> When pre-processing structural data, I have found a VBM Manual that uses
> DARTEL and a VBM8 Manual that requires downloading the VBM8 toolbox. What is
> the difference between the two, if there is one at all, and are there
> certain conditions when I would use either one? Also, what is the difference
> between the VBM8 toolbox and other cnn toolboxes. I have provided links for
> clarification on what I am talking about.
VBM8 is a toolbox for conducting VBM analyses using SPM (including
DARTEL), but a slightly different approach than the built-in SPM
functions. There is probably not a clear reason to use one or the
other - they are both good approaches. It's worth reading the VBM8
webpage to understand some of the differences. I'm not aware of any
studies directly comparing the results.
> My second question deals with the implicit/explicit masking options given
> under the analysis. When would I include an explicit mask? I also notice the
> default is to include an implicit mask? How do either of these have an
> effect when I want to analyze the structural scans?
Masks will limit the voxels included in your analysis. Let's say that
you are only interested in left frontal cortex in your VBM study; if
you had a mask of this region, including it as an explicit mask would
be one way to focus your analysis here. Another common reason to use
an explicit mask is to restrict analyses to "gray matter" - although
using a threshold mask (as you mention) often does a very reasonable
job of this, so it's not really necessary (but some people prefer the
explicit mask approach).
> My last question is about the threshold masking. The default "specify value"
> option is 100 but the manuals I have looked at recommend either 0.1 or 0.2.
> Which is the best to use and why (I know it might depend on the
For an absolute threshold, the number will depend on the kind of data
being entered. For functional MRI data this will be different than
structural MRI data. The recommendations of something like 0.1 or 0.2
reflect that most values for structural MRI data are based on
probabilities, and so lie between 0-1 (although if images are
modulated, perhaps a bit greater than 1). There is no clear standard
here - the idea is to exclude things that are unlikely to be brain.
Empirically 0.1 and 0.2 work well for many people. You can try this
both ways, and compare the resulting mask.img - this will give you an
idea of the areas that differ, and what is being masked out. (For fun,
you could also try it with a threshold of 0.9, or with no threshold
masking, and then you can see the value of including this mask.)
Hope this helps!
Dr. Jonathan Peelle
Center for Cognitive Neuroscience and
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104