> I have a group of AD patients that have had medication for a year. I want
> to compare PET measurements before and after treatment. I have done
> summations from each PET-scan occasion. I now have 2 image for each
> patients (one before treatment and one after) When I normalized the
> pictures (in SPM99) I took one image at a time and used the same one to
> "determine parameters" and "to write normalized". Is this correct? I
> consulted a fellow worker and she said I need to do a mean image of the to
> images from each patient and then normalize the two images to that mean
> image (do you follow?). Do I need to do this? And if so, how do I do this?
> Can I do it with SPM?
I agree with your collegue. It is better to rigidly register the images
together, thus allowing the same nonlinear warps to be applied to both. The
mean of the two images should have a slightly better signal/noise ratio,
giving better spatial normalisation results.
You can use the realign button to do this, as it gives the option of creating
a mean of the realigned images (best if you download the updates first). This
mean can be used to estimate the warps (via spatial normalisation), which can
then be applied to your individual PET images.
>
> A last question. What statistical method would you use to compare the data
> (before and after treatment). I think that a paired t-test would work?! My
> results is hard to interpret, that's why I am asking. But I don't know if
> that is because of the statistical method or something with the pictures
> (i.e. the normalization).
A paired t-test should work fine. Be careful about how the images are scaled
relative to each other. I'm not sure if the images represent some meaningful
units (such as CBF, volume of distribution, or some rate constants, or function
of the rate constants). If they do, then you should be able to just use the
images as they are. If not, then you may need to think about what is the best
form of intensity normalisation.
If the scans are taken a variable time apar, then instead of using a simple
paired t-test design matrix based of (e.g.) ones, zeros and minus ones, you
may instead want to try including a function of the time interval in the design
matrix.
Best regards,
-John
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|