Importing fieldtrip data

mheers · February 20, 2022, 7:58pm

Dear Francois,

I have fieldtrip dataset that contains sensor MEG data, virtual sensors from MEG and EEG data. I would like to review raw data of all modalities in brainstorm using the fieldtrip data importer. So far it loads the MEG channels as 'MEG' separate from EEG and virtual channels, which are loaded as 'OTHER'. Do you have a suggestion how to include the virtual channels here? Can I share with you an example dataset?

Best regards,
Marcel

Francois · February 21, 2022, 9:17am

Hi Marcel,

You can change the type of the sensors easily: Right-click on the channel file > Edit channel file.
https://neuroimage.usc.edu/brainstorm/Tutorials/Epilepsy#Prepare_the_channel_file

Best,
Francois

mheers · February 22, 2022, 9:46pm

Thanks, Francois! This is really a simple solution of my request.

mheers · May 31, 2022, 6:39pm

Hi Francois,

There is another question related to the data set from fieldtrip. When I am trying to display the MEG channel topography I get the following error:

I can display the channel positions without any problem:

You can download an example dataset from here:
https://bwsyncandshare.kit.edu/s/ytfa2okF3fNNm2m

Thanks for looking into this!
Marcel

Francois · June 1, 2022, 8:09am

Hi Marcel,

There was some sort of bug in the Brainstorm code, that led the MEG magnetic extrapolation (smoothing of the MEG topography) not working when there are different numbers of coils among the sensors.
This is something I fixed here: Inverse: MEG channel extrap robust to different number of coils · brainstorm-tools/brainstorm3@676d77b · GitHub

However, there is also another problem with the file you shared. The Brainstorm functions do not seem completely equipped to read it... If you look in the Matlab command line after reading the file, you'd see:

Error: Wrong number of coils for channel MRT41: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT51: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT52: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT53: Cannot import this file correctly...

The positions of these sensors are defined in an unusual way. I'll need to ask the FieldTrip team what this can possibly mean, to know if we should get rid of these of these sensors, or handle them in a different way. I'll get back to you on this when I get an answer.

Francois · June 1, 2022, 1:01pm

Update:

I fixed the code with an empirical fix, that should work for reading the file you shared:
Import FieldTrip channel structure: Binarization of grad.tra · brainstorm-tools/brainstorm3@a83914b · GitHub
Since I am not sure why the previous version of the reader does not work on this file, I also opened an issue on the FieldTrip github:
Getting the coil-channel correspondence from grad.tra · Issue #2040 · fieldtrip/fieldtrip · GitHub

Right now, I think you can update Brainstorm and safely link again this file to your database.
I'll post further updates here if any (depending of the answer on the FieldTrip side)

mheers · June 1, 2022, 7:53pm

Thank you Francois, it seems to work now!

Just one more comment: For me it looks like there is currently no clear definition whether negative or positive deflections of traces point upwards or downwards when reviewing MEG/EEG data. Considerung also the amplitude maps it looks like you start with positive amplitudes upwards initially which you can flip using "flip y-axes". The clinical standard is that negative amplitudes point upwards. Can you use this as standard setting for MEG/EEG traces and can you note somewhere in the traces window which definition is currently chosen? This will be very helpful for reviewing of epilepsy data sets.

Francois · June 2, 2022, 9:49am

Thank you Francois, it seems to work now!

It works for your file, but my work is not exactly finished, as I still need to understand if I can improve the heuristic I'm using to connect channels and coils...

You removed the file too soon for the FieldTrip developers to be able to download and test your file with my code. See: Getting the coil-channel correspondence from grad.tra · Issue #2040 · fieldtrip/fieldtrip · GitHub
Can you please share it again?

I won't use the reverse Y axis as this specific to clinical EEG and SEEG. It would be too confusing for other users. But I will add something to indicate graphically the direction of the Y axis.

Francois · June 3, 2022, 2:33pm

Can you use this as standard setting for MEG/EEG traces and can you note somewhere in the traces window which definition is currently chosen?

I added a button that indicates which is the direction of the Y axis in the figure.
If you click on it, it flips the Y axis.

