[Eeglablist] Removing artifacts using ICA on single epochs

[ADRIAN AVILA GARIBAY]

Dear EEGLAB list and all researchers,

I am looking for knowledgeable advice regarding the approach to removing
artifacts from EEG data using ICA decomposition and eliminating components
corresponding to artifacts.

Which of the following ways would be more appropriate in theory and
practice?

(Data sampling frequency is 500 Hz. 26 channels)

One could use epochs that are 30 seconds long each with the event of
interest in the center (-15 to 15 seconds), apply ICA on each single epoch,
inspect each epoch individually and eliminate the components corresponding
to artifacts on each epoch separately. This might cause to have epochs with
different number of components removed and different components removed
because maybe not all epochs have blinks and there might be epochs where it
was not necessary to remove any component (after cleaning the epoch will be
cut to two seconds beginning at the event of interest, so the 30 seconds is
only to have sufficient data for ICA, if there is no artifact within those
two seconds of interest why should you remove a component?). I know this
would be highly time consuming considering many epochs, many subjects and
different conditions.

This is in contrast to having 30 epochs that are 2 seconds long, apply ICA
to the epoch data, identify components to remove and eliminate the same
components from all of the epochs.

I would really appreciate any advice and comment on this issue. Thanks!

Adrián Ávila