[Mne_analysis] Conflict of preprocessing recommendations: ICA and high-pass

Brunner, Clemens (clemens.brunner@uni-graz.at) clemens.brunner at uni-graz.at
Tue Mar 13 07:43:35 EDT 2018
Search archives:

Hi Stefan!

What you are describing is in fact a standard approach to ICA. A good explanation why this is OK is given in Winkler et al. (2015): In theory, filtering does not change the ICA coefficients at all. Therefore, you can compute ICA on a 1Hz-filtered signals and then apply it to the same 0.1Hz-filtered signal. In practice, however, filtering does make a difference, because slowly changing drifts violate the stationarity assumption of ICA. Furthermore, filtering out nuisance signal parts tends to improve ICA decomposition because we are usually not interested in these low-frequency drifts. Taking these considerations together, computing and applying ICA on data filtered with different HP filters is a valid approach.

These references describe this approach in some detail:

I. Winkler, S. Debener, K.-R. Müller, M. Tangermann. On the Influence of High-Pass Filtering on ICA-Based Artifact Reduction in EEG-ERP. In: Proceedings of the International Conference of the IEEE Engineering in Medicine and Biology Society, 2015:4101–4105, 2015. https://doi.org/10.1109/EMBC.2015.7319296.

A. Hyvärinen, J. Karhunen, E. Oja. Independent Component Analysis. New York: John Wiley & Sons, 2001.

S. Debener, J. Thorne, T. R. Schneider, F. C. Viola. Using ICA for the analysis of multi-channel EEG data. In: M. Ullsperger, S. Debener (Eds.), Simultaneous EEG and fMRI: Recording, Analysis, and Application, pp. 121-133. New York: Oxford University Press, 2010.

J. M. Pignat, O. Koval, D. V. D. Ville, S. Voloshynovskiy, C. Michel, T. Pun. The impact of denoising on independent component analysis of functional magnetic resonance imaging data. Journal of Neuroscience Methods, 213(1), 105-122, 2013.

There are lots of published studies that do not describe why this works, but which use it in their analyses, for example:

C. L. Baldwin, J. D. Lee, N. Lerner, J. S. Higgins. Detecting and quantifying mind wandering during simulated driving. Frontiers in Human Neuroscience, 11, 406, 2017.

E. Jungnickel, K. Gramann. Mobile Brain/Body Imaging (MoBI) of physical interaction with dynamically moving objects. Frontiers in Human Neuroscience, 10, 306, 2016.

S. Meyberg, W. Sommer, O. Dimigen. How microsaccades relate to lateralized ERP components of spatial attention. Neuropsychologia, 99, 64-88, 2017.


> On Mar 12, 2018, at 10:59, Stefan Appelhoff <stefan.appelhoff at mailbox.org> wrote:
> Dear MNE users,
> I have a problem, that I would like to briefly state in three parts:
> Situation:
> - I am currently working on EEG data (64 channels, 500Hz sampling frequency).
> - I want to clean the data for blink and horizontal eye movement artifacts.
> - For this purpose, I would like to use ICA first for a single subject, to identify "templates", and then apply these templates to all subjects using the "corrmap" approach.
> Problem:
> The conflict I encounter is when it comes to preprocessing my data:
> - For ICA, a high-pass filter of 1Hz is recommended (see here)
> - For EEG data in general, high-pass filters with a cutoff > 0.1Hz have been shown to be problematic (see here)
> Question:
> Let's assume the following:
> - Given my raw data, I create two pre-processed versions: One version high-pass (hp) filtered at 1Hz and the other version hp filtered at 0.1Hz
> - I calculate the ICA on the 1Hz hp data to obtain mixing and unmixing matrices.
> - What are the caveats to be considered when I then apply the mixing and unmixing matrices obtained from the 1Hz hp data to the 0.1Hz hp data? 
> Help, and especially pointers to literature or examples are appreciated.
> I am also open to alternative approaches to addressing my situation.
> Best regards,
> Stefan
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
> The information in this e-mail is intended only for the person to whom it is
> addressed. If you believe this e-mail was sent to you in error and the e-mail
> contains patient information, please contact the Partners Compliance HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you in error
> but does not contain patient information, please contact the sender and properly
> dispose of the e-mail.

More information about the Mne_analysis mailing list