[Eeglablist] Inquiry about clean_artifacts.m function

[Cedric Cannard]

Dear Ingmar,

Interesting approach, thank you for sharing. 
You can also adjust the number of randsac samples to gain robustness (but lose computing speed) with built-in 'numSamples' input. It would be nice if someone could compare the results between that paper's approach vs increasing this value. If equivalent, it would be easier to implement in EEGLAB and GUI. 

Here is the help description:
NumSamples: Number of RANSAC samples. This is the number of samples to generate in the random sampling consensus process. The larger this value, the more robust but also slower the processing will be. Default: 50.


Cedric



------- Original Message -------
On Tuesday, January 24th, 2023 at 9:36 PM, ibrilmay at uni-koeln.de <ibrilmay at uni-koeln.de> wrote:


> Hi,
> 
> This topic comes up repeatedly on the list. There is a publication by
> the mobilab people from Berlin
> (https://urldefense.com/v3/__https://www.biorxiv.org/content/biorxiv/early/2022/10/06/2022.09.29.510051.full.pdf__;!!Mih3wA!D-VayU4D76htcPMSpfjj3jbRjEeNEUyjoVHHm3sG2_4CyNjLKMgVDLn2kFQ7hSWkS4HxjBNNB1gymhv2PNr2olAQWpU$ ), that
> says:
> "Subsequently, we repeatedly run the clean_artifacts function of the
> clean raw data EEGLAB plugin, with the number of repetitions specified
> by the bemobil_config.chan_detect_num_iter parameter. This is
> necessary because clean_artifacts uses a random sample consensus
> (RANSAC) approach that does not necessarily converge to the same
> results when repeated. The function stores the sampling in a micro
> cache that will be accessed when restarting the function without
> restarting MATLAB or clearing the micro cache beforehand, resembling a
> stable result. But hen using the function with a cleared micro cache,
> the detected channels might differ. To ensure a reproducible ad
> channel detection, we thus clear the micro cache and repeat the
> detection several times, with a recommended minimum of 10 iterations.
> Only channels that were flagged as ‘bad’ more than a given proportion
> of the processed data (specified in
> bemobil_config.chan_detected_fraction_threshold) are then detected for
> final removal. We exclude all EOG channels from the detected bad
> channels because their statistical properties will often lead to false
> positive detection."
> 
> Maybe this is something you might want to look into.
> 
> Best
> Ingmar
> 
> Zitat von Velu Prabhakar Kumaravel velu.kumaravel at unitn.it:
> 
> > I think it is the RANSAC algorithm that produces inconsistent removals. You
> > might take a look at Makoto's page here
> > https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Channel_rejection_using_RANSAC_in_clean_rawdata.28.29_.2803.2F21.2F2022_added.29
> > .
> > 
> > Best,
> > 
> > Velu Prabhakar Kumaravel, PhD Student
> > Center for Mind/Brain Sciences,
> > University of Trento, Italy
> > 
> > On Mon, 23 Jan 2023 at 17:07, Charalampos Georgios Lamprou <
> > 100063082 at ku.ac.ae> wrote:
> > 
> > > Dear EEGLAB developers,
> > > 
> > > I hope this mail finds you well.
> > > 
> > > My name is Charalampos Lamprou and I am currently working on EEG analysis,
> > > using EEGLAB. I would like to inquire about the clean_artifacts function
> > > and the way it works. I have noticed that even though I give as input the
> > > same data and using the same parameters, I don't always recieve the same
> > > results. In particular, in each realisation the function removes different
> > > channels. Hence, I would like to ask if the the function works in a
> > > stochastic way and if so, where exactly the stochasticity lies and why in
> > > different realisations, different channels are rejected.
> > > 
> > > Thank you in advance!
> > > 
> > > Kind regards.
> > > 
> > > Charalampos Lamprou
> > > _______________________________________________
> > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > To unsubscribe, send an empty email to
> > > 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
> > 
> > _______________________________________________
> > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > To unsubscribe, send an empty email to 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
> 
> 
> 
> 
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to 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