[Eeglablist] removing eye blinks from 256 channel EEG

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Thu Dec 4 19:02:03 PST 2014


Dear Jeff,

If your data is 10 min and 1000Hz sampled, you have 600,000 data points.
Divide 600,000 with 30 and square root it. You get 140. So use 'pca', 140.
I would recommend 'pca', 70 or lower to give it a good margin.

Makoto

On Thu, Dec 4, 2014 at 6:50 PM, Eriksen, Jeffrey :LGS Neurodiagnostics <
JEriksen at lhs.org> wrote:

>  Makato,
>
>
>
> Thanks for reminding me of the formula. Now that I have looked at 10 and
> 20 minute decompositions, it seems to be working better. Of course, I am
> still well below the 30x factor. I will try your ‘pca’ option as well, but
> I do not understand what X is. Is X the 30 (or a lower or higher value), or
> is X the number of data points?
>
>
>
> -Jeff
>
>
>
> *From:* eeglablist-bounces at sccn.ucsd.edu [mailto:
> eeglablist-bounces at sccn.ucsd.edu] *On Behalf Of *Makoto Miyakoshi
> *Sent:* Thursday, December 04, 2014 6:30 PM
>
> *To:* K Jeffrey Eriksen
> *Cc:* eeglablist at sccn.ucsd.edu
> *Subject:* Re: [Eeglablist] removing eye blinks from 256 channel EEG
>
>
>
> Dear Germano and Jeff,
>
>
>
> SCCN's rule of thumb is that you need 30*(number of channels)^2 datapoints
> for good ICA results. The first 30 should go up as the number of channels
> goes up.
>
>
>
> > Do you mean I should get more than 1-3 eye movement related components?
>
>
>
> Yes.
>
>
>
> > What do you suggest to try to improve the decomposition? I know there
> are some free parameters I could try, but what should I try first?
>
>
>
> Try 'pca' option when you run infomax. 'pca', X (X is a number) so that X
> meets the above equation. It does not guarantee to improve your
> decomposition though, since decomposition quality also depends on how clean
> your data are.
>
>
>
> Makoto
>
>
>
> On Thu, Dec 4, 2014 at 3:56 PM, K Jeffrey Eriksen <eriksenj at ohsu.edu
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__mailto-3Aeriksenj-40ohsu.edu&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=JN7WdfuI4xbLDmbGZLvRzolYZ6Ps1ODU2zbaUytWa-Q&e=>>
> wrote:
>
> Makoto,
>
>
>
> Thanks for your response. I believe I successfully obtained eye movement
> components 10-12 years ago with 10 or 11 minutes of 256-channel, 250 sps
> EEG.
>
>
>
> Do you mean I should get more than 1-3 eye movement related components?
>
>
>
> What do you suggest to try to improve the decomposition? I know there are
> some free parameters I could try, but what should I try first?
>
>
>
> Thanks,
>
> -Jeff
>
>
>
> *From:* Makoto Miyakoshi [mailto:mmiyakoshi at ucsd.edu
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__mailto-3Ammiyakoshi-40ucsd.edu&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=itl5_2QIm-QQ8wrv2IyZ13cYYwm4gKnPIGTPmesu42w&e=>]
>
> *Sent:* Thursday, December 04, 2014 11:33 AM
> *To:* K Jeffrey Eriksen
> *Cc:* eeglablist at sccn.ucsd.edu
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__mailto-3Aeeglablist-40sccn.ucsd.edu&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=ZP83Yy9d1NyZcAOgolQh1Oa1acrC6-JyuZg4DIfxSQY&e=>
> *Subject:* Re: [Eeglablist] removing eye blinks from 256 channel EEG
>
>
>
> Dear Jeff,
>
>
>
> > My epochs range from 3 to 20 minutes.
>
>
>
> Sounds too short for 256ch decomposition. You need at least > 1 million
> datapoints for it. Using high sampling rate such as 1000Hz increases
> apparent datapoints, but I'm not sure how it helps.
>
>
>
> > I generally find 1-3 ICs for each subject.
>
>
>
> This sounds too few for 256ch decomposition.
>
>
>
> > general background noise goes UP after these 1-3 components are removed.
>
>
>
> If the removed components are coupled with other components that cancels
> out each other, this could happen. However that's kind of rare and I don't
> see it very often... most likely your decomposition has problems.
>
>
>
> Makoto
>
>
>
> On Wed, Nov 26, 2014 at 3:18 PM, K Jeffrey Eriksen <eriksenj at ohsu.edu
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__mailto-3Aeriksenj-40ohsu.edu&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=JN7WdfuI4xbLDmbGZLvRzolYZ6Ps1ODU2zbaUytWa-Q&e=>>
> wrote:
>
> EEGlabbers:
>
>
>
> Does anyone out there have particular experience from removing eye blinks
> from 256 channel EEG data using ICA? I have started to apply it to 11
> subjects with data recorded at 1,000 frame/sec, bandpass filtered &
> downsampled to 250 fps. My epochs range from 3 to 20 minutes. I generally
> find 1-3 ICs for each subject. When I remove these, there is still visible
> remnants around the time of each eye blink. Of more concern is that the
> general background noise goes UP after these 1-3 components are removed.
>
>
>
> Any ideas that might improve my results? I can post some screenshots if
> need be.
>
> Thanks,
>
>
>
> -Jeff
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__sccn.ucsd.edu_eeglab_eeglabmail.html&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=4cQxIxtsUTujLB6Q1DYRg64Tjz6kN11URw_wj-okW04&e=>
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__mailto-3Aeeglablist-2Dunsubscribe-40sccn.ucsd.edu&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=orz8u32bGNdJZ7qFpKpcuE_UXVvcqGpYE3brKQIlHwg&e=>
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
> <https://urldefense.proofpoint.com/v2/url?u=http-3A__mailto-3Aeeglablist-2Drequest-40sccn.ucsd.edu&d=AAMFaQ&c=5aMxbibQyX6B-W2mpD2HYH5lwK3h7A1hIsNQieXCZKA&r=RW9gWwlQ_echeF9GoAB-bw&m=1L894JApoM0afiTll4HrtL0s3mCkc9RKPnHk_qjFlaY&s=P8xxeyL3kNWLJnOjS25s1nntuk7oam0F5KWbLO6sLGU&e=>
>
>
>
>
>
> --
>
> 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
>



-- 
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/20141204/876da09e/attachment-0001.html>


More information about the eeglablist mailing list