Dear Trine,
How much do you know about your data? For example, how reliable is your BP measurement in test-retest analyses?
Some very brave people have used unscaled absolute measures of receptor binding, and that's historically the norm in ROI analyses. In our experience, however, all kinds of uncertainties (cross-calibrations, internal diameters of tubing, etc. etc.) tend to add up to make globally scaled receptor concentration estimates the much more sensitive option at least for SPM, even for the tracers we've used which achieve Intraclass Correlation Coefficients for test-retest data of 0.85 or so.
As epileptologists, we've been lucky enough to have some pretty robust a priori ideas of what to expect where, and I think the way we've conducted analyses over the years has stood the test of time and has good face validity. I personally find AnCova preferable over proportional scaling as the latter tends to amplify the noise in images with lower parameter values.
For some recent pointers, try Hammers A, Asselin MC et al. Neuroimage 2007 and Hammers A et al. Brain 2007 - sorry haven't been searching for long but have given work from our lab!
Hope this helps,
Alexander
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Trine Hjørnevik
Sent: 26 November 2008 11:04
To: [log in to unmask]
Subject: [SPM] PET receptor analysis
Dear SPM users
I am working on a PET receptor study where I have 8 subjects with 2 scans
(baseline/activation) each. I have calculated binding potentials (BP) in
PMOD, and now I want to compare my two conditions in SPM. I am using paired
t-test, but alternating with the different options give me quite different
results. I was wondering if anyone in here have some experience in analysing
similar data? Should I use global normalisation, and if so overall grand
mean scaling or normalisation with proportional or AnCova?
Thank you for your help!
Best regards,
Trine Hjørnevik
|