[FieldTrip] Question re: TMS-EEG FT Tutorial

Wed Apr 16 09:44:46 CEST 2014

Hi Bingshuo,

Be careful with categorizing IC’s based on their amplitudes alone. The kind 
people on the mailinglist may correct me if I am wrong but the exact 
amplitude of sources cannot be extracted using ICA. A very nice introduction 
to ICA by Arnaud Delorme can be found here: 
http://sccn.ucsd.edu/~arno/indexica.html. It contains a nice example where 
you can mix two sources and decompose them using fastICA (make sure to add 
the fastica toolbox to your matlab path, e.g. fieldtrip/external/fastica):

% Create source time course

A = sin(linspace(0,50, 1000));   % A

B = sin(linspace(0,37, 1000)+5); % B

figure;

subplot(2,1,1); plot(A);         % plot A

subplot(2,1,2); plot(B, 'r');    % plot B

% Mix sources

M1 = A - 2*B;                  % mixing 1

M2 = 1.73*A+3.41*B;            % mixing 2

figure;

subplot(2,1,1); plot(M1);      % plot mixing 1

subplot(2,1,2); plot(M2, 'r'); % plot mixing 2

% Decompose using ICA

figure;

c = fastica([M1;M2]);              % compute and plot unminxing using 
fastICA

subplot(1,2,1); plot(c(1,:));

subplot(1,2,2); plot(c(2,:));

Using the above line of code you can change the amplitude of the source time 
courses (e.g. by multiplying A and/or B by a constant) and you can see that 
the amplitude of the independent components remains unchanged.

I am not 100% certain but what I think Korhonen et al. are doing is 
transforming the individual IC components back to sensor level and then 
looking at the EEG amplitudes, which would be valid. In fieldtrip you could 
realize this be using ft_rejectcomponent with ‘cfg.component’ set to all 
components except the one you would like to keep.

I personally prefer to see whether the topography matches the location of 
areas that are stimulated by the coil and to check whether the averaged IC 
time courses contains the large exponential decay I would like to remove.

Could it be that after removing these high amplitude ICs your TEPs only look 
more noisy because the vertical scaling changes while plotting? In other 
words, the noise was already there before but you are just zooming in more 
because the plotting function does not have to fit the high amplitude 
components in the figure?

Best,

Jim

From: fieldtrip-bounces at science.ru.nl 
[mailto:fieldtrip-bounces at science.ru.nl] On Behalf Of Bingshuo Li
Sent: dinsdag 15 april 2014 13:22
To: FieldTrip discussion list
Subject: Re: [FieldTrip] Question re: TMS-EEG FT Tutorial

Hi Jim,

Thank you very much again for your thoughtful response! I truly appreciate 
it! I see your point now why you interpolate after ICA. It makes sense and I 
think your argument is very valid. Indeed, bad channels always come up 
fairly independently anyways. I can for sure do my channel inspection after 
the ICA.

Another question along the same line: Would you suggest to categorize all 
the IC with more than, say 30 uV, on the time-locked averaged IC, as muscle 
artifacts as suggested by Korhonen et al. (2011)? I often feel very puzzled 
when it comes to picking up ICs for removal. I feel that if I remove those 
high amplitude ICs, my averaged TEPs look worse. However, by keeping those 
high amplitude ICs, I am letting in signals with no biological plausibility. 
Any thoughts on that?

Thanks very much!

Regards,

Bingshuo

-----
Bingshuo Li (MSc. candidate)
Systems Neurophysiology Group
Centre for Integrative Neuroscience
University of Tuebingen
Otfried-Mueller-Str. 25
D-72076 Tuebingen, Germany
bingshuo.li at student.uni-tuebingen.de
+49-7071-29-89029

On Fri, Apr 11, 2014 at 12:48 PM, Herring, J.D. (Jim) 
<j.herring at fcdonders.ru.nl> wrote:

Dear Bingshuo,

I am forwarding your e-mail to the Fieldtrip mailinglist so that others can 
benefit and/or contribute from our discussion.

