Hi Lucas

> at the moment I am unsure about whether correcting for total brain volume (TIV) in
> a VBM analysis of unmodulated gray matter segments (density) does make sense or not.
>
> On Dr. Gaser's homepage --- I have the feeling --- there are two slightly opposite statements:
> statement (a) (please refer below) clearly states that TIV should not be used as a nuisacne factor with unmodulated data.
>
> On the other hand, I read statement (b) (please see below) as correcting unmodulated data for TIV gives me relative
> density after correcting for TIV, which seems to make more sense to me.
>
> Which procedure is the more valid one, or for which kind of question would I use which procedure respectively?


One important issue that you bring up is the use of modulated gray matter (GM) images and what an analysis using these images would tell you. I think the prevailing view on this is as follows: if normalization were perfect, and there was no modulation, then all GM images would be perfectly aligned. Because we haven't modulated to account for changes in size, the images will be identical. Thus, any differences we see in an unmodulated analysis are likely due to errors in registration. So the first issue, i think, is to make sure that unmodulated images are answering the question you want to answer.

I think that the most common view of TIV and unmodulated analyses is that unmodulated GM images represent a property of the tissue, which (unlike its volume) should not depend on head size. Thus, the advice you came across not to include TIV in an unmodulated analysis.

[One thing I wonder about the above interpretation of unmodulated analyses, though, is the application to patients with various types of neurodegenerative disease. If there are small pockets of damage (e.g. spongiform encephalopathies), then I would think that an unmodulated analysis might pick this up. Which is to say, I think there may be things other than registration differences that may lead to differences in unmodulated analyses. However, I would be very interested to hear from someone who has had some experience with this.]

So, as to your specific question, if you don't have a reason to think that what you are looking at (e.g. tissue density) would be impacted by head size, then I wouldn't think there is a good reason to include TIV. (Although you may want to do some exploratory analyses, preferably in another group of subjects, and include it—then you could see empirically what variance in density, if any, was accounted for by TIV, which might be informative for you, and others.) :)

Best regards,
Jonathan




--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/