[Eeglablist] unusual ICs

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Sun Jun 7 17:50:55 PDT 2015


Dear Isaiah,

> 1. Do an intial rejection of the data, removing paroxysmal sections that
would be hard for the ICA to interpret.
2. Run ICA, use the component activations scroll  to identify epochs that
are bad and should be rejected.
3. Run ICA again on the cleaned dataset. Now reject ICs that correspond to
blinks, eye movement, etc . . .

Yes, it looks good.

Check also Delorme et al. (2007) NeuroImage to see why you want to reject
epochs after ICA.

Makoto


On Fri, Jun 5, 2015 at 8:16 AM, Isaiah Innis <isainnis at indiana.edu> wrote:

> Thank you for your help - it seems it was a rank problem.
>
> If I understand the pipeline correctly 2 ICAs are performed, but the first
> is used only to help identify bad epochs. So the steps would look like this:
>
> 1. Do an intial rejection of the data, removing paroxysmal sections that
> would be hard for the ICA to interpret.
> 2. Run ICA, use the component activations scroll  to identify epochs that
> are bad and should be rejected.
> 3. Run ICA again on the cleaned dataset. Now reject ICs that correspond to
> blinks, eye movement, etc . . .
>
>
> Would that be the recommended path to follow?
>
>
> Thank you again,
>
> 2015-05-21 13:19 GMT-04:00 Makoto Miyakoshi <mmiyakoshi at ucsd.edu>:
>
>> Dear Isaiah,
>>
>> That's a subspace. You ran the second ICA after IC rejection, correct?
>> Actually, the second ICA is NOT after IC rejection. The results from the
>> 1st ICA should be used for cleaning using icaact and not IC rejection. I
>> know it is a very tempting idea to run 1st ICA, reject ICs, then run the
>> 2nd ICA to find new faces... but it never works because when you reject ICs
>> your forward-projected scalp EEG data are rand reduced.
>>
>> See also 're-reference' part of the following wiki page
>> http://sccn.ucsd.edu/wiki/Makoto%27s_preprocessing_pipeline
>>
>> Makoto
>>
>> On Wed, May 20, 2015 at 5:59 PM, Isaiah Innis <isainnis at umail.iu.edu>
>> wrote:
>>
>>> Hello eeglab list:
>>>
>>> After running a 2nd ICA on a dataset we sometimes run into into strange
>>> ICs in the first few positions, usually ICs 1-5.
>>>
>>> Here is an example (images of IC1-4 for a single subject):
>>>
>>> https://www.dropbox.com/sh/zz9zt7qoa8coq85/AAD7J3FaEx7QWEe_ti9OuLxda?dl=0
>>>
>>> It looks like it might be the same thing in the "How to deal with
>>> 'corrupted' ICA decompositions section on the wiki but I don't know for
>>> sure. We do normally reduce our 64ch recordings to 32ICs using the 'pca'
>>> option.
>>>
>>>
>>> Does anyone have any experience with this? Is something wrong with our
>>> settings?
>>>
>>> Thank you,
>>>
>>> --
>>> Isaiah Innis
>>> Indiana University '13
>>> EEG Technician, IUB IRF
>>>
>>>
>>> _______________________________________________
>>> 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
>>
>
>
>
> --
> Isaiah Innis
> Indiana University '13
> EEG Technician, IUB IRF
>
>


-- 
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/20150607/1b4d5b90/attachment-0001.html>


More information about the eeglablist mailing list