Dear Brainstorm community,
My apologies if this has been reported already. Recently I've experienced a problem when visualizing EEG recordings and SSP projectors. This problem started after I ran the last update of Brainstorm. The problem is the following. When I close the window showing the SSP projectors of a dataset and next I try to open the recordings of a new dataset, Brainstorm shows me the image of the projectors instead of the EEG recordings. When I close this image and I try to open the recordings again a screen with flat SSP projectors is shown.
This is how it looks when I'm working on the recordings:
And this is what happens later:
The only way to make it work again is closing Brainstorm and starting it again.
My apologies if the solution is too evident, but I haven't find the way to find it. I hope you can help me with this issue.
All best,
Simon
Hi Simon,
In the Record tab, you can change the current montage from “ICA components” to “All”.
This should also get reset if you close the figures using the big cross “Close all figures” on the top-right corner of the Brainstorm window (not the one that closes the program).
Btw, ICA is usually a better option than SSP for EEG. Tutorials coming soon…
Francois
Hi Francois,
Many thanks for the response. The problem persists, though. When I change to "All", the channels appear flat.
Beforehand, many thanks for the solution.
All best,
Simon
My apologies, I left something unexplained. The problem I have if I use the tab “close all figures” is that I have to re-open the recordings every time I want to see the projectors. This is because when I close the image of a projector and I open a second one, the same flat activity is shown.
All best,
Simon
Hi Simon,
After look again at your screen captures, I realized that something was wrong in your SSP projectors: you should never see a scale showing “NaN uV” on the right of the screen.
I would recommend you delete ALL the SSP projectors from this file, and try to recompute new ones (after clicking on the button “Close all”).
Or try with ICA instead, which is usually better indicated for cleaning EEG (use it on continuous recordings, unlike the SSP).
Let me know if you can reproduce exactly the same behavior multiple times, there is probably some special unstable cases that are not correctly handled in the interface.
I don’t understand how you manage to have either the EEG recordings or the SSP components on the screen. In the typical workflow, you would always see the two simultaneously.
Open the recordings, then compute new projectors or select existing ones with the menu “Select active projectors”: it should open new figures for everything, and never replace the first figure with the EEG.
See the attached screen capture as a typical example. Can you tell me how you get to different configurations of figures?
I also noticed that the button [AS] is not selected in your figures (“Auto-scale”): clicking on it should adapt automatically the scale to the data in the figure (but maybe it doesn’t work if you have NaN values anywhere in your recordings).
Cheers,
Francois
Hi Francois,
Thank you for your reply. I know how to work with ICA, but in this occasion I would prefer to use ssp, as I have to analyse ~400 datasets, ssp is faster and I don't run a high risk of taking off non-artefactual components, as it might happen if I inspect ICA components too quickly. The other advantage is that ssp reduces the absolute noise value by a huge factor, which might be an advantage, as I'm analysing datasets of several experiments, and these might vary considerably in recording procedures and referencing options.
However, the main problem is not related to this. The problem is related to Brainstorm's GUI I'm afraid, because it persists when I use ICA.
This is how it looks when I select the ICA componentes generated by Jade:
However, as soon as I close the window and I try to re-open the recordings or I open another recording, this happens:
As you recommended me, I closed all the windows using the big "X" at the right corner in the main GUI. Although this solves the problem momentarily, the problem persists and I have to close all windows again. It ins't a big deal to work this way, it is just slower. But I imagine that you might want to solve this problem in the future. This has happened to me in a computer running Windows with Matlab 2015 and in a different computer running Linux with Matlab 2014. Could it be related to the last Brainstorm's upgrade? Could this improve if I use Matlab 2016?
Many thanks fr your help.
All best,
Simon
By the way. The Nan value you is exactly related to the problem. The image I printed is a frozen image, is not showing real activity and I cannot work with it. This frozen images are occurring every time I re-open recordings, projections or components. My apologies if I haven’t explained well the problem before. You can see a similar example in the second image of my previous post. That image appeared when I double-clicked the recordings after closing the images of ICA componentes and recordings that appear in the first image I’ve posted. So Brainstorm is basically opening the wrong image.
In the example you show here, are you really expecting to remove almost all the ICA components, and keep only component #9?
If so, it’s normal that you have flat signals for all the ICA components, and for most of the signals.
