Dear List:
I have processed data using SPM2 for 13 subjects using global
normalization and have since become aware that this is not such a
good idea. Nevertheless, I have spotted an anamoly that I thought
the list might be interested in and able to cast some light on.
Our design involved two tasks that were contrasted with each other
or with rest. About 25-50% of the time, the activation patterns were
massive, involving almost the entire brain and often causing SPM to
crash when trying to generate the images (leaving “Segmentation
Fault” in the command window, I believe it is a stacking issue?).
This problem went away when I omitted the global normalization in
subsequent analyses. I have a few questions:
Why does global normalization seem to make such a big difference for
some, but not all subjects?
1. This problem with whole brain activation that I believe is
associated with global normalization seems to be more prevalent in
one task versus the other. The task where I ran into this problem
more frequently was the first task they performed while in the
scanner. Might that have something to do with it?
2. This data was collected on a GE 1.5T scanner using reverse
spiral. I also collect fMRI data using the same tasks on a Siemens
3T scanner collecting EPI images. With these data I have never run
into this trouble even though I processed my data with global
normalization (which I will go back and adjust given my new
enlightenment). Does anyone have any suggestions for why this might
be the case? We believe it could be due to the spirals having quite
prominent bright spots due to point spread functions and other
mathematical blips, but we are not sure.
Many thanks in advance for your help.
Sent by Stuart Derbyshire for Patrick Fisher.
Sent by Medscape Mail: Free Portable E-mail for Professionals on the Move
http://www.medscape.com
|