[Eeglablist] Spread blinks and eye artifacts over IC

[Stefan Debener]

Dear Mahsa,

Use a 1 Hz high-pass filter and get rid of non-stereotypical segments 
(and bad channels, if you have any) before submitting the data to ICA, 
this can dramatically improve decomposition quality. Avoid PCA 
dimensionality reduction. You can apply the unmixing weights to the 
unfiltered raw data and thereby correct blinks from near-DC signals of 
interest, like this:


EEG2 = pop_firws(EEG, ...); % submit only continuous data to hp filter

EEG2 = eeg_regepochs(EEG2); % segment data into 1sec epochs

EEG2 = pop_jointprob(EEG2,...); % get rid of those epochs containing 
non-stereotypical artifacts

EEG2 = pop_rejkurt(EEG2,...);

EEG2 = pop_runica(EEG2, ...); % run ICA on pruned, high pass filtered data

EEG.icaweights = EEG2.icaweights; % apply decomposition weights to raw 
data.

EEG.icasphere = EEG2.icasphere; % Since unmixing weights are W = 
EEG.icaweights * EEG.icasphere, update both fields!

clear EEG2;

EEG = eeg_checkset(EEG);


The following paper includes a toy example (figure 3) and a proper 
evaluation of the benefits of hp filtering:

https://www.ncbi.nlm.nih.gov/pubmed/26737196

Note that the eye blink ICA will not contribute zero voltage in between 
blink events. Whether this is due to imperfect unmixing, miniscule eye 
movements/saccades, or (most likely) both, is not entirely clear. In my 
experience, proper preprocessing improves both, eye blink attenuation 
sensitivity and attenuation specificity to an extent that is difficult 
to reach with other correction procedures.

Best,

Stefan


Am 10.11.17 um 00:54 schrieb mahsa shalchy:
> Dear Experts,
>
> I am trying to correct EEG data blinks and saccades using ICA, 
> however, in some cases (i.e. when the subject is double blinking), ICA 
> finds multiple eye related components.
>
> The problem is that except for the main blink component, other eye 
> contaminated ICs, have neural information and I can not easily discard 
> them ( please find an image of ICs here:
> https://www.dropbox.com/sh/qjz2v3ke6zdr1j1/AABoSzG4s6ZyAlUz_vYC8slsa?dl=0)
>
> I tried using PCA before ICA but it didn't seem to work.
>
> For the preprocessing I did the following steps before applying ICA:
> Resampling, Filtering (0.1-40), re-referencing to average 
> mastoids,epoching.
>
> I highly appreciate your help on this problem.
>
> Many thanks,
> Mahsa
>
>
> _______________________________________________
> 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

-- 
Prof. Dr. Stefan Debener
Department of Psychology
Neuropsychology	Lab
University of Oldenburg
D-26111 Oldenburg
Germany

Office: A7 0-038
Phone: +49-441-798-4271
Fax:   +49-441-798-5522
Email: stefan.debener at uni-oldenburg.de

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20171114/8f44704f/attachment.html>