Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
Chat of first Brainstorm User Workshop / Q&A, via Zoom (April 29, 2020)
11:07:50 From Kiran : sorry i missed it, whom do we send our questions to?
11:08:09 From Marc (host assistant) : here :)
11:08:57 From Tatjana Liakina : The Brainstorm doesn’t have an approval for clinical use yet?
11:09:34 From Marc (host assistant) : (noted)
11:10:40 From Yvonne Buschermöhle : What is the best way to combine EEG and MEG data analysis in brainstorm? Apparently inverse problems cannot yet be solved combined
11:11:49 From Marc (host assistant) : Noted Yvonne. Let's keep that one for later, maybe after Martin has gone through the basics. :)
11:12:14 From Yvonne Buschermöhle : of course! :)
11:13:05 From Coffman, Brian A : Is there any way to apply SSP or EEG reference projection items to epoched data?
11:13:38 From Eleonora Tamilia : Hello! first of all, thanks for all your great work.
11:13:39 From Alfredo Spagna (he/him) : Another question for later: CONNECTIVITY.
11:13:44 From Eleonora Tamilia : I have a first specific question about the interpolation that is used when plotting, for instance, the power estimated for each scalp EEG sensor over the head or 2D disc. What kind of interpolation is used? I would need one that is less smooth, is it possible to modify it?
11:14:52 From Coffman, Brian A : Is Duneuro better than SimNIBS?
11:16:21 From Marc (host assistant) : I'm noting questions. thanks all!
11:17:26 From Carsten Wolters : SIM-NIBS contains Continuous Galerkin FEM (CG-FEM), while DUNEuro has CG-FEM, but also new FEM approaches such as Discontinuous Galerkin FEM (DG-FEM).
11:17:36 From Renzo Lanfranco : I have another specific question. Brainstorm has the option to run phase synchronization, using PLI. I was wondering if you are planning to make available the weighted phase locking index (wPLI) too any time soon. Thanks! :)
11:18:47 From Marc (host assistant) : There has been recent work on connectivity methods, so it could be. I'll ask!
11:19:13 From Coffman, Brian A : Thanks Carsten Wolters!
11:19:36 From Charly Billaud : Since there are specific protocols for infants, is that also the case for older children (aged 3 to 14 in my case?)
11:19:40 From Camille Bouhour : For time-frequency maps, how can I design wavelets to cover two frequency ranges (11-16Hz and <60-150Hz) ?
11:19:49 From Davide Momi : Thanks for this first of all!! I have question regarding the difference between source computation solutions. Specifically in the tutorial you say that dSPM, current density and sLoreta should give similar results. However I still get very different results specifically with dSPM I rarely get some cortical activation (more are subcortical). Could you please expand on the difference between all these methods?
11:20:15 From Coffman, Brian A : Any plans for functionality for nonhuman primate anatomy/neurophysiology
11:21:02 From Camille Bouhour : For time-frequency maps, how can I design wavelets to cover two frequency ranges (11-16Hz and <60-150Hz) ? — meaning the central frequency and the time resolution parameters, what could be optimal ? Or should I switch to a Hilbert Transform option
11:21:39 From Sylvain Baillet : https://www.nature.com/articles/s41597-019-0242-z
11:22:24 From Coffman, Brian A : Ah, I really meant for anatomical segmentations and source models of EEG.
11:22:26 From Eleonora Tamilia : Another specific practical question: When an imported epoch has a portion within it that is marked as “bad”, that is automatically excluded from any possible process option, isn’t it? Is it possible to run analysis excluding only that “BAD” portion or to automatically re-segment the epoch so that only the bad part is excluded? thanks!
11:24:01 From Marwa El Zein : When reconstructing sources, for EEG in a past study I had computed 1 inversion matrix for each participant based on all the trials - for MEG (my new study) I understand I should compute one for each run as head position moves, but is it correct to then merge the different inversion matrices into 1 for each subject? (I use the matrix to then do trial by trial GLM in each vertex - instead of each electrode)
11:24:09 From Rita Oliveira : Is it advisable to do source normalization before extracting information from the scouts? Thank you!
11:24:11 From Alessandra Pizzuti : I have three question about source estimation for resting state EEG data. The first question is about the estimation of noise covariance: one option given in the brainstorm tutorial is that you can estimate it from a long segment of resting state signal. I was wondering if this segment needs to be pre-processed or must be raw data. The second question is about the SNR value: is the value of 3 (default) good enough if I have resting state data? The third question is about the dimension of inverse operator: If I have an inverse operator unconstrained [3*Nvertices x Nchannels], how can I recombine the three values into one?
11:24:23 From Wired Brain : Hi, I would like to ask about the spectral granger causality. How to consider the maximum Granger model order which is default as 10. Thanks
11:25:02 From Yvonne Buschermöhle : What is the best method to reconstruct auditory source as dipoles (since we have 2 sources)?
11:25:28 From Ottavia : yes
11:28:50 From thomasrd : It seems that the right click does not work on a Mac (usually a Control - click mimics right click but I cannot get the menus to open like this)
11:30:39 From Sylvain Baillet : @Yvonne: the main Brainstorm tutorial uses auditory presentation data. This could be a good start for you to go to.
11:30:47 From Davide Momi : As for the fibres file, how does it work the source computation? Does it take into account where the bundles are propagating?
11:30:48 From Megan Schendel : I've also had the right click issue working on a chromebook to a remote desktop in linux
11:31:06 From Sylvain Baillet : @davide: No.
11:31:35 From Sylvain Baillet : This is for visualization of functional source connectivity results using realistically plausible representations.
11:31:41 From Tomoya Kawashima : Is it possible to select specific sensors for ERF or time-frequency analysis in MEG data?
11:32:14 From John Mosher : To Alessandra Pizzuti's question above: Briefly, (1) the noise covariance should be calculated from the processed data, whatever your processing pipeline (particularly frequency filters and resampling), so that it is the same as the data under investigation. (2) the SNR of 3 in the min norm is from Matti Hamalainen's extensive experience with the results. An amplitude ratio of 3 between the signal and noise matrices in the overall min norm kernel works under a wide range of conditions, somewhat independent of what you might think the true SNR really is. It's more of an imaging resolution control variable than a true SNR. (3) We have that under active discussion that includes the larger problem of multiple time series from a larger scout. You might think of the unconstrained 3 dipoles as a subset of the dozens of time series from a scout (whether constrained or unconstrained). For today, if you need individual time series, you might try switching back to cortically constrained modeling.
11:32:16 From Davide Momi : @Sylvain I see…thank you
11:32:29 From Sylvain Baillet : @Tomoya: absolutely. You can even define montages: https://neuroimage.usc.edu/brainstorm/Tutorials/MontageEditor
11:34:00 From Sylvain Baillet : @Marwa: you are right. For MEG, one imaging kernel per run due to changing head positions. Then all stats (including individual averages) need to be computed in source space.
11:34:00 From Carsten Wolters : Does BrainStorm have an automatized pipeline to segment a pair of MRI such as T1 and T2-MRI? For example through the embedding of SIM-NIBS!?
11:35:21 From fabsarah : The matrix output values are very small…does the transfer rescale (ie. from uV to V)?
11:35:23 From Ottavia : I was wondering if there are quantitavive measures for the source localization accuracy. If not, are you planning to add them in brainstorm? thank you!
11:35:40 From Takfarinas Medani : @Carsten, Yes it does, it calls the headreco process from SimNibs
11:37:06 From John Mosher : @fabsarah, in general, everything in Brainstorm is stored in MKS, so yes, volts, not microVolts.
11:37:34 From Takfarinas Medani : @Brian, SimNibs is also integrated to brainstorm for FEM mesh generation from T1 and T2
11:37:35 From Antonio : Do you plan to increase the documentation regarding how the Brainstorm database work and how the database request should be made according to the different data types, to make the scripting easier?
11:38:58 From fabsarah : @John merci bien!
11:39:09 From Carsten Wolters : thanks, Takfa!
11:39:31 From Marc (host assistant) : @Ottavia This is a very complex question, active topic of research. There are many papers evaluating and comparing various methods. So the short answer would be unfortunately no.
11:39:37 From Francois : @Carsten: Yes, we can call SimNIBS and Brain2mesh to generate meshes from T1+T2: https://neuroimage.usc.edu/brainstorm/Tutorials/FemMesh
11:40:54 From Francois : @Antonio: Please read this Scripting tutorial (https://neuroimage.usc.edu/brainstorm/Tutorials/Scripting) and ask more specifically on the forum what kind of information you could not find
11:41:56 From tale_andrea : is it possible to run ICA not on continuous data but on epoched data? thanks
11:42:17 From lorenzo : Is it possible to create condition folders for different subjects, say “Condition before task” and “after task” in order to perform statistics (permutation analysis) directly on the EEG data? (i.e. functional connectivity measures, pre vs. post)
11:43:00 From AlexChala : There is growing advance in automatic/semi-automatic artifact removal Tools, such as ICLabel for remove non brains components. Autoreject to automatically remove bad epochs and interpolate bad channels. Is there any plans to implements these methods ?
11:43:08 From Solofo : Hello
11:43:42 From Marc (host assistant) : @AlexChala Yes we've been looking into this! Interesting work.
11:43:48 From IMAGING WORKSTATION : how about if we dont have EOG sensors in our original recordings and want to identify/correct for eye artifacts?
11:43:59 From Sylvain Baillet : @camille: you can design an optimized wavelet transform for each frequency band of interest. Alternatively, you can indeed use Hilbert transforms for each band.
11:44:19 From Solofo : How do you take Signal to noise ratio when removing component? How to normalize between many subjects not affected in the same way with artefacts during recording if we want to compare them in terms of ERPs after?
11:45:00 From John Mosher : @Ottavia, @Marc, In the dipole modeling component, we have included in the dipole scanning results the basics of chi-square, goodness of fit, etc, similar to Neuromag's XFIT results, but not as complete. But the overall goal of more quantitatively assessing source localization in all modalities (not just dipole scanning) is part of active research. For now, the community tends to rely on the z-score value in the imaging scans to declare significance.
11:45:23 From fabsarah : (Repeating @Eleonora’s question): when marking segments in raw or epoch data as “BAD, is it possible to get rid of the time series of the data marked BAD, but keep the epoch?
11:45:43 From Sylvain Baillet : @camille: you can also use the default wavelet transform in Brainstorm for both bands. It depends on how much time vs frequency resolution you are seeking.
11:47:16 From Ottavia : @John Mosher thank you!
11:48:54 From KATIA ANDRADE : @Solofo : great question !
11:49:01 From Fran López : What about bad channels? are they removed from all the files from a subject if you remove them from one file? Even if some files are links (raw) and others are already in brainstorm database (epochs)
11:49:12 From Jenkrz : it there a way to interpolate a bad channel ?
11:49:44 From Megan Schendel : yes please talk about bad channels
11:49:50 From Francois : For questions related with bad segments, please search the forum, there are plenty of specific cases discussed. All your questions have already been asked.
11:49:58 From IMAGING WORKSTATION : would you recommend eye artifact identification before or after epoching?
11:50:12 From Marc (host assistant) : Before according to the basic tutorials
11:50:32 From IMAGING WORKSTATION : thanks
11:50:40 From Francois : interpolate a bad channel : process "Standardize > Interpolate bad electrodes"
11:50:51 From Jenkrz : thanks
11:51:24 From Francois : Bad channels are never removed from the recordings: https://neuroimage.usc.edu/brainstorm/Tutorials/BadChannels
11:52:46 From IMAGING WORKSTATION : follow up to kristina question, i would like to ask the same which one i should use between AF3 and AF4?
11:53:49 From eleononora tamilia : Will you make public all this written discussion? I lost connection and loose previous replies. Thanks!
11:56:43 From Francois : Know what, sorry?
11:57:40 From Davide Momi : Not sure if this is off-topic but I have question regarding the difference between source computation solutions. Specifically in the tutorial you say that dSPM, current density and sLoreta should give similar results. However I still get very different results specifically with dSPM I rarely get some cortical activation (more are subcortical). Could you please expand on the difference between all these methods?
12:00:09 From John Mosher : @Davide Momi, dSPM and sLORETA begin with the exact same min norm solution, they just scale the answer somewhat differently (measured noise covariance vs theoretical data covariance). Hence our general statement that they should give similar results between the two.
12:01:23 From Alessandra Pizzuti : @John Mosher. Thank your for your reply. 1) In your opinion, do you recommend to use this option or the second one proposed in your tutorial?(the one that suggests to use identity matrix) 2) can I run simulations ranging SNR values from 3 to 10 and then use the best solution as SNR value?
12:04:34 From Marc (host assistant) : @Davide, this is a more conceptual question difficult to answer in this demo. I would recommend reading the relevant literature about the methods to understand the differences. Different people have different "favorite" methods that they'll recommend. :)
12:05:10 From Sylvain Baillet : @eleononora: yes - on Twitter and FB
12:05:31 From Rita Oliveira : Regarding estimating the noise covariance for computing the source solution, we have to choose an option to remove the DC offset. Could you comment on the differences between the two methods and if one is more suitable than the other?
12:05:36 From Fran López : Are there any plans to allow processing of multiple simultaneous subjects in parallell using multithreading? thanks :)
12:06:10 From Marc (host assistant) : @Fran Yes!
12:06:20 From Filippo Gambarota : What is the best way to learn brainstorm functions to write our scripts directly? without the GUI
Thanks!
12:06:57 From John Mosher : @Alessandra Pizzuti (1) I would used the identity matrix only in theoretical studies. At least in MEG, the noise is too colored to simply use an indentity matrix, so we really need to get some sort of sample of the noise in order to better know the background condition. (2) I'm not a big fan of tinkering too much with the SNR value in the min norm solution. That would make comparisons between sets more difficult. But I'm not too experienced with adjustments to the min norm kernel, so I'm not the best expert for that approach.
12:08:13 From Takfarinas Medani : @Filippo : you can start from this tutorial https://neuroimage.usc.edu/brainstorm/Tutorials/Scripting
12:08:34 From Marc (host assistant) : @Filippo But the GUI is one of the strengths of Brainstorm! ;)
12:10:29 From Marc (host assistant) : @Rita You should apply the same pre-processing to your noise (for the covariance) than to your data, including DC offset. (Not exactly an answer to your question - I don't remember the 2 options you mention.)
12:11:28 From Alessandra Pizzuti : @John Mosher. Thank you again! And what about EEG? I am working on 256 channels EEG data.
12:12:22 From Steven Beumer : @Rita you can let brainstorm remove the DC offset over the whole file, though EEG baseline tends to drift over time and that will introduce artificial covariance. The block by block option removes DC offset over a block of samples and then the next etc., not the whole recording and thus will be a bit slower, but the data should be detrended better
12:12:43 From Filippo Gambarota : Thank you very much!
12:13:02 From Tatjana Liakina : Will it output intermediary files in generated after each of the steps of the pipeline?
12:13:27 From Martin Cousineau : Yes, all intermediary files will be saved as well!
12:14:53 From Francois : Discussion about smoothing: https://neuroimage.usc.edu/forums/t/plot-actual-values-on-2-d-disk/4541/6
12:15:45 From John Mosher : @Alessandra Pizzuti EEG has the challenge that there is no "empty room", since the array only measures when it's active on the head! So you have to instead find a time region that is considered to be "baseline", which may be at a completely different time in the data file. As also mentioned in @Rita answer above, you should ensure drifts are removed. And because each electrode has a somewhat different noise level, this measured noise covariance is much better than an identity matrix assumption.
12:16:57 From eleononora tamilia : is it possible to mark a channel as bad only for a portion of data?
12:20:07 From tale_andrea : @eleononora Yes very good question... any advice with brainstorm ? thanks
12:20:19 From Alessandra Pizzuti : @John Mosher. Maybe I did not mention it before, but I do not have evoked potential so it is difficult to find a baseline since I have resting-state recording! Thank you again.
12:25:21 From John Mosher : @Allesandra Pizzuti, yes you have the most challenging problem. But, as in fMRI, you could find some contrast in your resting states, such as between sleep and awake. The "beamformer" class of algorithms should work directly with the data, but it has it's own issues of sensitivity to the accuracy of the source model, and in EEG, the lead fields are more difficult to accurately calculate. So, interesting problem, with more research to do.
12:26:47 From Konstantinos N : Any update on Calcium Imaging integration?
12:27:07 From Martin Cousineau : Still work in progress, stay tuned! :)
12:27:38 From Marc (host assistant) : I personally would recommend wPLI vs PLI and I've used it before, so I'll put it on my to do list. ;)
12:29:51 From IMAGING WORKSTATION : is it recommend to to do source localization with 14 channels data? I have 14 channels data.
12:30:32 From John Mosher : @all, a plug for a new paper showing Brainstorm's excellent performance among other research beamformer software:: https://www.sciencedirect.com/science/article/pii/S1053811920302846
12:31:07 From SERGIO OSORIO GALEANO : wPLI would be an asset. As it reportedly minimizes spurious connectivity due to volume conduction, wPLI seems to be the way to go for anyone attempting do Functional Connectivity on EEG data.
12:31:13 From Marc (host assistant) : @IMAGING I believe most answers would be no. You could only find 14 independent signals/sources so not really amenable to source imaging.
12:32:34 From IMAGING WORKSTATION : thanks @marc
12:33:17 From Paola : And source localization with 64 channels?
12:33:37 From John Mosher : @IMAGING, even 23 channel EEG data (classic 10-20 in epilepsy) yields poor localization results. See for example: https://www.sciencedirect.com/science/article/abs/pii/001346949390043U, where we show error bounds for a variety of sensor densities.
12:33:51 From Marc (host assistant) : I think this is doable, with 64.
12:33:58 From Paola : thanks
12:34:13 From John Mosher : @Paola, yes 64 whole head coverage should be acceptable.
12:34:44 From SERGIO OSORIO GALEANO : Just quick question, how about subcortical sources?I understand you have now included subcortical anatomy. When and how do you suggest to project sources on subcortical structures? Any paper published to date using this BST implementation?
12:36:01 From Solofo : Thanks for answers
12:36:01 From IMAGING WORKSTATION : My name is Rabnawaz. I did not tried to changed the name here. sorry for that
12:36:47 From IMAGING WORKSTATION : thanks for detailed answere
12:37:30 From Martin Cousineau : Refer to this tutorial page for source modeling on deep structures, including some paper references: https://neuroimage.usc.edu/brainstorm/Tutorials/DeepAtlas
12:38:10 From Isabelle Arseneau-Bruneau’s iPad : :)
12:38:25 From Sylvain Baillet : Deep sources: https://www.nature.com/articles/ncomms11070
12:38:48 From Saeed : Hi,
I am trying to perform "dipole scanning". Right now I have about 200 individual dipoles generated for each of the epileptic-spike-epochs. Currently the procedure I follow is to carefully open the parent folders for each of the dipoles and select each of these dipoles (200 dipoles!) and finally right click and merge them.
Is there easier way to directly select them?
Thank you
12:40:45 From John Mosher : @Saeed, in the Brainstorm tutorials, advanced, for the Neuromag phantom, you'll see a script where we individually fit 32 dipoles, then merge them in one bundle
12:41:51 From Rita Oliveira : Are you planning on allowing to simulate the recordings from sources with other methods rather than MN (for the model evaluation)? Because now it only works for MN, right? Thank you!
12:41:57 From Francois : @Saeed: Merging dipoles: Use the database search to get all the dipoles files then select them all and right-click>Merge
12:43:40 From SERGIO OSORIO GALEANO : Thanks a lot for the link. I was thinking more about the technical aspects (of subcortical source projection). To be more specific, could I attempt it with a 64 EEG channel set up or is this something that would be restricted to MEG data only?
12:44:12 From Sylvain Baillet : In principle, EEG is even more sensitive to deeper sources than MEG - so, yes!
12:44:45 From Saeed : @Francois:
Yes but the problem is that each of the dipoles are in their specific folders, and I need to open those folders one by one..
You suggested in the forum to use the process file-> delete file-> delete folder
but it seems that this process deletes both folders and files within them.
Please check this:
https://neuroimage.usc.edu/forums/t/problem-selecting-specific-files/17571
12:45:40 From Martin Cousineau : Unmute please
12:46:57 From Marc (host assistant) : @Sergio you also have to design your study to be able to reach the SNR for your deep sources, typically many more trials!
12:47:25 From Songhee Kim : Sorry if it is a redundant question. Will this meeting be available for viewing later?
12:47:40 From SERGIO OSORIO GALEANO : Wonderful! thanks a lot!
12:48:41 From SERGIO OSORIO GALEANO : Another question. How can I import other atlases that are not available in bst by default? I have tried in the past to import functional atlases unsuccesfully.
12:48:47 From Marc (host assistant) : @Songhee yes if the quality is ok
12:49:13 From SERGIO OSORIO GALEANO : Thanks in advance and hope we can make the BST workshop in Santiago work! : )
12:49:40 From Martin Cousineau : @SERGIO: Open a cortex surface file, and go to the Scout tab -> Atlas -> Load atlas
12:50:56 From HaydeeGL : I would be very interested on the Stats for Sources
12:51:10 From SERGIO OSORIO GALEANO : Certainly!
12:52:16 From Sahana Nagabhushan : For microstate analyses, do you have an option for extracting resting-state microstates? Do you have the option to use the meta-criterion to select the optimal number of microstates?
12:53:06 From HaydeeGL : Thanks. I will check it.
12:53:10 From Marc (host assistant) : @Sahana Good question, but I don't think there is any micro-state analysis implemented in Bst.
