[Eeglablist] Using ICA with EEG, ECG and EOG.

[Alois Schloegl]

Hi Dario,


if you want to do a fully automated EOG/ECG correction, you can use 
REGRESS_EOG.M from biosig/t250 (I guess it comes with eeglab). In my 
experience, regression will work better than fully automated ICA (for 
details see [1]).

It works this way:
   [s,HDR]=sload(filename);
   [M,N]=size(s);

   eocg = sparse(N,3);
   eocg(#H+,1)=1; 	% replace #H with the number of your H++ channel)
   eocg(#H-,1)=-1; %
   eocg(#V+,2)=1;
   eocg(#V-,2)=-1;
   eocg(#EKG+,3)=1;
   eocg(#EKG-,3)=-1;

A fully automated identification of these channels is needed to automate 
that step. The function identify_eog_channels() is trying to do this, 
but because of a lack of standardization it's likely it won't work 
correctly. Though, You might want to adapt it to your data.

   eegchan = [... ]; % list of your EEG channels
   R = regress_eog(s, eegchan, eocg);
   s_corr = s*R.s0; 	% corrected eeg channels


A view remarks:

- You can expect the best performance when R = regress_eog(..) is 
computed from a data segment with large EOG and EKG artifacts (ask the 
subject to perform different eye movements at the beginning of the 
recording).

- ECG has 3 spatial components, but only one is removed. Therefore, you 
can expect only a limited performance on the ECG artifact.

- if the function identify_eog_channels() works properly on your data, 
the method can be fully automated. See also
    help regress_eog
    help identify_eog_channels
    help get_regress_eog
    help sload

    Alois

[1] 
http://precedings.nature.com/documents/3446/version/1/files/npre20093446-1.pdf



On 05/17/2012 05:57 PM, Dario Madeo wrote:
> Hi all,
>
> I'm a new user in the fantastic EEGLAB world :) !
> I have a question about the removal of ECG and EOG artifacts from EEG data.
>
> Here is my scenario: I have 40 channels, where:
> - 34 are EEG of which I know the location on scalp
> - 4 are EOG (namely H++, H-, V++, V-)
> - 2 are EKG (namely EKG++, EKG-).
>
> I load the dataset, and after this operation I would like to immediately
> run an ICA decomposition
> (I need to do something else before this?).
> But actually I really don't now how to correctly use this dataset with the
> ICA decomposition.
> Should I run the ICA only on the 34 eeg channels, or maybe on all 40
> channels? (in the
> relative GUI i can decide to do ICA on a certain subset of channels).
> The aim is to exclude the influence of the ECG/EOG artifacts from EEG data
> to accomplish further analysis.
>
> Can anyone help me?
>
> Thanks a lot!
>
> D.M.
>
>
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu