[Eeglablist] **preprocessing

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Sat Jul 22 17:32:47 PDT 2017

Hello Zhaleh, notes below best wishes. In general the two ways should be
leading to approximately the same results/patterns.


You can drop really noisy eye channels, and ICA should still pickup eye
artifact ICs regardless. It's not clear whether your noisy eye channels
comes from eye movements, and from something like facial EMG and/or
tension/channel noise at eye channel locations. ICs shouldn't be used to
control for noise at a few or single channel. However, an IC source that
covers that region, upon removal of that IC, may decrease the total amount
of noise at such focused (eye) channels.

Your first way seems fine to get data that should be relatively clean. You
can add the following to the end of the first way:
Re-checking the data for residual artifactual epochs/segments [including
removal of epochs/segments)

Your second way, it's not clear from your notes where you start and stop
the "do the works on each segment". I assume you mean you stop doing stuff
for each segment, and then do ICA on all remaining epochs/segments.

Extra notes:
One should re-check the rebuilt EEG data after IC rejection using visual
and semi-automatic methods (e.g., ICA based detection of artifactual

One can re-apply the ICA to the full continuous data, and then remove ICs,
and then remove any still dirty time periods or epochs/segments. This can
leave you with more data, as some epochs/segments that are rejected before
ICA will be much cleaner and can be kept in.

Caveat: with only 32 channels you may want to focus on common neural ICs
across subjects instead of rebuilding the EEG data after dropping ICs.

On Sat, Jul 22, 2017 at 4:29 AM, Jaleh Mohammad alipoor <sugmad973 at yahoo.com
> wrote:

> Hello,
>  I’m doing a project on the effects of music on the functional
> connectivity of the brain using EEG. But l came across a confusing problem
> during preprocessing of my data!
> We all know that there isn’t a confirmed and stable pipeline for
> preprocessing so that all experts agree on. So I tried two different ways
> to preprocess my data that l’ve explained below but adding to the confusion
> is the fact that after finishing preprocessing and doing the main process,
> data acquired by each of these two ways gave me completely different
> results and now l wonder which of these ways and subsequently final results
> could be correct!!
>  but first l want to give you a little about my data and process. I took a
> 32-channel continuous data of length 350000 datapoints from each person(my
> subjects are children between 10 to 13 years old). These data consist of 12
> segments during which 12 music excerpts was played. The lengths of these 12
> segments vary between 12000 And 27000 datapoints. As you can see below the
> only major difference between these two pipelines that l’ve taken is doing
> segmentation as the final step or doing it earlier!!  My reason to abandon
> the first way and trying the second was the fact that the final segments in
> first way, after doing all of these steps and after segmentation were
> sometimes noisy and had noisy channels that made me confused!!
>  And now the two ways l took to clean the data and prepare it for the
> final process:
> *First way:*
> Define channel location file
> Band-pass filter between 1 to 40 Hz
> Rejection of paroxysmal artifacts
> Remove bad channels
> Run ICA
> Run ADJUST toolbox to find bad components
> Interpolate removed channels
> Re-reference to average
> Segmentation
> *second way:*
> Define channel location file
> Band-pass filter between 1 to 40 Hz
> Segmentation
> *And doing these works on each segment*:
> Remove paroxysmal artifacts
> Remove bad channels (but l didn’t remove if bad channels were fp1 and fp2
> because l thought they were bad channels because of eye blinks and eye
> movements (by visual inspection) that ICA would remove them properly and
> data after running ICA and ADJUST showed me that l thought correct)
> Run ICA
> Interpolate if there were removed channels
> Re-reference to average
> Re-checking the data for residual artifacts
> Whether  my reason to not remove bad channels if they are due to eye
> movements and running ICA  on them is correct?
> And how can l get sure that my process is correct and the data are really
> clean?
> l would be very grateful for your kind help.
> Best regards,
> Zhaleh
> _______________________________________________
> 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170722/2d679544/attachment.html>

More information about the eeglablist mailing list