CTF275 coherence

[jan-mathijs schoffelen]

Dear Cristiano,

Very quickly: the gradiometer description is present in the data-
structure as data.grad
(and will be passed on to freq.grad when calling freqanalysis).
This at least should be there when you read in your data using
preprocessing. You do
not have to specify cfg.grad before your call to sourceanalysis,
because it will take the
gradiometer structure from the data. This will ensure that the csd
and gradiometer
description really match. As long as you use your own grad-structure
you have to be
absolutely sure that this one corresponds to the specifications of
the dataset at hand.
I cannot judge that this really is the case, so I would drop the
cfg.grad from your configuration
to sourceanalysis in the first place.

I totally agree with you that a 'good emg-meg correlation is not only
the peak in the spectrum
but also two controlateral dipolar 'spots' in the topography', and I
did not claim otherwise. Only:
the sensor-level csd is much richer in structure than the part you
plot on the topography and a
good topography is no guarantee for a good source result (please take
my word on it: I have
looked at these things a couple of times myself, so I should know ;o) ).

The number of trials is sufficient I guess.

Yours,

JM



On Jun 27, 2008, at 4:11 PM, Cristiano Micheli wrote:

> Dear Jan-Mathijs
> In the preprocessing i performed baseline correction for meg data
> and 10 Hz
> hi-pass filtering for emg (sampling at ~300 Hz).
> However since normally the correlated activities for isometric
> pinch take
> place in the beta band, and eye-blinks and other artifacts are in
> other
> bands or uncorrelated with the emg, i did not think about applying
> artifacts
> removal.
> For sure i will have a look at it but what for me constitutes the
> signal of
> a good emg-meg correlation is not only the peak in the spectrum but
> also two
> controlateral dipolar 'spots' in the topography, which i took as a
> good
> result. And i expect dics to be more sensible than emg-meg
> coherence in
> detecting sources, with the right settings.
> I specify the settings of frequency analysis like that:
>     refch    =  'EMG_lH';
>     cfg       = [];
>     cfg.output     = 'powandcsd';
>     cfg.method     = 'mtmfft';
>     cfg.foilim     = [max_f max_f]; % peak of coherence spectrum
>     cfg.tapsmofrq  = 5;
>     cfg.keeptrials = 'yes';
>     cfg.channel    = {'MEG' refch};
>     cfg.channelcmb = {'MEG' 'MEG';'MEG' refch};
>     freqcond       = freqanalysis(cfg,data);
>     cfg = [];
>     fd             = freqdescriptives(cfg,freqcond);
>
> Maybe i should use another method for csd calculations or another
> smoothing
> factor.
> Alltogether i have 250 subtrials and 9 tapers. I collect 25 trials
> of 10
> seconds each and then i cut each trial in segments of 1 second each.
> The reason why i use my own gradiometers definitions is because i
> already
> have this information ready in my code from a CTF routine. Also i
> did not
> get how to look for it in the documentation. How can i do it easier?
> Then I run the dics with the following parameters:
>     cfg           = [];
>     cfg.grad     = grads;
>     cfg.grid      = grids_;
>     cfg.method    = 'dics';
>     cfg.projectnoise = 'yes';
>     cfg.lambda    = 0;
>     cfg.refchan   = refch;
>     cfg.frequency = max_f;
>     cfg.hdmfile   = 'myhead.hdm';
>     sourcecond    = sourceanalysis(cfg, freqcond);
>
> In the simulations i get the same problem whereas the spectra look
> reasonable and the dipole fit localizes the source in the right
> place from
> coherence peak.
> The simulations were done with one source (virtual emg) coherent in
> phase
> and amplitude with the source time course. Source is modelled by a
> dipole in
> position [0  5  10] cm (head coordinates) and moment [10 0 0] nA.m
> How do i perform a good frequency analysis in order to have a good
> localization?
> If it does not depend on fr. analysis which other factors can cause
> this
> problem?
> Thank you for the patience
>
> Best
> Cristiano
>
> ----------------------------------
> 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.

----------------------------------
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.