[Eeglablist] ICA for removal of EOG artifacts

[angel]

Hi all,

I'm using the 128 channel GSN system (sampling rate = 200Hz) for EEG
recording on hand movement and imagined hand movement tasks.  The total
recording time per subject is 5 minutes, and during the recording, the
subject alternated between clenching and imagine clenching a ball with one
hand in 10 second intervals (giving 15 x 10s epochs for imagined-movement
and 15 x 10s epochs for actual-movement EEGs ).

I would like to use ICA for removal of EOG artifacts.  According to the
tutorial, I will need more than 3*N^2 sample data points for N channel
components, therefore I will need 3*128^2 = 49152 sample data points.  If I
run ICA on 5 minutes continous data I will have 5*60*200 = 60000 data sample
points, but If I run ICA on the 2.5 minutes concatenated epoched data then
I'll only have 30000 data points, which is less then the required sample
data points.  However, it was mentioned in previous discussions that ICA
will only return cleaner components for larger sample data points if the
quality and content of data remains roughly the same.

my question is:

Since there were 2 different conditions in my recordings (actual-movement
and imagined-movement), should I run ICA on the whole 5 minutes continuous
data (with both conditions) or should I run it for concatenated epoched data
(2.5 minutes for each condition) even though the data points were abit
insufficient?

Thank you

regards,
Angel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20061123/3c1fe5a5/attachment.html>