[Eeglablist] ICA for ocular artifacts: split components?

[Makoto Miyakoshi]

Dear Agnese,

Thank you for sharing an interesting result.

My prediction is that there was a event-related external stimulus at 250 ms
that can affect the eye.

Another thing to check is that if these are not the 'ghost IC' pairs--read
this article.
https://sccn.ucsd.edu/wiki/Makoto%27s_useful_EEGLAB_code#How_to_avoid_the_effect_of_rank-deficiency_in_applying_ICA_.2803.2F31.2F2021_added.29

However, this does not explain the result. Basically, when ICA is forced to
decompose rank-deficient data, it shows pretty strange behavior, which is
not even well described (one example is shown here
https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Examples_of_artifact_from_ICA_.28Modified_09.2F08.2F2020.29;
another example I know of is that the ICA weight matrix was with complex
numbers). So practically anything can happen. But as ICA is just a spatial
fitler, it does not know about the timing of the stimulus at all. It can
still detect a change in a spatial filter, which is why I suspect an
event-related spatial fitler shift. This is as if you slightly shift the
entire EEG cap at 250 ms, then put their position back before the next
stimulus onset.

I'm very curious to know what happened!

Makoto


On Wed, May 5, 2021 at 10:23 AM Agnese Zazio via eeglablist <
eeglablist at sccn.ucsd.edu> wrote:

> Dear all,
>
> I am using ICA to correct for ocular movements in a TMS-EEG dataset with 74
> EEG channels, recorded with a sampling rate of 9600 Hz with
> online reference on FPz.
> Before ICA, I applied the following preprocessing steps: time interpolation
> around the TMS pulse (from 1 ms before to 2 ms after the TMS pulse), high
> pass filter at 1 Hz, downsampling at 4800 Hz, epoching from -200 to 500 ms
> after the TMS pulse (n=180 epochs), baseline correction (-100 -2 ms), SOUND
> algorithm for noise removal (Mutanen et al., NeuroImage 2018), artifact
> rejection. For ICA, I am using this code:
>
> EEG = pop_runica(EEG, 'chanind', [1: size(EEG.data,1)], 'interupt','on');
>
> When exploring the ICA components, In a few cases I found a 'strange'
> pattern: two components with a similar topography (but reversed in
> polarity) seem to reflect the same component but in distinct time-windows.
> As can be seen in the plot trials-by-time, the first component shows an
> ocular pattern up to about 250 ms, and the second component displays a
> similar pattern but from 250 ms on. What seems strange to me is that the
> temporal interruption at about 250 ms is very sharp. At this link (
>
> https://urldefense.com/v3/__https://drive.google.com/file/d/1YiRjx4BPEPC6-NFySPbN-sNHwN6MmbLC/view?usp=sharing__;!!Mih3wA!TSRqGkavVV4G_JOcKWIJFYOeQf68i2IF5NhBGlcT8W-_EeR9wq5KX0qMgBMeeuXsM_AxCw$
> )
> you can see a figure which makes it much easier to understand the problem.
>
> Is anyone familiar with this issue?
> I'm using Matlab 2020b and eeglab v.2020.0.
>
> Thank you in advance.
>
> Best,
> Agnese
> --
> Agnese Zazio, PhD
> Neurophysiology Lab
> IRCCS Centro San Giovanni di Dio Fatebenefratelli
> Via Pilastroni 4, 25125 Brescia, Italy
> Phone number: (+39) 0303501-596
> Site: *
> https://urldefense.com/v3/__http://www.cognitiveneuroscience.it__;!!Mih3wA!TSRqGkavVV4G_JOcKWIJFYOeQf68i2IF5NhBGlcT8W-_EeR9wq5KX0qMgBMeeuVh3da2rQ$
> <
> https://urldefense.com/v3/__http://www.cognitiveneuroscience.it/__;!!Mih3wA!TSRqGkavVV4G_JOcKWIJFYOeQf68i2IF5NhBGlcT8W-_EeR9wq5KX0qMgBMeeuVKp4TURw$
> >*
> E-mail:
> *agnese.zazio at gmail.com <agnese.zazio at gmail.com>*
> _______________________________________________
> 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
>