[Eeglablist] Artifact rejection by ICA for continuous data

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Tue Sep 27 08:10:54 PDT 2016


Hello Nasrin, a few notes below about your questions, don't worry about not
having experience, it just takes a few days of learning/practice to
understand the basics, plus there are many tools you can use, all described
below. cheers!





**********BEGIN FOR NASRIN

If you are a newbie, review and learn from Makoto's pipeline for
eeglab/ica, just google "Makoto eeglab pipeline" for a list of expert steps
and suggestions. It's been mentioned many times on eeglablist and updated
recently with even more cool info for users. It's a good way to learn about
basic rules to follow, especially for beginners.

If you have no experience, it's okay, just take the time to download the
eeglab data sets, go through the eeglab tutorial for a fewdays to really
learn how to use eeglab, + watch the eeglab school videos, + review the
basic steps used by researcher with eeglab/ica in high-quality ICA/eeglab
articles via Google Scholar. Also, google your topic, such as "detecting
artifacts + eeglablist" to get many past eeglablist messages related to
your questions.

ICA does not need the data to be epoched (it can be continuous or epoched).
If you want to epoch it use the eeg_regepoch function.

Data should be cleaned at least a little before ICA to remove the
largest/worst multi-channel noise which will confuse ICA and make it give
worse results.

Be sure to review the visual artifact rejection tutorial on the eeglab
online tutorial. It's easy, just PLot the EEG data, then select periods to
remove, then save. It however requires trained eyes.

You can also try the Tools > Automatic Continuous data cleaning options
available via the eeglab gui. Another options is the more recent PREP
pipeline from Kothe/Mullen, which uses the same tools.

also try SASICA, ADJUST, IC-MARC (Frolich), and MARA eeglab plugins for
detecting bad ICs to remove from data. All are installable via the eeglab
gui, and the articles can be found via Google Scholar. One can learn a lot
about artifacts and ICA from the articles, and the examples given in the
eeglab tutorial.

You can also train yourself to detect good/bad ICs on the ICA
classification training site from eeglab/sccn developers. The documentation
and examples there are especially useful. I recommend doing at least 200 or
300 classifications to get your white belt in ICA kung-fu.

*http://reaching.ucsd.edu:8000 <http://reaching.ucsd.edu:8000/> *


If you want to exactly repeat processing in an article and can't do it by
yourself, try contacting the authors of the article or a local eeg expert
to give you some tips. To get some ideas from eeglablist, you can ask about
specific steps in a new email.


**********END FOR NASRIN
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20160927/ed24a24e/attachment.html>


More information about the eeglablist mailing list