Working with white matter is a bit of an unknown quantity. With VBM,
most of the differences are found at edges. Because the cortical
sheet is generally thinner than the amount of smoothing used, the
partial volume means that an SPM analysis of the data gives reasonable
localisation of where there are volumetric differences.
With WM, there are large continuous areas of solid tissue. If tissue
is lost in some region, it is difficult to say exactly where it
happens. Therefore, it is probably more likely to be localised around
the ventricles etc., even if the tissue was actually lost elsewhere
from the WM. Scaling by the Jacobians helps a bit, although the
spatial distribution of the expansion and contraction estimated by the
warping will depend on the details of the registration algorithm. See
figures 4 and 5 of
http://www.frontiersin.org/Brain_Imaging_Methods/10.3389/fnins.2012.00197/full#h4
to get an idea of how estimates of where volume changes occur can
vary, depending on the sort of regularisation used.
Best regards,
-John
On 18 February 2013 20:27, Jie Yu <[log in to unmask]> wrote:
> Hi
> In the multiple regression basic model design after WM VBM,
> there is a MASKING option.
> I set " absolute threshold" with threshold 0.2.
> Is that ok?
>
> This is the GM SPM8 manual suggested setting.
> But I am doing a WM now.....so really don't know if that's ok?
>
> THANKS
|