[Eeglablist] BurstCriterion in clean_rawdata

[Makoto Miyakoshi]

Dear Srini

> Do you have citation yet?

It was not my data, and it was for the internal data discussion. I guess we
have no plan to publish it.

SD==20 became the best given that specific set ups. The method Nima, my
former colleague and now Syntrogi CTO, used there can be found in these
functions.

https://github.com/bigdelys/pre_ICA_cleaning/blob/master/detect_artifacts_by_robust_sphering_MIR.m

https://github.com/bigdelys/pre_ICA_cleaning/blob/master/robust_sphering_matrix.m

https://github.com/bigdelys/pre_ICA_cleaning/blob/master/mututal_info_reduction_time_course.m

Basically he used a sliding window mutual information reduction (MIR) as a
measure of ICA weight matrix performance and compared how much MIR
improvement was observed among 19 methods.

For psychiatric children's data I use SD==4, not 20. SD==20 would be most
appropriate if the data are already good and only big rare glitches are the
problems.

Makoto



On Wed, Sep 2, 2015 at 3:04 PM, Srinivas Kota <svkota at gmail.com> wrote:

> In clean_rawdata usage it was mentioned that 'BurstCriterion' standard
> deviation cutoff for removal of bursts is 3 for more aggressive without
> loosing much EEG, conservative is 5.
>
>
> Makoto, in the past you mentioned that SD==20 showed the best results in
> your unpublished study. Do you have citation yet? I am interested in
> learning how you made the comparison. I will like to do the same with my
> data with different SD values.
>
> I tried BurstCriterion values 3 and 20 on EEG during fast walking. With a
> value of 3, I could preserve whole data, where as with value 20, more than
> 50% of data was lost. I did frequency spectrum with 'pwelch' in MATLAB.
> Frequency spectrum prior and after ASR process differ with SD == 20 (not
> sure significant or not) compared to SD == 3.
>
> My EEG data was bandpass filtered 1 to 40 Hz.
>
> Here is the other parameters for ASR process
>
> arg_flatline = []; % default : 5 seconds
> arg_highpass = 'off';% in Hz.
> arg_electrode = [];% minimum electrode correlation, default : 0.85
> arg_noisy = [] ;% line noise, default: 4;
> arg_burst = 3;%[];% standard deviation cutoff for removal of bursts,
> default: 5, agressive 3
> arg_window = [];%default: 0.25
>
> Thank you for your suggestions.
>
>
> Best
> Srini
>
> ***********************************************************
> Srinivas Kota, Ph.D
> Research Scientist
> Movement and Neurosciences Center
> Institute for Rehabilitation Science and Engineering
> Madonna Rehabilitation Hospital
> 5401 South St, Lincoln, NE 68506
> Email: svkota at gmail.com
> Ph:  (618) 319-0471 (cell)
> www.madonna.org/research_institute/index.html
> <https://owa.madonna.org/owa/redir.aspx?C=RNnR5lrQvEmYU7lVtl6kLHIjwe3PiNAI8ouI90lD2Vnn4niuQ5QMBdLYDKmByFhfb525xS5XXT8.&URL=http%3a%2f%2fwww.madonna.org%2fresearch_institute%2findex.html>
> ***********************************************************
>
> _______________________________________________
> 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
>



-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20150904/d68c9752/attachment.html>