[Eeglablist] filters, ICA and erp

Thu Oct 13 07:08:41 PDT 2011

Dear all,
thank you for all the replies and discussions started on these topics. It was very useful for me.
I have still some doubts though.
Scott, when you mentioned to keep 128 ICs without using the PCA, this implies that then you has 128ICs to analyze and to select. I guess that you are proposing to use some software to recognize the good ICs. I am always sceptical to use software for this kind of judgements. Which software would you suggest? Did you compare software and human selection?
Furthermore I was looking for some papers in which infomax and fastica were compared for IC separations possibly in different paradigms (like ERP analysis, PSD analysis, on continuous data) but I could not find any. Are you aware of any papers doing this? Or do you have any personal experience about it?
Thank you very much
Best
Sara
________________________________
From: Scott Makeig [smakeig at gmail.com]
Sent: 07 October 2011 15:44
To: Sara Graziadio
Cc: David Groppe; japalmer29 at gmail.com; eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] filters, ICA and erp

Sara - I strongly believe in keeping as many degrees of freedom as possible when performing ICA. How many are possible?  This depends mainly on the length of the data. With 128 channels, ICA must learn a 128x128 unmixing matrix (or AMICA, the 128x128 mixing matrix plus other coefficients) -- to learn these 16k (or more) variables requires >>16k data points (perhaps 300,000 or more). At a 25- Hz sampling rate,  300k time points requires 20 min of data recording - usual EEG/ERP experiments are at least this long.

Using PCA to reduce the dimension of the data unfortunately 'folds in' or 'squashes down' information in the smaller dimensions into the lower dimensional subspace -- I use the analogy of a 3-D data cloud shaped like a pizza with one small anchovy sticking up in the middle. After reducing the dimension of the data to 2 using PCA, the 'anchovy' portion of the data will be 'squashed down' into the (2-D) pizza (e.g., only its projection onto the pizza surface would be retained).  If the 'anchovy' dimension (i.e., that independent component and EEG source) were of interest, ICA could no longer find it. Even if not, it's presence in the (2-D) pizza would contribute noise to the (now 2-D) ICA decomposition...

Scott

On Fri, Oct 7, 2011 at 7:28 AM, Sara Graziadio <sara.graziadio at newcastle.ac.uk<mailto:sara.graziadio at newcastle.ac.uk>> wrote:
Do you mean that I should keep the external electrodes (that I was actually removing before the ica to have less noise) to have a better decomposition through  ica? I have 128 channels overall but I usually reduce the dimension to obtain around 20-30ICs (more than  97% of variance explained). The problem is though that the IC that have the muscle activity, when the muscle is silent, has cortical activity instead, if you know what I mean…don’t you have the same problem independently on the number of channels?
Thanks

Sara

From: Scott Makeig [mailto:smakeig at gmail.com<mailto:smakeig at gmail.com>]
Sent: 07 October 2011 15:20
To: Sara Graziadio
Cc: David Groppe; japalmer29 at gmail.com<mailto:japalmer29 at gmail.com>; eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>

Subject: Re: [Eeglablist] filters, ICA and erp

Sara,   Given enough channels (i.e., deg of freedom), ICA separates out each scalp/neck muscle activity -- Note there may be information of interest in their independent activities as well! Therefore, EEG montages including low electrodes need relatively high density there, since the spatial source (muscle) density is high around the insertions of the neck muscles into the skull (this is where the EMG activity appears on the scalp, with a polarity reversal in the direction of the muscle and centered at the muscle end).  This is quite likely why in your experience muscle activity "usually ... spreads on half of the components," as you say.

Scott Makeig
On Thu, Oct 6, 2011 at 2:50 AM, Sara Graziadio <sara.graziadio at newcastle.ac.uk<mailto:sara.graziadio at newcastle.ac.uk>> wrote:
Hello,
Thanks for your suggestion.

As I was planning to do also a PSD analysis on the data I guess that to remove the mean is not the best method if it works as a non-selective high pass filter, am I right?

