain.
If I understand them correctly, recent discussions on the SPM list have
stated that SPM scales all volumes to a mean of 100 by default and
that selecting global scaling during model estimation makes the
scaling proportional. Is this true? If so, then would anyone be able to
advise me on how to turn off global scaling entirely? Alternatively,
would anyone have any other ideas about what might be causing the anomalous
rings of deactivation I am observing?
Any advice would be greatly appreciated.
Thanks and Cheers!,
Daniel Weissman, PhD
Center for Cognitive Neuroscience
Duke University
Durham, NC 27708
|