[FieldTrip] beamformer results smaller than brain

[Akiko Ikkai]

Thanks, Arjen and Gio,

I played around with cfg.inwardshift during ft_sourceanalysis, and
adjusting it to somewhere between .25 and .5 gives decent results (it was
originally set to 1), i.e. good coverage with no shrinkage.

I'm actually a little confused about why one might need cfg.inwardshift
parameter... Assuming the segmentation into different tissue types goes
correctly, is it best to set the inwardshift parameters as close to 0 as
possible (or not specify), since we'd want to put dipoles ranging the whole
volume of the brain?

Thanks! Akiko

On Thu, Mar 29, 2012 at 5:13 AM, Gio Piantoni <g.piantoni at nin.knaw.nl>wrote:

> Hi Akiko,
>
> Your headmodel looks pretty good and your beamformer plot seems to
> cover most of the gray matter.
>
> However, it's good to check the location of your dipoles in respect to
> your headmodel. For example, you can plot the .pos field of your
> leadfield. Something like:
>
> ft_plot_mesh(bnd(3), 'facealpha', .5)
> hold on
> plot3(lead.pos(:,1), lead.pos(:,2), lead.pos(:,3), '.')
>
> It might well be that your grid is too coarse and some dipoles just
> happen to be over the edge of the brain. Remember that your
> "beamformer_result.png" is just an interpolation of the values of this
> dipole grid.
>
> You can either try to make your grid more refined (but remember that
> your computation time will significantly increase). Or you can
> manipulate the way dipoles are considered inside or outside the brain
> with "inwardshift" for ft_prepare_leadfield (although I'd advise
> against using a negative value as some dipoles might really end up
> outside of your brain mesh, with some numerical instabilities).
> Another option (the easiest) is to nudge your grid by, say, a few
> millimeters, so that the parietal dipoles will be just inside the
> brain mesh.
>
> HTH,
>
> Gio
>
> --
> Giovanni Piantoni, MSc
> Dept. Sleep & Cognition
> Netherlands Institute for Neuroscience
> Meibergdreef 47
> 1105 BA Amsterdam (NL)
>
> +31 20 5665492
> gio at gpiantoni.com
> www.gpiantoni.com
>
> On Thu, Mar 29, 2012 at 00:22, Akiko Ikkai <akiko.ikkai at gmail.com> wrote:
> > Hi Fieldtrip users,
> >
> > I'm hoping that someone could give me advice on segmentation and
> beamformer
> > on EEG data. I have EEG data set based on 128 channel cap. Thanks to the
> > help I got in mid-Feb, I'm now able to create a decent
> > segmentation (seg_results image attached) and volume conduction model.
> > However, when I run beamformer based on these models, resulting
> beamforming
> > image is often smaller than the brain (beamformer_result attached).
> > Particularly posterior parietal and occipital map is NaN.
> >
> > I have tried going back to segmentation and expanded brain by
> using imdilate
> > after ft_volumesegment such as:
> >
> > newbrain = imdilate(seg2.brain,strel_bol(1)); % seg2.brain is original
> brain
> > tissue from segmentation
> >
> > seg2.brain = newbrain;
> >
> > seg2.seg = seg2.scalp + seg2.skull*3 + seg2.brain*6;
> >
> > and run ft_prepare_mesh_new on seg2. Of course, I have to make sure
> there is
> > no intersect between different tissue types, so using imdilate has
> > limitation.
> >
> > Could someone explain why this shrinkage might be happening, and how I
> could
> > fix it?
> >
> > Thanks in advance! Akiko
> >
> > --
> > Akiko Ikkai, Ph.D.
> > Postdoctoral Fellow
> > Department of Psychological and Brain Sciences
> > Johns Hopkins University
> > Ames Hall, 3400 N. Charles St.
> > Baltimore, MD 21218
> >
> >
> >
> > _______________________________________________
> > fieldtrip mailing list
> > fieldtrip at donders.ru.nl
> > http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>



-- 
Akiko Ikkai, Ph.D.
Postdoctoral Fellow
Department of Psychological and Brain Sciences
Johns Hopkins University
Ames Hall, 3400 N. Charles St.
Baltimore, MD 21218
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20120402/d498cf2b/attachment-0001.html>