problems with DICS

Wed Nov 19 15:07:31 CET 2008

hi jan-mathijs,

thanks once again for your quick help.

it's about 60 trials which was sufficient in the past. the decrease is
bilateral, however i can't imagine how this could lead to a more or
less pronounced increase after sourceanalysis. in case of a DICS
problem wouldn't the decrease just be placed somewhere in between the
two spots?

the labels in data.label, data.grad.label match. they also have the
same order as in the .xyz file from the BTI system.

for the sourceanalysis we precompute the leadfield. as a side note we
also use predefined grid points so that they are exactly the same
within one subject across conditions.
one issue that we noticed is that we recalculated a leadfield that we
calculated a long time ago and compared the values. they appear not to
be exacly the same. when looking at the topographies (forward
solution) yieldied by each source, they are similar but not exactly
the same. is this possible?
I'm attaching a plot. for each dipole i calculated the vector sum over
orientations and averaged the energy over sensors. this i did for the
old and the newly calculated leadfield. in the plot difference between
the two leadfields is shown on y and dipole number on x.

here's the code:
%% checke tzvetans grid mit neuer FT

cd /home/nadia/Matlab/zvetanskopf

tvol=load('vol'); %old vol
tgrid=load('grid'); %old grid

%%
cfg=[];
cfg.grid.pos=tgrid.grid.pos;
cfg.grid.dim=tgrid.grid.dim;
cfg.grid.inside=tgrid.grid.inside;
cfg.grid.outside=tgrid.grid.outside;
cfg.vol=tvol.vol;
cfg.grad=tgrid.grid.cfg.grad;
cfg.channel=tgrid.grid.cfg.channel;

tgrid.grid2=prepare_leadfield(cfg); %new grid

%%

for i=1:length(tgrid.grid2.inside)
diffLF
(i)=mean(sqrt(sum(tgrid.grid.leadfield{tgrid.grid2.inside(i)}.^2,2)))-
mean(sqrt(sum(tgrid.grid2.leadfield{tgrid.grid2.inside(i)}.^2,2)));
end

plot(diffLF)

thanks for any input.

nathan

On 19.11.2008, at 14:13, jan-mathijs schoffelen wrote:

> Dear Nathan,
>
> Are you using precomputed leadfields? The reason I ask is because
> there could be a discripancy between the assumed order of the coils
> in your leadfields, and the coil-order in your data. The issue with
> the BTI system is that the channel order is somewhat erratic
> (references ending up all over the place, and no nice alphabetical
> ordering of the magnetometer coils). Prepare_leadfield (the low-
> level function which computes the leadfield) just computes the
> solution to the forward model for the list of sensors in the input,
> the order of which is specified by the order in the data, or by the
> order in the gradiometer structure, if no data is supplied.
> I recently (about a month ago) made a change in bti2grad, which
> changed the order of the coils in the grad-structure. Initially, I
> thought it would be nice to have them ordered alphabetically, but
> this led to problems later on when using precomputed leadfields (and
> making implicit assumptions about the matching sensor order in both
> data and gradiometers). It could be that your problems are related
> to this.
> On the other hand: could this be replicated in other datasets? How
> many trials is your csd-matrix based on? Isn't there any hint of a
> bilateral temporal decrease in alpha activity?
>
> Yours
>
> Jan-Mathijs
>
>
>
> On Nov 19, 2008, at 11:00 AM, Nathan Weisz wrote:
>
>> hi,
>>
>> i'm not sure whether the following question is a fieldtrip-related
>> question or a rather general question.
>>
>> we use a 148 sensor BTI system.
>> since a couple of days we're struggling with a data-set in which
>> we'd like to localize auditory cortical alpha desynchronizations.
>> on a sensor level they are clearly observable (see attached Pitcure
>> 2). however when running DICS it seems like the brain is more or
>> less increasing power post-stimulus.
>> what makes me wonder is that it's exactly the same script (i could
>> provide more details of course) which worked very successfully
>> previously. our notion was initially that something is weird with
>> the leadfield. but -illegally- testing it with a leadfield from
>> another subject where things worked out basically gave more or less
>> the same picture. now we assume that something is fishy with the
>> data itself, that is not clearly observable when looking at the
>> data on a trial-by-trial basis, leading to bad spatial filters. any
>> suggestions how this could be diagnosed?
>>
>> are there any other suggestions? or has anybody had similar
>> problems lately?
>>
>> cheers,
>> n
>>
>>
>>
>>
>>
>> <Picture 2.png>
>>
>> <Picture 1.png>
>>
>>
>> --------------------------------------------
>> Dr. Nathan Weisz
>>
>> OBOB-Lab
>> University of Konstanz
>> Department of Psychology
>> P.O. Box D23
>> 78457 Konstanz
>> Germany
>>
>> Tel: ++49 - (0)7531 - 88 45 84
>> Email: nathan.weisz at uni-konstanz.de
>> Homepage: http://www.uni-konstanz.de/obob
>>
>> "Nothing shocks me. I'm a scientist." (Indiana Jones)
>>
>> ----------------------------------
>>
>> The aim of this list is to facilitate the discussion between users
>> of the FieldTrip toolbox, to share experiences and to discuss new
>> ideas for MEG and EEG analysis.
>>
>> http://listserv.surfnet.nl/archives/fieldtrip.html
>>
>> http://www.ru.nl/fcdonders/fieldtrip/
>>
>

--------------------------------------------
Dr. Nathan Weisz

OBOB-Lab
University of Konstanz
Department of Psychology
P.O. Box D23
78457 Konstanz
Germany

Tel: ++49 - (0)7531 - 88 45 84
Email: nathan.weisz at uni-konstanz.de
Homepage: http://www.uni-konstanz.de/obob

"Nothing shocks me. I'm a scientist." (Indiana Jones)

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/fcdonders/fieldtrip.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20081119/1dc3d722/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Picture 1.png
Type: image/png
Size: 40889 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20081119/1dc3d722/attachment.png>