[Eeglablist] ICA misinformation

Tue Jun 13 23:15:41 PDT 2017

Dear Robert,

Below is the EDF file with the components removed.
There was 19 channels, so I got 19 components, of which I removed 2 (components 1 and 2). 1 was blink. 2 was lateral eye movements.
I have rerun the decomposition keeping all the data. This way you can align it with the dataset you sent me.

https://sccn.ucsd.edu/~arno/download/Australia-Post-ICA_delorme.edf <https://sccn.ucsd.edu/~arno/download/Australia-Post-ICA_delorme.edf>

Best wishes,

Arno

> On Jun 13, 2017, at 4:11 PM, Robert Thatcher <rwthatcher2 at yahoo.com> wrote:
> 
> Arnaud,
>    Thank you. Please attach the raw digital data produced after ICA reconstruction.  Images are not sufficient and also what is of concern is the phase differences between all of the channels and this cannot be evaluated without having access to the raw digital data following ICA reconstruction..
> 
> I look forward to receiving an EDF formatted data file with the post ICA reconstruction.  Also, how many and which ICA components did you delete to re-create the 19 channels, was it 18 or 17 or 16, etc?
> 
> Thank you,
> 
> Robert
> 
> 
> On Tuesday, June 13, 2017, 6:47:15 PM EDT, Arnaud Delorme <arno at ucsd.edu> wrote:
> 
> 
> Hi Robert,
> 
> I have imported your dataset in EEGLAB and I have looked at the data. The ICA components I get are actually beautiful.
> 
> I have taken a couple of screen captures. The first 2 components are eye blinks and lateral eye movements. 
> 
> Then I used one of the EEGLAB functions to compare before (in black) and after (red) ICA rejection. There are small captures of alpha oscillation before and after removing ICA. You can see that the phase of alpha is preserved (there is no need for spectral decomposition to see that the peaks are almost perfectly aligned which means that the phase is preserved).
> 
> I attached the code that you can use to get these results in EEGLAB (you will need to install the BIOSIG extension to import EDF files). I have only used EEGLAB default (except manual rejection of 2 portions of data totaling 3 seconds or so but you can try without that and it will most likely return very similar results). Imported the data, ran ICA (Infomax as in the paper) using default parameters and plotted results using EEGLAB menus. I did not tweak the data in any way.
> 
> Let me know if anything is unclear.
> 
> Best wishes,
> 
> Arno
> <Screen Shot 2017-06-13 at 3.31.50 PM.png><Screen Shot 2017-06-13 at 3.31.28 PM.png><Screen Shot 2017-06-13 at 3.30.56 PM.png><Screen Shot 2017-06-13 at 3.30.04 PM.png><Screen Shot 2017-06-13 at 3.29.02 PM.png><Screen Shot 2017-06-13 at 3.28.26 PM.png>
> 
> % import the data
> EEG = pop_biosig('/Users/arno/Downloads/Australia-Pre-ICA.edf’);
> 
> % remove some bad portion of data
> EEG = eeg_eegrej( EEG, [11 922;21141 21618]);
> 
> % look up channel locations
> EEG=pop_chanedit(EEG, 'lookup','/data/matlab/eeglab/plugins/dipfit2.3/standard_BESA/standard-10-5-cap385.elp','rplurchanloc',1,'changefield',{1 'labels' 'FP1'},'changefield',{2 'labels' 'FP2'},'changefield',{3 'labels' 'F7'},'changefield',{4 'labels' 'F3'},'changefield',{5 'labels' 'Fz'},'changefield',{6 'labels' 'F4'},'changefield',{7 'labels' 'F8'},'changefield',{8 'labels' 'T3'},'changefield',{9 'labels' 'C3'},'changefield',{10 'labels' 'Cz'},'changefield',{11 'labels' 'C4'},'changefield',{12 'labels' 'T4'},'changefield',{13 'labels' 'T5'},'changefield',{14 'labels' 'P3'},'changefield',{15 'labels' 'Pz'},'changefield',{16 'labels' 'P4'},'changefield',{17 'labels' 'T6'},'changefield',{18 'labels' 'O1'},'changefield',{19 'labels' 'O2'},'lookup','/data/matlab/eeglab/plugins/dipfit2.3/standard_BESA/standard-10-5-cap385.elp','rplurchanloc',1);
> 
> % run ICA (default)
> EEG = pop_runica(EEG, 'extended',1,'interupt','on’);
> 
> % plot properties of the first two components
> pop_prop( EEG, 0, [1  2], NaN, {'freqrange' [2 50] });
> 
> % Use menu Tools > Remove components to compare the data before and after ICA
> 
>> On Jun 13, 2017, at 2:32 PM, Robert Thatcher <rwthatcher2 at yahoo.com <mailto:rwthatcher2 at yahoo.com>> wrote:
>> 
> 
> Hi Arno,