I am applying the PCA before applying the ICA to reduce the number of components. How the data rank would be modified in this case?
I have to admit that it never happened to me that the muscle artefact is put in a single source with the ICA. Usually it spreads on half of the components, is this only my experience?

Thanks again

Best wishes

Sara

>-----Original Message-----
>From: David Groppe [mailto:david.m.groppe at gmail.com<mailto:david.m.groppe at gmail.com>]
>Sent: 05 October 2011 23:10
>To: Sara Graziadio
>Cc: eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>
>Subject: Re: [Eeglablist] filters, ICA and erp
>
>Hi Sara,
>   I found that a good way to improve the performance of ICA for ERP
>analysis is to
>1) Epoch your data into one or two second chunks time locked to the
>event of interest
>2) Remove the mean of each epoch at each channel
>3) Run ICA to remove artifacts
>4) Use a standard pre-event time window to baseline your data
>5) Compute your ERPs
>
>Removing the mean of each epoch acts as a crude high-pass filter.
>It's not nearly as selective as a "true" high pass filter but it
>doesn't distort the ERP waveforms as much either.  Moreover we've
>found that the procedure described above massively improves the
>reliability of ICA when compared to standard ERP prestimulus
>baselines:
>
>Groppe, D.M., Makeig, S., & Kutas, M. (2009) Identifying reliable
>independent components via split-half comparisons. NeuroImage, 45
>pp.1199-1211.
>
>Hope this helps,
>       -David
>
>
>
>On Wed, Oct 5, 2011 at 10:46 AM, Sara Graziadio
><sara.graziadio at newcastle.ac.uk<mailto:sara.graziadio at newcastle.ac.uk>> wrote:
>> Hello,
>> I would like just a suggestion about some data cleaning/analysis I am doing. I
>am doing an ERP analysis and I want to clean my data first with the ICA. In
>theory, though, I should not use an high-pass cutoff higher than 0.1 Hz to not
>reduce the erp amplitude. On the other side the ICA does not work well if the
>high-pass cutoff is lower than 0.5 Hz...what is then the best method to apply?
>Has anybody tested how robust the ica is with a 0.1Hz filter?
>> I have also another question: I am doing the analysis on 94 electrodes
>referenced to Fz. I planned to average reference the data but actually there is
>quite a large spread of noise on all the electrodes with this method (muscular
>artefacts for example from the temporal electrodes). But actually almost all
>the papers are using the average reference so I was surprised, am I the only
>one having this problem of noise? Would not be better just to keep the Fz
>reference and then perhaps to average the erps for every different cortical
>area and do the analysis on these averaged erps?
>>
>> Thank you very much
>>
>> Best wishes
>>
>> Sara Graziadio
>> Research Associate
>> Newcastle University
>>
>>
>>
>> _______________________________________________
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>> To unsubscribe, send an empty email to eeglablist-
>unsubscribe at sccn.ucsd.edu<mailto:unsubscribe at sccn.ucsd.edu>
>> For digest mode, send an email with the subject "set digest mime" to
>eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu>
>>
>
>
>
>--
>David Groppe, Ph.D.
>Postdoctoral Researcher
>North Shore LIJ Health System
>New Hyde Park, New York
>http://www.cogsci.ucsd.edu/~dgroppe/

_______________________________________________
Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu<mailto:eeglablist-unsubscribe at sccn.ucsd.edu>
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu<mailto:eeglablist-request at sccn.ucsd.edu>

--
Scott Makeig, Research Scientist and Director, Swartz Center for Computational Neuroscience, Institute for Neural Computation; Prof. of Neurosciences (Adj.), University of California San Diego, La Jolla CA 92093-0559, http://sccn.ucsd.edu/~scott<http://sccn.ucsd.edu/%7Escott>

--
Scott Makeig, Research Scientist and Director, Swartz Center for Computational Neuroscience, Institute for Neural Computation; Prof. of Neurosciences (Adj.), University of California San Diego, La Jolla CA 92093-0559, http://sccn.ucsd.edu/~scott<http://sccn.ucsd.edu/%7Escott>