[Eeglablist] EOG channels included in ICA and DIPFIT

Arnaud Delorme arno at ucsd.edu
Thu May 11 10:54:13 PDT 2017

```Hi Benedikt,

ICA returns linear decompositions/equations to reconstruct brain and artifactual sources. If you have an EOG channel which has the same reference as the other channels, it is all good, because any electrical activity (artifact or brain source) will project linearly to all channels (because of Maxwell equations) assuming no change in brain tissue conductance over time (which is a reasonable assumption). In other words, if you assume that brain tissue conduct electricity the same way over time, any electrical activity will project linearly to scalp channels (say with a factor 0.124 on Fz, 0.372 on VEOG1, etc…). Blink electrical potential at time t would for example be equal to 0.124 times the activity of Fz at time t plus 0.372 times the activity of EOG1 at time t etc…

Now the second case. Say you scalp channel reference is Cz. If you have bipolar montage for EOG channels (bipolar VEOG signal recorded from physical channels EOG1 and EOG2) then the linearity assumption is broken. You will first need to reconstruct the signal for EOG1 and EOG2 with reference to Cz. VEOG is equal to EOG2-EOG1 (EOG2 minus EOG1) and you want EOG1-Cz and EOG2-Cz so you can go back to the first case above. If you were include the VEOG channel in your ICA decomposition as is, ICA might not be able to find source that project linearly to VEOG and other scalp channels (because they each have a different references).

Note that the truth is a little bit more complex, because in the second case, ICA might be able to reconstruct EOG1-Cz and EOG2-Cz from bipolar VEOG, simply because it is a linear operation and reconstructing these channels might be part of the solution that maximizes independence for ICA. In other words, at the cost of using one additional ICA component to reconstruct the missing reference for VEOG, ICA might still be able to find a solution (and component scalp topography) as if you had recorded the EOG1 and EOG2 channels with the same reference as the data. However, this is not guaranteed. This would require more detailed investigation on real data. In practice, it is simpler to record EOG channels with the same reference as the other channels (you can always compute bipolar VEOG signal after the fact if you want to).

Hope this helps,

Arno

> On May 10, 2017, at 1:51 PM, Benedikt Ehinger <behinger at uos.de> wrote:
>
> Hi Arno,
>
> could you elaborate on why to not include them? Did you experience
> multiple eye-components, one for the EEG-electrodes and one for the
> bipolar eye-electrodes or are there other problems?
>
> Best,
> Benedikt
> Am 10.05.2017 um 18:22 schrieb Arnaud Delorme:
>> Dear Alessio,
>>
>> Yes, include the EOG channels but as long as they have the same
>> reference as the scalp channels. Otherwise, do not include them.
>> Best wishes,
>>
>> Arno
>>
>>> On May 10, 2017, at 7:33 AM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu
>>> <mailto:mmiyakoshi at ucsd.edu>> wrote:
>>>
>>> Dear Alessio,
>>>
>>> Getting back to you after 3 month!
>>>
>>>> how to deal with ICs where EOG contributes for example for 20%, or
>>> 40%, or 60%?
>>>
>>> This is a good question.
>>> Usually EOG is decomposed well, so the result is rather black and
>>> white. You can reduce it by optimizing settings (using proper
>>> preprocessing and parameters.)
>>>
>>> However, sometimes this mixture of brain EEG and non-brain artifact
>>> happens. In this case, you are forced to choose either false positive
>>> (include them) or false negative (exclude them). I tend to include
>>> them. If you exclude them, you typically find no frontal ICs.
>>>
>>> Makoto
>>>
>>>
>>>
>>>
>>> On Sun, Feb 12, 2017 at 1:31 AM, Alessio Matiz <muec at inbox.com
>>> <mailto:muec at inbox.com>> wrote:
>>>
>>>    Very dear EEGLAB-list,
>>>
>>>    is it correct to fit dipoles (via DIPFIT2-plugin) to ICs that have
>>>    been found including EOG channel in the ICA decomposition (as Arno
>>>    suggested in
>>>    https://sccn.ucsd.edu/pipermail/eeglablist/2007/001801.html
>>>    <https://sccn.ucsd.edu/pipermail/eeglablist/2007/001801.html>)?
>>>
>>>    From my datasets I removed ICs where EOG channel contributes for
>>>> 90% of IC power (considering them artifactual ICs), but how to
>>>    deal with ICs where EOG contributes for example for 20%, or 40%,
>>>    or 60%?
>>>
>>>    Thanks,
>>>    Alessio
>>>
>>>
>>>
>>>    _______________________________________________
>>>    Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>>>    <http://sccn.ucsd.edu/eeglab/eeglabmail.html>
>>>    To unsubscribe, send an empty email to
>>>    eeglablist-unsubscribe at sccn.ucsd.edu
>>>    <mailto: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
>>>    <mailto:eeglablist-request at sccn.ucsd.edu>
>>>
>>>
>>>
>>>
>>> --
>>> Makoto Miyakoshi
>>> Swartz Center for Computational Neuroscience
>>> Institute for Neural Computation, University of California San Diego
>>> _______________________________________________
>>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>>> To unsubscribe, send an empty email to
>>> eeglablist-unsubscribe at sccn.ucsd.edu
>>> <mailto: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 <mailto:eeglablist-request at sccn.ucsd.edu>
>>
>>
>>
>> _______________________________________________
>> 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
>>
> _______________________________________________
> 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

```