12:53:19 From Francois : Microstates: https://neuroimage.usc.edu/brainstorm/Tutorials/MicrostatesCena
12:53:29 From Marc (host assistant) : I stand corrected! :)
12:53:30 From Kiran : Will the recording of this session be available for viewing later?
12:53:31 From Martin Cousineau : Statistics: https://neuroimage.usc.edu/brainstorm/Tutorials/Statistics
12:54:04 From Kiran : Thanks :)
12:54:17 From eleonora tamilia : When drawing a scout on the cortex, if i'm not wrong, you can open the MRI and draw it slice by slice. In the volume that's not possible. Is that something you are working on? or would be interested in doing?
12:55:04 From Francois : Microstates: Not working at the moment, but please contact them directly, it would motivate them to work on it :)
12:55:21 From Sahana Nagabhushan : Will do! Thank you!
12:56:07 From Francois : @Saeed: Use the new search tools (https://neuroimage.usc.edu/brainstorm/Tutorials/PipelineEditor#Search_Database) with the option "Hide parent nodes" selected
12:56:36 From eleonora tamilia : any chance the conf. volume will be added as well?
12:56:52 From IMAGING WORKSTATION : is it possible to do some prcocessing (e.g. connectivity, TF) and then do stats in brainstorm?
12:56:54 From Francois : @Martin: Just realized that this new option is not documented yet the tutorial. Could you add this?
12:57:57 From Francois : Scout tab, menu Scout > Edit in MRI
12:58:12 From Carsten Wolters : Thank you, John and Sylvain, for the AEF/AEP dipole fitting explanations! Yvonne and I will try it.
12:58:12 From John Mosher : @Elenora, I want to add, both Matti Hamalinen and I have been talking between us how to resurrect that older code and bring confidence volume back into both MNE and Brainstorm.
12:58:12 From eleonora tamilia : not sure how to raise my hand
12:58:12 From Francois : This is difficult to use
12:59:31 From Francois : Volume scouts: https://neuroimage.usc.edu/brainstorm/Tutorials/TutVolSource#Volume_scouts
12:59:48 From tale_andrea : what's the most reliable way to downsample the number of vertices in the (default/individual) cortex to about 5000 ?? (thanks again
13:03:22 From Christina : Thank you so much, this is a very useful workshop! How can I know the exact time of the next one?
13:04:01 From Konstantinos N : I think it pops up after you select the number of vertices
13:05:13 From Alfredo Spagnav (he/him) : Thanks for this, everyone. Always great to follow the latest on BST. Keep it up and let’s meet again soon;
13:05:36 From eleonora tamilia : @John that would be great. thanks
13:06:04 From Marc (host assistant) : @Christina it would be posted on https://neuroimage.usc.edu/brainstorm/Training
13:06:07 From Carsten Wolters : Thank you very much to the whole team for this very good workshop!
13:06:07 From KATIA ANDRADE : Thanks for this very interesting webinar!
13:06:11 From Kiran : Thanks to you too. It was great meeting up!
13:06:13 From Rita Oliveira : Thank you for the workshop!
13:06:16 From Vardan Arutiunian : Thank you very much for this meeting!
13:06:18 From Filippo Gambarota : thank you very much!
13:06:21 From GAURAV SINGH : thank you
13:06:22 From Francesco Di Gruttola : Thanks everyone!
13:06:30 From davidparker : Thank you!
13:06:30 From Oscar : Thank you for putting this together! :)
13:06:36 From IMAGING WORKSTATION : Thanks for this opportunity
13:06:41 From H@Bremen : Thank you very much!
13:06:44 From Charly Billaud : Thanks for this. It is especially useful to have these online sessions when in Europe.
13:06:45 From Isabelle Arseneau-Bruneau’s iPad : thanks for organizing this :)
13:06:48 From mihai : Thank you for offering this workshop. Best wishes to everyone. Stay healthy!
13:06:52 From Thea Giacomini : Thanks for this useful workshop!
13:06:53 From Maëva Michon : Merci beaucoup!
13:06:59 From GAURAV SINGH : thank you very much for every one for meeting
13:07:08 From Zaira Romeo : Thank you very much!
13:07:12 From SERGIO OSORIO GALEANO : Thanks a lot to all the organizers. This was a great session.
13:07:15 From naazaiez : Thank you!
13:07:19 From eleonora tamilia : thanks!!
13:07:19 From HaydeeGL : Thanks!!!!
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment