Independent Component Analysis on EEG

Independent Component Analysis (ICA) is often used to improve EEG signals. See blind source separation for the algorithm.

1. Selecting your channels

You typically will not want to do ICA on all 80 channels of your signal, some of which may not even be EEG channels. To create a smaller EEG with e.g. only the 64 scalp channels, use EEG: Extract channels.... From here on we will assume that your reduced EEG has 64 channels.

2. How to get a MixingMatrix

Once you have your reduced EEG, you can start to do ICA on it. This starts by creating a MixingMatrix: select your EEG object and choose To MixingMatrix.... The resulting MixingMatrix has one row for each of your 64 EEG channels, and columns called "ic1" through "ic64".

3. How to see the independent components

Select your EEG and your MixingMatrix together and choose Unmix. The resulting ICA-EEG will have 64 channels called "ic1" through "ic64".

4. Just checking back

If you select your ICA-EEG together with your MixingMatrix and choose Mix, the resulting EEG should be very similar to your original 64-channel EEG.

Links to this page


© ppgb, May 2, 2018