>    I agree that the study by Montefusco-Siegmund  et al does not give a lot of detail and they used simulated EEG and simulated eye movement and a very large time series and they used phase unwrapping which is not necessary when one uses JTFA like the Hilbert transform.  They produced reliable phase changes due to the ICA reconstruction, albeit small in magnitude.  The analyses by Georges and colleaques used real human EEG and standard ICA software and more realistic simulations with much larger changes in phase differences due to ICA reconstruction.  Large changes in phase differences by ICA is best demonstrated by regular type EEG data samples and with more realistic mathematical simulations.  The EEG samples from the Australian workshop are better examples of ICA phase distortion and these examples are commonplace and WinEEG is commercial software used world wide.   Georges and colleaques as well as myself and colleaques can easily reproduce large phase distortion and will do so in a future publication.  I expect that we will share our analyses on this forum because of the rare expertise and interest by member of this forum.
> 
> It would be helpful for all of us together to mathematically and empirically test the changes in phase differences due to ICA reconstruction to allegedly remove artifact.   I think that at the end of the day we will find agreement and this will strengthen the proper uses of ICA.   
> 
> An example of important advantages of ICA are the last few postings on this forum about 3D source vectors and the best way to use the cross-spectrum and the imaginary numbers to derive reliable estimates of network hub/node coupling.  ICA feature detection is excellent, the problem is limited to the mathematically "ill-posed" reconstruction when used for the purposes of artifact rejection.
> 
> Best wishes,
> 
> Robert
> 
> 
> On Tuesday, June 13, 2017, 3:15:51 PM EDT, Arnaud Delorme <arno at ucsd.edu <mailto:arno at ucsd.edu>> wrote:
> 
> 
> Hi Robert,
> 
> Thank you for sharing the manuscript.
> 
> Regarding the actual EEG data (figure 1), component activities seem quite noisy for blink components and the scalp topographies also appear patchy which is not typical of ICA blink or eye movement components which are usually quite smooth. It might be that the ICA decomposition was suboptimal (preprocessing of the data is not detailed and this might be the reason why - although sometimes ICA also fails to return meaningful components for reasons which are not well understood). The units on this figure are wrong by the way (ICA component activity or scalp topography are not in microvolt). So, the poor quality of this decomposition might be responsible for the phase distortion you are mentioning.
> 
> Regarding the simulated data, projection to scalp channels are usually done using a leadfield matrix that models the conductivity of the different brain layers. In this paper, it seems to be a simple linear combination (and the coefficients and the positions of the sources are not indicated which makes it hard to reproduce these results). The phase difference shown in Figure 5 (on the order of 0.001 radian for most of them) even if significant, do not appear convincing - is a change in phase of 0.001 radian really that critical. Changing the ICA stop threshold might actually dramatically affect these results.
> 
> I will take some time to download your dataset and have a look at it.
> Best wishes,
> 
> Arno
> 
>> On Jun 13, 2017, at 11:34 AM, Robert Thatcher <rwthatcher2 at yahoo.com <mailto:rwthatcher2 at yahoo.com>> wrote:
>> 
>> ICA is a wonderful tool for feature extraction but it has limitations when used to allegedly remove artifact.   The brain is not a large balloon filled with saline – there are hubs and clusters of neurons connected in networks (Brodmann areas, etc).  The phase differences between hubs are due to differences in the synaptic rise times and synaptic integration times and differences in conduction velocity, etc.   These differences are vital and critical to understanding brain function (see Nunez, “Electrical Fields of the Brain”, Oxford Univ. Press, 1981 or Walter Freeman and many others).
>> 
> 
> 
> 
> <Screen Shot 2017-06-13 at 3.30.04 PM.png><Screen Shot 2017-06-13 at 3.30.56 PM.png><Screen Shot 2017-06-13 at 3.28.26 PM.png><Screen Shot 2017-06-13 at 3.29.02 PM.png><Screen Shot 2017-06-13 at 3.31.28 PM.png><Screen Shot 2017-06-13 at 3.31.50 PM.png>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170613/49c90891/attachment.html>