[Eeglablist] PREP: how to best use for ICA?

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Mon Jun 29 13:40:09 PDT 2015

Dear Michael,

I confirmed it with Nima and Kay. The interpolated channel list is stored
at EEG.etc.noiseDetection.interpolatedChannels.all., so removing all the
channels listed there is the solution.

Kay also said that in the next release they will support the option to
discard bad channels rather than to interpolate. Nima suggested one should
always check the rank whether to use 'pca' option in performing ICA since
potential electrode bridging may be there, which is a good point is missing
from the conventional EEGLAB pipeline (ICA seems to have it's own rank
checker, but Jason told me that determining the data rank is not


On Fri, Jun 26, 2015 at 3:09 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu>

> Dear Michael,
> Yes, I was wondering the same thing too.
> For ICA point of view, interpolating channels is meaningless. If one can
> simply reject channels instead of interpolating, that's the best for us.
> I'll ask Nima if no-interpolation is in the option.
> My impression is that they care channel-level cleaning more than I do. For
> example, when you use clean_rawdata, which Christian Kothe developed, the
> suit spend long time to detect and reject bad channels.
> I'll get back to you when I hear back from Nima.
> Makoto
> On Sat, Jun 20, 2015 at 2:12 PM, Michael Spezio <
> mspezio at scrippscollege.edu> wrote:
>>  Makoto, thank you for updating your pipeline page and pointing out the
>> new PREP paper.
>> It's great to have the very helpful updates from that paper of using
>> double precision and using 1 Hz highpass filtering prior to using Cleanline.
>> For anyone on the EEGLAB list who can help, I have a question.
>> With regard to ICA downstream, however, I'm wondering if the new PREP
>> pipeline could be used to detect and eliminate bad channels, but without in
>> the end using the interpolation results and the resulting average
>> reference? It seems to me that if PREP-with-interpolation results in a
>> dataset that is intended to be used for ICA and subsequent component source
>> inference, the nonlinearities introduced by PREP's interpolation-and-rereferencing
>> will lead to inexact outcomes.
>> On the other hand, if PREP is a robust method of finding and eliminating
>> bad channels, one could use it for that and then proceed only with the good
>> electrodes, using their post-Cleanline signals (pre-PREP) for the
>> downstream steps of calculating the average reference and ICA.
>>  Does this sound right? Any insight that anyone could share on this?
>>  Best,
>>  Michael
>> _______________________________________________
>> 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
> --
> Makoto Miyakoshi
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego

Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20150629/ef003be2/attachment.html>

More information about the eeglablist mailing list