[Eeglablist] removing eye channels before ICA?

[Chang Gu]

Dear colleagues,

I'm using ICA to clean the data, but I'm wondering that if we could remove
the eye channels before conducting ICA.

We're using 128-channel EGI recording system and there're 8 eye-channels on
the face picking up eye signals like blink and movement.
(1). For these 8 channels, they're usually recording huge potentials
comparing with regular channels on the sculpt: the signal amplitude from eye
channels could be 100 times to channels close to vertex before average
re-reference.* If I run ICA on all channels, is it possible that the big
variability in the eye channels influence the decomposition result?* From
our data, it seems that signals from eye channels draw most of the attention
of ICA and ICA decomposes many components of eye artifacts. However, we're
interested in decomposing brain-signal rather than eye-signals. Also, is it
possible that data-scale differences from eye channels (around 100 micro
volts) and regular channels (less than 3 micro) may influence the
performance of ICA algorithm?

(2). Since the eye-channels are on face, very often they have poor contact
with skin and dry out easily. From the ICA 2D topographic results, we often
found maps with a single red/blue dot on eye channel, suggesting the
distinct temporal character of the signal from this channel.* I'm wondering
that if this 'bad' channel may affect the overall ICA deposition. *

We tried removing these 8 eye-channels: other channels nearby also record
eye potentials (but not as strong as eye-channels for sure) and ICA are able
to separate eye-components from the EEG signal...
I'm wondering that, is there any theoretical concerns about removing
eye-channels?

Thanks in advance!

Chang

-- 
Chang Gu
Psychology & Human Development
Vanderbilt University
Nashville, TN
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110607/278dad8f/attachment.html>