[FieldTrip] source localising activity coherent with muscle channels not present in leadfield

Wed Oct 2 16:27:36 CEST 2019

Hi Kirstin,

This is hard to debug without any data to look at… Yet, to answer your questions: I wouldn’t do interpolation for missing channels. Rather, I’d ensure that I compute a forward model only for the EEG channels that are available. This can be tweaked with a cfg.channel argument for ft_prepare_leadfield. Regarding your second question: I’d expect ft_sourceanalysis (with cfg.method = ‘dics’, and cfg.refchan = ‘thenameofyouremgchannel’) to run through fine if your data is well-behaved, in the sense that it contains a cross-spectral density estimate (or fourier transformed data) for the EEG channels + the EMG, AND if your forward model (leadfield) has been computed on exactly the same EEG channels.
BTW, are you aware of this tutorial? http://www.fieldtriptoolbox.org/tutorial/coherence/ It might provide some pointers into the right direction.

Best wishes,
Jan-Mathijs

On 30 Sep 2019, at 14:40, Kirstin Heise <kirstin.heise at kuleuven.be<mailto:kirstin.heise at kuleuven.be>> wrote:

Thank you for your response, Jan-Mathijs and sorry for my response latency…

The script throws an error when using the pre-computed filter.
I do indeed have a mismatch between the channel number because the filter is computed for the possible number of EEG electrodes (in my case 127) and I am intending to run the next step in which I use the filter with the available data.

The first question is, do I in any case need to interpolate missing channels (I wanted to avoid this)? And if not, what would be the best approach to solve this, taking care of the channel labels and order?

The second question is related to the source localisation of the EEG activity correlated with the EMG. At this step, there are EMG channels in the data - I tried the tutorial and am confused which information is actually necessary for using the filter given that the sensor/channel labels must diverge between the filter (EEG only) and the data including the reference (EMG) channel.

Thanks again very much for your support.
Best regards,
Kirstin

On 12 Sep 2019, at 17:35, Schoffelen, J.M. (Jan Mathijs) <jan.schoffelen at donders.ru.nl<mailto:jan.schoffelen at donders.ru.nl>> wrote:

Hi Kirstin,

The EMG channels are not supposed to be added to the forward model.

The error that is thrown suggests indeed a mismatch between the channels in the data and/or their order, that are used for the covariance/csd matrix and the channels and/or their order in the precomputed forward model. I would start by inspecting the list of channels at the line in which the error occurs (either putting in an explicit breakpoint, or use the command 'dbstop if error’) in cfg.channel, and sourcemodel.label. I’d expect the cfg.channel to only contain the EEG channels, unless you put in a weird cfg.channel into the ft_sourceanalysis’ cfg. Note that you don’t need to put in a cfg.channel at this stage.

Best wishes,
Jan-Mathijs

On 12 Sep 2019, at 16:27, Kirstin Heise <kirstin.heise at kuleuven.be<mailto:kirstin.heise at kuleuven.be>> wrote:

Dear FieldTrip community,

I am following the pipeline of the  http://www.fieldtriptoolbox.org/tutorial/beamformingextended/#loading-of-the-data tutorial to source localise EEG activity that is coherent with muscle/EMG channels. I have a pre-computed leadfield (using simbio), which does not include the muscle channels that I want to use.

When attempting to run ft_sourceanalysis, I receive the following error message:

Error using ft_sourceanalysis (line 542)
There's a mismatch between the number/order of channels in the data,
with respect to the channels in the precomputed leadfield/filter.
This is not easy to solve automatically. Please look into this.

Is there a way of solving this post hoc? I am not sure which information is required in which part of the leadfield structure. I saw that in the tutorial data set, there are more sensors in the data set than included in the forward model and wonder whether it is just a matter of adding the additional channels “somehow” in the forward model?

Also, not having interpolated bad channels and hence having a different number of channels in the EEG data set, at what point can I control the numbering/order of the actual channels in the source reconstruction procedure?

Thanks a lot for your support.
Best,
Kirstin

_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
https://doi.org/10.1371/journal.pcbi.1002202

_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
https://doi.org/10.1371/journal.pcbi.1002202

_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
https://doi.org/10.1371/journal.pcbi.1002202

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20191002/9ddc9676/attachment.html>