[Eeglablist] Any tips for unsupervised (sic) ICA ECG / EOG component rejection?

[stauros dimitriadis]

Dear Panagiotis,
There is a simple way for automated detection of artifactual components.
Firstly, you can calculate the correlation of each component with your ECG, EOG channel.
Secondly, you  can employ high-order statistics applied to the time course of indepedent components like kurtosis (see Nadia Mammone, Francesco Carlo Morabito "Independent Component Analysis and High-Order Statistics for Automatic Artifact Rejection").

Another practical tip is to use the following criterion in order to get an optimal number of ICAs.To find Nstable components (from N-channel data) typically requires more than kN^2
 data sample points (at each channel), where N^2 is the number of 
weights in the unmixing matrix that ICA is trying to learn and k is a multiplier. A practical criterion is  points, k/N^2 > 30 pts/weight 
points. So it is better to apply first PCA in order to meet the above requirements.
Finally, it is better to follow the above unsupervised technique as a guide and not as a automatic procedure.

Best regards
Dimitriadis Stavros
PhD candidate in NeuroInformatics, Dept. of Physics, University of Patras

1)Electronics Laboratory, Department of Physics, University of Patras
          2)Artificial Intelligence Information Analysis lab Department of Informatics 
          Aristotle University of Thessaloniki

http://users.auth.gr/stdimitr/




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110727/b408ead5/attachment.html>