[Eeglablist] train ICA on part of the data

Iversen, John jiversen at ucsd.edu
Tue Aug 3 06:35:46 PDT 2021

Hi Dan,

Yes, you can certainly train ICA on a segment of data then copy the results back to the original dataset then remove components.

See: https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Tips_for_copying_ICA_results_to_other_datasets_.2806.2F26.2F2018_updated.29



PS: Out of curiosity, what do you mean by 'mess up the triggers'?

From: Dan <dan.beniamini at mail.huji.ac.il<mailto:dan.beniamini at mail.huji.ac.il>>
Subject: train ICA on part of the data
Date: August 3, 2021 at 5:17:37 AM PDT
To: "eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>" <eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>>

Hi all,

I am using pop_runica() to clean blinks etc out of EEG data.
For some of my subjects the beginning of the data is very noisy- introducing a lot of noise into the ICA and corrupting the components.

Is there a way to train pop_runica() only on part of the data, but remove the components from the whole data.
(I will later mark these segments as artifacts but I don’t want to trim the data so I won’t mess up the triggers)


More information about the eeglablist mailing list