[FieldTrip] inverse imaging

[David Pascucci]

Thanks again Julian,
About the covariance, I am not sure about its usage in the reconstruction
of single-trials activity.
According to the example, this is done by multiplying the spatial filters
with the EEG data.

Whereas the covariance (Cf) is used to compute the avg.pow and ori in
ft_eloreta (line 160-168)

% get the power
dip.pow = zeros(size(dip.pos,1),1);
dip.ori = cell(size(dip.pos,1),1);
for i=1:size(dip.pos,1)
  csd        = dip.filter{i}**Cf**dip.filter{i}';
  [u,s,vv]    = svd(real(csd));
  dip.pow(i) = s(1);
  dip.ori{i} = u(:,1);
end

It does not seem to be considered when creating and storing spatial filters
(later used for single-trials reconstruction)
(line 152-158, ft_eloreta)

% use existing filters, or compute them
if ~isfield(dip, 'filter')
  filt = mkfilt_eloreta_v2(leadfield, lambda);
  for i=1:size(dip.pos,1)
    dip.filter{i,1} = squeeze(filt(:,i,:))';
  end
end

My question is, am I getting this wrong?
and if not, should I ignore the covariance estimation in the case of
single-trials reconstructed via filters*data?


Cheers,
David


2017-09-13 13:41 GMT+02:00 Julian Keil <julian.keil at gmail.com>:

> Hi David,
>
> regarding the lambda, I think there are different ideas floating around
> the fieldtrip discussion-list. I suggest searching for the term „lambda“ to
> get a rough idea. Personally, for our EEG-data I usually use 10%.
>
> What is your question exactly regarding the covariance as input?
>
> Cheers,
>
> Julian
>
> Am 13.09.2017 um 13:22 schrieb David Pascucci <psc.dav at gmail.com>:
>
> Thaks Julian,
> that is the approach I was using, with eLoreta.
> I am not sure about two steps,though.
> One is the estimate and use of the signal covariance to input for single-trial
> activity in source space.
> The other is the choice of the optimal lambda.
>
> If you have some advice, that wold be very helpful.
>
> Thanks,
> David
>
> 2017-09-13 12:22 GMT+02:00 Julian Keil <julian.keil at gmail.com>:
>
>> Hi David,
>>
>> do you want to obtain single-trial activity in source space? In that
>> case, have you looked at the „virtual sensors“-tutorial? http://www.
>> fieldtriptoolbox.org/tutorial/shared/virtual_sensors
>> In the tutorial, LCMV is used for the source analysis, but it should also
>> work with sloreta, as the output-structure of the source-analysis is
>> identical. I’m not sure about MNE though.
>>
>> Good luck,
>>
>> Julian
>>
>>
>> Am 12.09.2017 um 20:47 schrieb David Pascucci <psc.dav at gmail.com>:
>>
>> Dear fieldtrip experts,
>>
>> I was wondering if anyone has experience with extracting single trials
>> estimates of source activity (using MNE or Loreta-based approaches) from
>> regions of interest, and what would be the best procedure…
>>
>>
>> Thanks in advance
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl
>> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>
>>
>>
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl
>> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>
>
>
>
> --
> ---------------------
> David Pascucci
>
> Postdoctoral Fellow
> University of Fribourg
> Department of Psychology
> Rue de Faucigny 2
> 1700 Fribourg
> Switzerland
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>



-- 
---------------------
David Pascucci

Postdoctoral Fellow
University of Fribourg
Department of Psychology
Rue de Faucigny 2
1700 Fribourg
Switzerland
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20170913/f48904ca/attachment-0002.html>