Source anaylsis

Juan Civeira jciveira at UNAV.ES
Thu Jul 27 18:54:27 CEST 2006


   Dear Robert:

   Thanks for your answer but my problems are still there. The main
problem is that i can not use DICS with a reference channel on EEG
data, no matter if  the reference is an EEG channel or EMG or some
other channel.

   I have taken my measures with 85 electrodes, some of them are
EEG1010, some EOG and some other that I have placed myself.

   I do the freq analysis and then the source analysis. Then I can
make another source analysis with refdip but there is always errors
when I use refchan. I do not think is a problem of the reference ,
cos always the program says that the  cross-spectral-density with
the reference channel is not complete .

   I have done the freq analysis making cfg.channelcmb combinations of
all channels with only one, so i make sure that the
cross-spectral-density is complete,but when I make source analysis
with that channel as a reference is the same.

   Searching in the code I have come up with the fact that the error
is always the condition in line 195 of the function
prepare_freq_matrices (version 2006/03/08). My idea is that there is
a confussion because if we have a set of N channels and we make  
combinations with channelcombination, the result is N-1 combinations  
because we do not include the combination of the reference with itself.

I hope i explain myself clear enough, anyway I am trying to fix the  
problem by myself but any help or suggestion will be great

Juan


Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
>> I have the same problem and I do not know how to fix it, even   
>> though I tried your solution with cfg.channelcmb.
>>
>> My case is : i have a set of 82 electrodes, I make freqanalysis   
>> with cfg.channelcmb = channelcombination({'all'
'all'},data.label);
>>  thus I get
>>
>> freq =
>>
>>        label: {82x1 cell}
>>       dimord: 'chan_freq'
>>         freq: [14 14.4000 14.8000 15.2000 15.6000 16
16.4000
>> 16.8000 17.2000 17.6000 18]
>>    powspctrm: [82x11 double]
>>     labelcmb: {3321x2 cell}
>>    crsspctrm: [3321x11 double]
>>          cfg: [1x1 struct]
>>
>> When I make sourceanalysis with a dipole as a reference and I
have
>> no problems but when I want to use a channel as a reference I   
>> always have the same error:
>> The cross-spectral-density matrix is not complete.
>>
>> Can you help me?
>>
>> Best regards
>> Juan
>
> Dear Juan
>
> Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
> expect the latter, since computing coherence between one particular
EEG
> electrode and sources in the brain with DICS will not lead to
> meaningfull results. The selected EEG electrode will pick up
> volume-conducted activity from all sources in the brain (as opposed
to
> an EMG electrode) and would be referenced to the same reference
> electrode as all other electrodes. So there will be a huge amount
of
> trivial coherence, whose spatial distribution in the brain will
> probably more reflect the volume-conduction than something
> physiologically interpretable.
>
> Although I have not used DICS with a reference channel on EEG data,
I
> would expect the code to work if you follow a procedure like this:
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EEG electrodes
> cfg.implicitref = name of implicit reference electrode used for the
EEG
> channels
> cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
> data_eeg = preprocessing(cfg)
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EMG electrodes, assuming that they are
> bipolar channels
> % cfg.reref may be used if you have two unipolar channels on the
muscle
> of interest
> cfg.rectify = 'yes'   % this makes the spiky nature of the EMG
more
> visible, and improves the coherence estimate
> cfg.baseline = 'yes'  % remove baseline prior to rectification,
not
> neccessary but still preferable
> data_emg = preprocessing(cfg)
>
> % combine EEG and EMG in one dataset, both are referenced
differently
> data = appenddata([], data_eeg, data_emg);
>
> cfg = ... frequency analysis stuff
> cfg.ouptu = 'powandcsd'
> cfg.channelcmb = {'all', 'all'}
> freq = freqanalysis(cfg, data);
>
> cfg = ... source specific settings
> cfg.refchan = one of your bipolar EMG channels
> cfg.channel = all your EEG channels
> source = sourceanalysis(cfg, freq)
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es



More information about the fieldtrip mailing list