Please update Brainstorm and let me know if it matches your request.

Can you please share it again?

Waiting for you to share the file again for the FieldTrip developers - thanks.

mheers · June 3, 2022, 2:50pm

Hi Francois,

thank you for adding this button!

I uploaded the dataset again. You can find it here:
https://bwsyncandshare.kit.edu/s/faA7smobMdxsZtg

Best regards,
Marcel

mheers · March 6, 2023, 10:17am

Hi Francois,

I am having problems importing fieldtrip structures, which I could import before. I get the following error messesage. I can share example data using a private link.

Thanks for your help!
Marcel

Francois · March 6, 2023, 10:39am

The function ft_datatype_sens.m, which is supposed to be in folder fieldtrip/utilities, is missing on your computer or is not in your path. Your installation of FieldTrip seems to be broken.

Either restore it manually, or install/update FieldTrip as a Brainstorm plugin:
https://neuroimage.usc.edu/brainstorm/Tutorials/Plugins

mheers · March 6, 2023, 11:04am

You were right. I just needed to reinstall fieltrip. Now it works!

Thanks a lot!
Marcel

mheers · April 26, 2023, 7:01pm

Hi Francois,

when importing fieldtrip.mat files into brainstorm, I get an outlier in the MEG channel topography. This outlier is not present in the initial fieldtrip file. Do you have an idea what might cause this error? I can share with you an example data set through a private linke. Thanks for your help!
Marcel

mheers · May 3, 2023, 9:26am

... In addition to this one wrong sensor position, I get the following error:
Error: Wrong number of coils for channel MLF12: Cannot import this file correctly...
Error: Wrong number of coils for channel MLF13: Cannot import this file correctly...
Error: Wrong number of coils for channel MLF14: Cannot import this file correctly...
Error: Wrong number of coils for channel MLF25: Cannot import this file correctly...
Error: Wrong number of coils for channel MLO53: Cannot import this file correctly...
Error: Wrong number of coils for channel MLT21: Cannot import this file correctly...
Error: Wrong number of coils for channel MLT31: Cannot import this file correctly...
Error: Wrong number of coils for channel MLT32: Cannot import this file correctly...
Error: Wrong number of coils for channel MLT41: Cannot import this file correctly...
Error: Wrong number of coils for channel MLT47: Cannot import this file correctly...
Error: Wrong number of coils for channel MLT57: Cannot import this file correctly...
Error: Wrong number of coils for channel MRF12: Cannot import this file correctly...
Error: Wrong number of coils for channel MRF13: Cannot import this file correctly...
Error: Wrong number of coils for channel MRF14: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT21: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT31: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT41: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT51: Cannot import this file correctly...
Error: Wrong number of coils for channel MRT53: Cannot import this file correctly...

... and the channel file shows gaps like this:

Is it possible that you have a look at this problem, Francois?

Thanks in advance!
Marcel

Raymundo.Cassani · May 3, 2023, 1:19pm

Hi Marcel,

Could you share the a sample file that causes the issue, to understand what is going on?
Upload the file somewhere and post the download links here.

Best,
Raymundo

mheers · May 3, 2023, 3:02pm

Thank you for offering your help, Raymundo!

You can find an example file here: https://bwsyncandshare.kit.edu/s/qwpnGFgaHjNH3xt

Marcel

Raymundo.Cassani · May 4, 2023, 3:09pm

Hi Marcel,

After looking at the data, the issues (outlier and blanks in channel file) have the origin:

