[Eeglablist] Questions about ICA component rejection in EEGLAB

[Yahoo]

Hi Pan,

What you need are FASTER (www.mee.tcd.ie/neuraleng/Research/Faster) and Fieldtrip which are completely compatible with EEGLAB.

With these two toolboxes, you could use the "component_properties" function in FASTER  to get more than three indexes (namely, median gradient value, spectral slope around the LPF band,Hurst exponent,kurtosis of spatial map and eye blink correlations ) to evaluate the quality of the ICAs. See the manual of FASTER for detail information.

In your analysis procedure, after running the ICA in EEGLAB, you may would like to do something like followings:

     icaProperties= component_properties(EEG,[64 65]);% in my case the 64, 65 channels are the  eye channels

    EEG.FASTER.icaProperties= icaProperties;

    icaProperties=(icaProperties-repmat(mean(icaProperties,1),size(icaProperties,1),1))./repmat(std(icaProperties,0,1),size(icaProperties,1),1);

    EEG.FASTER.rejectICAs = find(sum(icaProperties>3,2)>0);





> (2). While doing ICA component rejection, how many of the original components would you reject? I sometimes rejected almost half of them, is it too much or not?


I 'm not sure whether "Half of ICAs" is too much. But  in my case, with 65 channels data, I usually reject less than 10 ICAs and do indeed get good results. 
Good luck !

Yang

 

On Feb 26, 2013, at 3:20 AM, Liu Pan wrote:

> Dear eeglabbers,
>  
> I have 2 questions about the ICA component rejection function of EEGLAB:
>  
> (1). I read the paper "Automatic artifact rejection for EEG data using high-order statistics and independent component analysis" by Drs. Delorme, Makeig, and Sejnowski. http://www.google.ca/url?sa=t&rct=j&q=&esrc=s&frm=1&source=web&cd=2&ved=0CEAQFjAB&url=http%3A%2F%2Finc.ucsd.edu%2Fica2001%2F117-delorme.pdf&ei=JakrUbjhCMe30gHevIGoCQ&usg=AFQjCNGwjRtjnTiRvUxguqJNAejvoD6a7g&sig2=9BQuz03YEzUOszEreafYgA
> In section3.3 the authors mentioned that in addition to manually and visually rejecting ICA components based on the component properties, there is a function in EEGLAB software by which one can set an adequate rejction threshold for three high-order statistical measures (the entropy of the activity of the component, the kurtosis of the activity, and the kurtosis of the components' spatial map), then one can automatically detect and reject components (insteady of mannually doing so). In search of this function in EEGLAB, I found that in the Tools--Reject data using ICA--Reject components by map function, there are 3 buttons at the bottom of the popped out window (set threshods, see comp. st..., see projection) which appear to be grey and inactive. I suppose these buttons, especially the on set threshods, are relevant to the automatic ICA component rejection function mentioned in that paper.  However, these buttons are not active and cannot be used. According to the Help file of this function, "if the function POP_REJCOMP is ran prior to this function, some fields of the EEG datasets will be present and the current function will have some more button active to tune up the automatic rejection."  I searched POP_REJCOMP function but failed to get a clue from the tutorial nor the previous discussion list. Could anyone give me any suggestion about this? Are these inactive buttons the ones for automatic ICA component rejection?  If yes, how can I get them to work?
>  
> (2). While doing ICA component rejection, how many of the original components would you reject? I sometimes rejected almost half of them, is it too much or not?
>  
> Thank you very much for any of your input in advance! Pan
> 
> -- 
> Liu Pan, PhD. Candidate
> Neuropragmatics and Emotion Lab (Pell Lab)
> School of Communication Sciences and Disorders
> Faculty of Medicine, McGill University
> Montreal, Canada
> 
> _______________________________________________
> 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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130226/9e3a76bc/attachment.html>