[FieldTrip] ICA in TMS-EEG

[Cristiano Micheli]

Dear J.D., Max and Bingshuo,

A brief comment: to be more confident about the correlative nature of an
artifact being an artifact (a.k.a. unwanted interference) I always consider
three things: topographies, time courses and spectra. The last one too,
since it might inform about spectral components not evident at the bare eye
from the time courses.
I'm not familiar with artifacts in TMS, but in all cases experience helps,
especially if you see the same pattern in many subjects, in other published
studies, from colleagues' data or from independent evidence (e.g.
concomitant electromiography recorded from the facial muscles, for muscular
artifacts identification).

HTH
Cristiano





On Wed, Nov 27, 2013 at 3:28 AM, Herring, J.D. (Jim) <
j.herring at fcdonders.ru.nl> wrote:

> Dear Bingshuo,
>
>
>
> I agree with Max that it would be beneficial to see the timecourses as
> well. Also, in case of ICA on TMS-EEG data I find it usefull to timelock
> average the time courses of the ICA components to the onset of the TMS
> pulse. That helps in identifying components that are related to the TMS
> pulse.
>
>
>
> Just from looking at the topographies I would say that component 7 is too
> posterior for being a cranial muscle artifact given that you are
> stimulating M1. Also, cranial muscle artifacts related to the TMS pulse
> usually last up to about 15ms after stimulation onset (see for example:
> http://fieldtrip.fcdonders.nl/_detail/development/tms/art_cranial_muscle.png?id=tutorial%3Atms-eeg) unless you are actually referring to a decay artifact (
> http://fieldtrip.fcdonders.nl/_detail/development/tms/art_decay.png?id=tutorial%3Atms-eeg)
> , which can last up to one second. You said that you cut and interpolated
> up to 18ms after stimulation so you should not see any cranial muscle
> artifacts related to TMS.
>
>
>
> Too me component 7 looks more like a topography related to posterior alpha
> oscillations, however, to be sure we need to see the time courses.
> Component 13 and 18 might be neck muscle artifacts, however, without
> performing a timelock analysis of the ICA components and inspecting the
> time courses it is hard to tell whether these are due to the TMS-pulse.
>
>
>
> Component 42 and 54 could indeed just reflect two bad channels, at least
> they seem to contain some variance that cannot be explained by a
> combination of the other channels.
>
>
>
> Best,
>
>
>
> Jim
>
>
>
> *From:* fieldtrip-bounces at science.ru.nl [mailto:
> fieldtrip-bounces at science.ru.nl] *On Behalf Of *Bingshuo Li
> *Sent:* dinsdag 26 november 2013 18:26
> *To:* fieldtrip at science.ru.nl
> *Subject:* [FieldTrip] ICA in TMS-EEG
>
>
>
> Dear FieldTrip Community,
>
> I recently started to analyze some TMS-EEG datasets and I encountered some
> questions regarding to using ICA to remove eye movement/muscle artifacts in
> our EEG data. As I am quite new to the analysis of TMS-EEG, I would like to
> inquire the FT community for some hints or suggestions. Below are the
> details of my questions:
>
> //Description of Data Processing//
>
> - EEG with 64 channel, sampling frequency 2500 Hz, electrode impedance
> less than 5 kOhm
>
> - Every epoch consists of 1s prior to and 1s after TMS (130-150 trials per
> subject)
>
> - TMS contaminated data points were cut out symmetrically -18ms to +18ms
> relative to TMS onset. Cubic spline interpolation is used to fill in the
> cut.
>
> - Bandpass 0.5 - 80 Hz, with BUT and filter order 3.
> - Discrete Fourier transform filter (cfg.dftfilter) to remove 50 Hz line
> noise
>
> - Visual inspection and rejection of trials with obvious unstable signal
> or channels.
>
>  //ICA//
>
> - ICA algorithm: runica
>
> - Demeaned data for ICA training (baseline is defined as the entire epoch
> -1 to +1s)
>
> - Unmixing matrix applied to non-demeaned data for component removal
>
> /////QUESTIONS/////
>
> Please see the image below for a typical result of ICA from a subject with
> TMS applied at M1 (32 epochs for ICA training):
> https://www.dropbox.com/s/chwo2jnwi72saba/ica1.png
>
> Q1: It seems obvious to me that component 1 and 2 are of eyeblink origin.
> However, what about component 5, 12, 20, 28? Topology-wise, they seem to
> have a very anterior origin, but data in the time domain does not seem to
> correlate with component 1 and 2 very well (judging visually..)
>
> Q2: What can you say about components 7, 9, 13 and 18? Are these cranial
> muscle artifacts?
>
> Q3: Also, for components 42 and 54, given their high focality, are these
> more or less a indication of bad/unstable electrodes?
>
> - I guess maybe I am asking too many questions. I think my main problem
> here is that I do not know what can be a good procedure / rules in manually
> selecting ICA components for rejection? (I tried to look in the literature
> but I couldn't find any that can answer my questions). And sometime I have
> the feeling that my ICA results look like a mess and maybe there were
> something wrong with my pre-processing or even data collection?
>
> Thank you guys in advance for any input! I look forward to hearing from
> you!
>
> Regards,
>
> Bingshuo Li
>
> MSc. Student, Neuroprosthetic Group, CIN, Uni Tübingen
>
> _______________________________________________
> 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/20131127/57f9f787/attachment-0002.html>