The reason why I like to postpone the interpolation to after the ICA is that 
I noticed that when interpolating before ICA the interpolated data would 
load onto one component. Removing this component did not make sense as it 
would leave a flat line in the TEP, keeping it in did not make sense either 
because the interpolation was done with, for example, the decay artifact 
still in the data. After removing the decay artifact with ICA you would be 
left with a sharp peak in the interpolated segment.

I see that it is difficult to reject trials and channels before 
interpolation but keep in mind that before having done ICA large amounts of 
variance of the data can be explained by the TMS related artifacts, these 
will most likely cloud the detection of other types of artifacts with 
functions such as ft_rejectvisual.

In my experience bad channels show up as separate independent components 
anyway and can be removed after having run the ICA, but that may depend on a 
number of factors I’m not aware of.

If you would like to reject bad channels prior to running ICA you could also 
try to run ft_rejectvisual and specify cfg.latency to contain only your 
pre-TMS period or a period post-TMS that does not contain any large 
artifacts. That way you should at least be able to run metrics on your data 
to remove bad channels. In any case you can always run ft_databrowser on 
your data to visually inspect your channel time courses for bad channels 
prior to running the ICA.

In any case you are of course free to try running the ICA after 
interpolation yourself and share your experiences, perhaps this works fine 
for you J

Best,

Jim

From: Bingshuo Li [mailto:bingshuo.li at cin.uni-tuebingen.de]
Sent: donderdag 10 april 2014 16:36
To: j.herring at fcdonders.ru.nl
Subject: Question re: TMS-EEG FT Tutorial

Dear Jim,

I inquired you about a TMS-EEG question a while ago and you referred me to 
the TMS-EEG tutorial on Fieldtrip's website. As I was following the steps of 
the tutorial, there is a new question that come up to me -- Is it really 
necessary to postpone the interpolation of the TMS artifact until the ICA is 
done? Would there be any bad consequences if I interpolate first and then 
run the ICA over the entire interpolated trial(s)? The reason I am asking 
this question is that I would like to visually inspect my data first and 
remove bad channels or trials (if any) prior to ICA. If the artifact is not 
interpolated, it is really difficult to run visual inspection or metrics on 
the data.. Following the logic of the tutorial, I can only visual inspection 
after the ICA, which I think might not be a good idea as I was told that a 
bad channel can easily bias the ICA result..

Do you have any insights in this? Thank you very much for your time in 
advance!

Sincerely,

-----
Bingshuo Li (MSc. candidate)
Systems Neurophysiology Group
Centre for Integrative Neuroscience
University of Tuebingen
Otfried-Mueller-Str. 25
D-72076 Tuebingen, Germany

bingshuo.li at cin.uni-tuebingen.de

+49-7071-29-89029 <tel:%2B49-7071-29-89029>

On Mon, Feb 10, 2014 at 3:24 PM, Herring, J.D. (Jim) 
<j.herring at fcdonders.ru.nl> wrote:

Dear Bingshuo,

Please have a look at the following tutorial: 
http://fieldtrip.fcdonders.nl/tutorial/tms-eeg

It deals with a number of TMS-EEG related artifacts including the ‘evil’ 
decay artifact.

Best,

Jim

From: fieldtrip-bounces at science.ru.nl 
[mailto:fieldtrip-bounces at science.ru.nl] On Behalf Of Bingshuo Li
Sent: maandag 10 februari 2014 15:06
To: FieldTrip discussion list
Subject: [FieldTrip] TMS-EEG Decay Artifact

Dear FT users and developers,

Does anyone have any experience in dealing with the so-called decay 
artifacts found in TMS-EEG? It is a relatively long lasting (up to 100ms) 
artifact that follows a waveform similar to exponential decay and it occurs 
sporadically in certain recording channels.

Any tips/hints/recommendations are greatly appreciated! Thank you!

Regards,

-----
Bingshuo Li (MSc. candidate)
Systems Neurophysiology Group
Centre for Integrative Neuroscience
University of Tuebingen
Otfried-Mueller-Str. 25
D-72076 Tuebingen, Germany
bingshuo.li at student.uni-tuebingen.de
+49-152-06054831 <tel:%2B49-152-06054831>

_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20140416/8b103e9c/attachment-0002.html>