[Eeglablist] Newbie needs help with evoked potentials and ICA

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Thu Aug 24 22:08:58 PDT 2017

​Greetings Malte, thanks for sharing about your adventure. ​ I've tried to
respond to all your thoughts below.

You're almost there, and trust that EEG will reward you. I would say first
pause for now and take a good week to really get into EEG methods and data
recording procedures. And then focus on one or two new/modified pilot tests
with 32 or 64 electrodes, probably using more trials for each condition,
and perhaps more trials for each nociceptive location.

First and perhaps most important: you don't have enough electrodes on the
head for good ICA. However you should be able to get good ERPs even from a
single channel (but then focus on single-channe-focused analyses, not ICA ,
which is for decomposing broad spatial patterns in scalp eeg into likely
"sources". Note that this approach has been used a lot (dense EEG / ica for
nocicpetive research), but then again so has traditional ERPs. Note that
eeglab can also do analysis that are just about single-channels.

To get better recordings, YES!! 32 or more electrodes equally on both sides
of the head,and then you could average reference. More than 32 that would
be better for ICA and science. 64 channels can be setup in 10 to 15 minutes
max with an properly trained data collector. For a montage, Google image
the 10-10 EEG montages, and mimic that. However, most systems come with a
cap of some sort, so hopefully you have caps. If you're placing electrodes
one by one without a cap, just make a proper grid on the head with a tape
measure. I would generally say make good use of your excellent EEG system.

To get better recordings, record more time during a resting-eyes-closed
period as a baseline before and after the pain blocks, and record more
trials of stimulation (pain, not pain, etc..). Also review more CHEPs
papers and methods (google 20 papers, and take the time to review methods
in all closely, see also their limitations and try to avoid major ones.
Benefit from past mistakes and limitations in papers.

Learn about ICA by googling (luca eeglablist reaching) for the excellent IC
classification training site. I recommend students beginning with ICA to
classify at least 200 ICs and to thorougly look through the tutorial on
that site, which has excellent examples of the "usual suspects" of
non-neural ICs, and some nice feedback in a practice mode. These ICs are
quite spatially stereotyped, and can easily be picked out of good ICA
decompositions. There also several good chapters online that cover what
neural ICs are and aren't, easily found via google or the SCCN articles
site. All ICA classification eeglab plugins also have papers where they
talk about and show examples of good and bad ICs. See for example, the Luck
ERP handbook chapter from Onton and Makeig on ICA, or the SASICA plugin
from delorme and colleagues. There are a lot of eeglab tutorials and videos
that are googlable and focus on ICA running, selection, etc..

Since you are saavy and a self-starter, you will benefit a lot from
reviewing/practicing with ADJUST, MARA, SASICA, plugins for ICA
classifiaction. Each will make recommendations about which ICs are
artifactual, and you can confirm easily (especially for eye and muscle
related ICs). Overall, don't worry the ICA classification plugins work
quite well if ICA and EEG expectations are met.

Note If you don't have adequate density of electrodes (prefarably 50 or
more) covering a good part of the whole head with even coverage, then ICA
will be unhappy. But, open secret, many researchers have published with ICA
results with less electrodes (but the electrodes always sample from across
the whole head, equally from front, back, left, and right quadrants of the
head. If you are interested in "sparse EEG or particular electrode sites,
that's fine, just record an appropriate number and spatial density of
channels first.

Beginners can really benefit from reading Steve Luck's and Mike Cohen's
books on EEG. (especially chapters on data collection). Both have online
chapters available, some with a focus on data collection. Both have
chapter-related videos online.

remember, if you haven't had a chance to yet,  to be sure to view EEGLAB
summer school videos, ERPLAB videos, powerpoints from EEGLAB and ERPLAB

Read a few more articles and mimic methods closely from the highest quality
paper. CHEPS are cool and pain science too. All easy to google.

It's common practice in the field to use common similar set of electrodes
sites on both sides of the head, ideally all channels you have, and all
channels should cover the head quite nicely and evenly.

Don't worry it's natural to make mistakes or weird chocices before having
already done ~50 recording session in at least a few studies, along with
data analysis of what you collect, under the guidance of a mentor.

you want to make sure the data is precleaned of some major artifacts before
ICA. (see eeglab tutorials and videos on this). and agian only with
approrpiate montage of channels.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170825/f11b9a56/attachment-0001.html>

More information about the eeglablist mailing list