Rebuilding the coil-channel matrix, since the 'reject' step essentially removed one (or a few) components from the data, so these cannot be anymore recovered. Here are detailed explanation:

github.com/fieldtrip/fieldtrip

Getting the coil-channel correspondence from grad.tra

opened 12:57PM - 01 Jun 22 UTC

closed 01:29PM - 08 Jun 22 UTC

ftadel

question

This problem is connected to a previous issue: https://github.com/fieldtrip/fiel…dtrip/issues/507 A user reported on the Brainstorm a bug related with reading a FieldTrip .mat MEG+EEG recordings file: https://neuroimage.usc.edu/forums/t/importing-fieldtrip-data/33631/4 The link to the dataset: https://bwsyncandshare.kit.edu/s/ytfa2okF3fNNm2m To read correctly this file I had to adjust the Brainstorm reader, which needs to reconstruct the coil-channel correspondence independently from any other mixing matrix (CTF gradient compensation, ICA, SSP...). https://github.com/brainstorm-tools/brainstorm3/commit/a83914bc934a169bd6406d810b383b182584aadc As suggested a few years ago by @schoffelen, I start by reverting the mixing documented in the `balance` field, to get the "original" `tra` matrix. ``` ftData = load('FCD-PA07-LCMV-TC-PCA-combined-sleep-Tr1.mat'); % Plot the original tra matrix grad = ftData.dat.grad; figure; imagesc(grad.tra); % Get the list of montages to undo montageList = {grad.balance.current}; if isfield(grad.balance, 'previous') && ~isempty(grad.balance.previous) montageList = cat(2, montageList, grad.balance.previous{:}); end % Undo montages one by one for iMontage = 1:length(montageList) % Make sure the montage is defined in the structure mont = montageList{iMontage}; if ~isfield(grad.balance, mont) continue; end % Remove all the fields that may cause incompatibility issues % Fix proposed by JMS for importing the HCP pre-processed files for f = {'chanunitnew' 'chanunitold' 'chanunitorg' 'chantypenew' 'chantypeold' 'chantypeorg'} if isfield(grad.balance.(mont), f{1}) grad.balance.(mont) = rmfield(grad.balance.(mont), f{1}); end end % Reverse transformation grad = ft_apply_montage(grad, grad.balance.(mont), 'keepunused', 'yes', 'inverse', 'yes'); % Plot "original" tra matrix figure; imagesc(grad.tra); end ``` The code above produces the following figures: ![image](https://user-images.githubusercontent.com/6920058/171406453-0aacc088-bf80-4358-9700-db08a5452412.png) ![image](https://user-images.githubusercontent.com/6920058/171406561-0fd0d6b8-9bb2-4e42-ac32-fad27fdb805a.png) ![image](https://user-images.githubusercontent.com/6920058/171406579-ef498182-b246-4a87-9840-242225d55a86.png) My problem here: there are still lots of values that are not close to zero at all after inverting all the montages. **Am I not supposed to obtain something very close to a binary matrix simply indicating the correspondence between the channels and the coils? What am I missing?** If this cannot be improved, I would need to reconstruct the coil-channel adjacency matrix by binarizing the "original" tra matrix above. I came up empirically with the following code: ``` diagVal = max(abs(grad.tra),[],2); coilChan = bst_bsxfun(@rdivide, grad.tra, diagVal); coilChan(abs(coilChan) < 0.7) = 0; coilChan = round(coilChan); figure; imagesc(coilChan); ``` **Does this make any sense? Would you have any better suggestion?** Thanks

To fix this issue an heuristic fix was implemented and worked for the previously shared data (Jun 2022). However, the newest shared data is a case where it fails. This is well reported in: Importing fieldtrip data - #6 by Francois

Details:

Outlier (MLF33). In this case, in the (imperfect) rebuild the coil-channel matrix, channel MLF33 is defined by two coils, but they are the two reference coils (in red)

image444×688 27.7 KB
Blanks in channel file In these cases, in the rebuild the coil-channel matrix, those channels are defined by more than 2 coils, thus the position is obtained from the channel position and not from the coils that comprise the channels.
How are electrodes, magnetometers or gradiometers described? - FieldTrip toolbox

@Francois, it seems an improved heuristic is needed at rebuilding the coil-channel matrix

mheers · June 12, 2023, 9:49pm

Hi Francois,
will you have time to look into this? There seem to be problems importing the MEG channel positions from Fieldtrip into brainstorm.

Best regards,
Marcel

mheers · July 5, 2023, 7:51am

As Francois seems to be unavailable: Is there anyone else who is able to reorganize the import of fieldtrip data into brainstorm so that it doesn't change the positions of the MEG sensors?

Many thanks in advance!