Hi Debbie,
On Thu, Aug 12, 2010 at 12:38 PM, dtalmi <[log in to unmask]> wrote:
> hi there,
>
> two questions on source localization.
>
> i tried to localize the source of the activations i saw at sensor level. the
> literature speaks about striatum and acc as the likely sourrces which i
> realize are not going to show in MEG. i used MSP and got bilateral insula
> (as well as lots of visual cortex). i was wondering if it is at all feasible
> with this technique that one deep source mascarades as two bilateral
> sources?
I've seen more examples of the other way around. It sometimes happens
(not necessarily in MSP) that a single central source shows up instead
of symmetric bilateral sources. Since a model with a single source is
a simpler one it will usually be preferred. The only reason I can
think of why it wouldn't happen is that the right location is not
included in the mesh but ACC is definitely included. As to striatum,
it's not in the mesh so theoretically it could happen that a signal
from striatum maps to the insula but there is no convincing evidence
of signals from striatum in the MEG and even if there are some, they
will only be visible with a large number (thousands) of trials and
would be probably very difficult to localize with MSP in the presence
of much stronger cortical signals. The only way that might perhaps
work in this case is 'localization of contrast' rather than 'contrast
of localization'. So what I would do if I were you is look at the
corresponding sensor-level contrast and:
1) Try to see by eye what is the scalp pattern that contributes to
those insular activations and whether it looks like two symmetric
dipoles or a single central dipole. You can compute the contrast for
each subject and grand-average to get a clearer picture. If you send
me some scalp maps I can give you my opinion.
2) This pattern can also be examined with VB-ECD and you can do model
comparison of a single source vs. two sources.
>
> the contrast i'm most interested in is quite complex (a 2*4 interaction),
> and i thought of perhaps trying to only compare 2 conditions. but i'm always
> working with a file that has all 8 conditions; how does one pluck out just
> two? is it possible to have a source-localization for just one single
> condition or does it always have to be a contrast?
>
When you do the inversion you'll be asked what conditions to include
and you can only select two. This can also be done via the batch tool.
You can localize a single condition but the problem is doing the stats
on it as the data is non-negative. There have been some discussions
about it on and off the list before and I started with suggesting
comparing two time windows:activation and baseline. Then from
experience we saw that it doesn't work very well so later I tried to
discourage it. In the latest SPM version one can define several time
windows together when doing a contrast which might result in more
sensible activation to baseline comparisons. So the best way is to
try.
Vladimir
>
|