[Eeglablist] basic question about artifact rejection using ICA

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Tue Apr 1 15:49:49 PDT 2014

Hi Vivian, hoping the ultimate Frisbee is going great! There are several
quick and short suggestions below. Cheers.

1. The online eeglab artifact rejection tutorial has good examples, as well
as the Recent chapter by Makeig and Onton in Luck's second handbook.
Another resource is the Delorme 2004? Paper on ica artifact detection.
Ic maps for eyeblinks , lateral eye movement , and muscle activity tend to
be relatively stereotypical,as are their spectra. The chapter and articles
are online via google scholar or Dr.Makeig's publications page.

2. With practice, and review/knowledge of scalp maps from erp articles, and
especially a dozen or two ica for eeg articles, one can build up additional
expertise in spotting bad ICs.

3. Usually healthy ica maps are dipolar, match existing known components,
and are definitely not focused on one or few single channels.

4. Eeglab toolboxes including adjust and faster, or corrmap, can help you
find or hunt for particular ica maps.
There are also models in these toolboxes that hunt for particular and well
defined spatial maps.

5. Some Patterns have specific spatial distributions and known cognitive
associates and spectral peaks., such as mefiofrontal maps with theta as the
dominant frequency, or occipital ICs dominated by alpha. Eyebli

6. There are differing opinions as to whether to stay in Ica space or
rebuild the eeg and how many components to reject. I would reject only
artifactual components that are within the first 20 ICs in a solution .
There is usually not a good rationale for removing small components from
the data.

7. A search on the eeglab.list should find similar topics previously
discussed which may be of use for you.

8. I was not able get the ic picture you put up, but I'll try to quickly
comment later. No worries, it quickly becomes second nature ,and you can
use specific metrics for most decisions on this topic.

9. One critical point imho is to make sure to compare the results of
different pre-ica cleaning rules and different ica rules on the ica maps,so
as to get a feel for ground truth in a collection of eeg recordings. One
should also see more or less of artifactual ics across different people.

Best wishes!

Let me know if this helps a bit.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20140401/597e89e1/attachment.html>

More information about the eeglablist mailing list