[Eeglablist] Jason Palmer on ICA preprocessing and bipolar montage configuration

[Scott Makeig]

Makoto,

Yes it is possible. The bipolar matrix defines up to nchan-1 bipolar pairs,
and the remaining rows are defined so as to make the matrix invertible. For
example if we have 5 channels, you can define the bipolar matrix as:

M =

    1.0000   -1.0000         0         0         0
         0    1.0000   -1.0000         0         0
         0         0    1.0000   -1.0000         0
         0         0         0    1.0000   -1.0000
   -0.2500   -0.2500   -0.2500   -0.2500    0.7500

Here the last channel is 1/(nchan-1) times itself minus the average of the
other channels. (It could also be [0 0 0 0 1]).  Then the inverse is:

>> inv(M)

ans =

     0    -1    -2    -3    -4
    -1    -1    -2    -3    -4
    -1    -2    -2    -3    -4
    -1    -2    -3    -3    -4
    -1    -2    -3    -4    -4

Then you can run ICA on M * EEG.data. And then look at component maps
inv(M) * EEG.icawinv.

Best,
Jason

-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0559, http://sccn.ucsd.edu/~scott