Dear SP,

The exclusion of bad channels from the reference channels can be done as illustrated below. As long as a channel is marked as 'bad', it is excluded from further analysis which means you don't have to delete it from the dataset. If you intend to extract sensor time-series for analysis outside SPM, simply use setdiff(D.meegchannels,D.badchannels) to index clean MEG/EEG channels.

I hope this helps

Martin

% re-reference (average)

% ---------------------------------------------------

 

if ~isempty(D.badchannels)

    G = setdiff(D.meegchannels,D.badchannels);

    R = G;

 

    T = eye(numel(D.meegchannels));

    T(G,R) = T(G,R) - 1/length(R);

 

    S = [];

    S.D = D.fname;

    S.montage.tra = T; % mxn transformation matrix

    S.montage.labelorg = D.chanlabels(D.meegchannels);

    S.montage.labelnew = D.chanlabels(D.meegchannels);

    S.keepothers = 'yes';

    S.updatehistory = 1;

    D = spm_eeg_montage(S);

end

On 22 Apr2013, at 12:17 PM, SP Ho wrote:

Dear experts

May I ask some basic questions about spm_eeg_montage ...

1. When we create the matrix for montage, how should we handle the bad channels if I would like to do average referencing ?  Should I exclude the bad channels (so that their effect will not be spread to those good channels when I do the average re-referencing) ?  As each subject / session may have a different set of bad channels, so I should create a different matrix for each subject ?

2.  I would also like to exclude some EEG channels in further processing.  I have created a script to do this and try a few things, but not sure if these are the correct ways:
  - Before montage, updated these channels as type 'Other', and put the flag keepothers to 'no' when running montage
  - Keep these channel in "labelorg" but remove in 'labelnew' in the montage matrix (becomes MxN), and put zeros to the columns (original labels) for these channels

And, in combining the effort to exclude both bad / other channel, I have generated the montage matrix so that the mean (1/N) is based on the no. of EEG channels that are marked as neither bad nor other ...

However, question is, after I generated the montage matrix and run the spm_eeg_montage, I found the output file have those channels that I changed to "other" reverted to "EEG", and all 'other' channels are still retained in the montage output file.

I found there are codes in spm_eeg_montage that set the channel type back to default.  Also, some of the channel that I set are as "other" are interspersed within the range of EEG channels (not only those at the very end), and seems the spm_eeg_montage will sort the channels and add some missing channels at the back (if the matrix is not MxM). So, I am worried that the steps I done are not correct and may have messed up the data.  How should I check these ?

Sorry that I may have get things confused, please kindly shed some lights.  Thank you so much!!

Best regards
SP Ho