You raise an interesting question. You should of course wait to see what
the experts say, but I was thinking about the following:
In one sense the comparisons are between voxels in the same place within
subjects, and thus attenuation and scatter correction wouldn't be
necessary. Yet, as I understand SPM, it also looks for significant
differences (variance in a way across voxels and subjects) between subjects
to determine both voxel and cluster significance levels. Thus,
"artificial" differences due in fact to differences i head size (lower or
higher counts particularly in central structures merely as a consequence of
differences in head size - ie, attenuation) might create errors in the
variance estimation. Of course, if you also later wanted to look for
correlations between clusters, you might get false positives that were
merely due to differences in uncorrected attenuation.
-- Mike
Dr. Michael D. Devous, Sr.
Professor of Radiology
Nuclear Medicine Center
University of Texas Southwestern Medical Center
Dallas, TX 75235-9061
USA
phone 214 648-3315
fax 214 648-5641
[log in to unmask]
At 10:57 AM 6/15/00 , Dr Peter Talbot wrote:
>Dear SPMers
>
>The FAQ doesn't appear to have a clear consensus on this issue so I'd be
>grateful for advice:
>
>For our clinical SPECT scans we routinely do scatter- and
>attenuation-correction. However, if I want to use SPM to analyse SPECT
>scans in
>the following simple design is it really necessary to perform either
>correction?
>
>I have 16 subjects, each scanned on two occassions and each acting as his own
>control. The circumstances of each scan are as identical as possible, the
>only
>difference being a different pharmacological intervention prior to each
>scan. I'm looking for changes in rCBF between the interventions analysed by
>simple paired t test.
>
>Under these circumstances will scatter- and/or attenuation-correction do
>anything other than remove counts and potentially add noise? Does anyone have
>any references?
>
>Thanks for your time
>
>Peter
>----------------------
>Dr Peter Talbot
>Specialist Registrar/Senior Tutor
>Department of Mental Health
>Queen's University Belfast
>Whitla Medical Building
>Lisburn Road
>Belfast BT9 7BL,
>NORTHERN IRELAND, U.K.
>
>Phone: +44 (0)28 9027 2166
>Phone: +44 (0)28 9033 5791
>
>Fax: +44 (0)28 9032 4543
>Email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|