If you select the first 5 or 6 first SSP components (to remove them), expect to have almost no signal left.
I’ll try to understand why this causes errors, and make the interface more robust for exploring the noise (instead of expecting mostly the full signal).
Hi Francois,
I printed that image just as an example, I arbitrarily selected as much components as possible (in any case I calculated just 20 components to make a fast analysis as an example and they were not removing much).
The real problem of the flat signal is that it is not signal at all. Brainstorm is not printing an image of the actual recordings when the problem occurs, but just a frozen image of flat channels. I really don’t know how to explain the problem better, I’m sorry.
All best,
Simon
Hi Simon,
I’m not sure I understand why you would remove more than one or two SSP components, but let’s try to fix this issue.
I could not reproduce this behavior. Could you send me one of the files so I can test it on my side?
- Pick a continuous file with the SSP or ICA decompositions that cause this problem, right-click on the link > File > Show file in explorer.
- Upload the corresponding .bst somewhere
- Send me the link in a separate email (click on my username on this forum)
- Describe step by step what I have to do to reproduce the error (succession of all the menus I have to click on)
Thanks
Francois
Hi Francois,
My apologies for the late response. I have installed Matlab 2016 and updated Brainstorm, but things went worst.
Now Brainstorm gives me this error when I try to plot the polarity maps:
Of course I've checked that all channels are good, and etc. I'll send you one of the datasets I'm analysing. However, I've tried this with datasets from other experiments and the same happens.
The steps I'm following are: Review Raw file, Band-pass and Notch filter, detect heartbeats and eyeblinks, ssp both.
thanks for the help. All best.
Hi Simon,
I looked at the file you sent me.
There are two this missing at the beginning of your pipeline:
=> Change the type of the EXG* channels (this is not EEG, is it?)
=> Setting the position of the electrodes
You get the last error message you reported (cannot produce the topography) because you haven’t set the electrodes positions.
To make this more explicit, I will change this error message to: “Nothing to display: All the ‘EEG’ channels are marked as bad or do not have 3D positions.”
Then, I’m not sure you need the notch filter or the band-pass filter…
Have you looked at the power spectrum? There is no 50/60Hz noise in your recordings (battery-powered EEG?)
What is the frequency range of interest in your analysis?
What channels are you using for detecting blinks and heartbeats?
I don’t see an ECG in your file, therefore I’m not sure what heartbeat detection or SSP you are talking about.
And in most cases, there is no correction heartbeats to do in EEG.
For blinks, I would definitely recommend an ICA approach instead.
I understand you tried to reproduce exactly the analysis pipeline that is illustrated in the tutorials. However, you have to adapt it to the type of information available in your file.
The EEG/Epilepsy tutorial is probably a better example for you to start with:
http://neuroimage.usc.edu/brainstorm/Tutorials/Epilepsy
Cheers,
Francois
Thanks for your answer. I will try to adapt the analysis and see what happens.
Hello,
I have encountered a similar problem while running through the tutorials. After completing the notch filter tutorials, and beginning on the SSP tutorial, now the time series plots all appear flat. And if I try and zoom in to increase resolution, the time series plots for each channel start migrating off screen and cannot be viewed. Topography figures all appear normal. Power spectrum plots appear normal. I’m using Matlab 2013. I’ll try restarting the tutorials and see if the problem persists. I don’t recall the issue occurring prior to the notch filtering tutorial.
I appreciate your thoughts when you have time.
Kindest Regards,
Jared
Hi Jared,
Please make sure that the following options are correctly set:
- In the Record tab, the button [DC] should be selected (it removes the average of the signal over the loaded time window, therefore centers the signals on zero in your figure)
- In the time series figure, try selecting the [AS] button (auto-scale).
If you’re getting any trouble with the amplitudes in the file, delete the SSP projectors and start over. If you suspect the notch filter step, I’d recommend you delete the link to the continuous file, and start over from linking the continuous file to your database (tutorial #4).
If you can reproduce this behavior, please post here a screen capture showing your database explorer and the figure with the wrong amplitudes.
Cheers,
Francois
Thank you very much, Francois.
I will give these tips a try and post the outcome shortly.
Cheers,
Jared
Hi Fancois,
I started over from scratch and the problem is gone. However, I have encountered a new issue. I am now on Tutorial #17 and am trying out the 3D sensor cap. The cap does not display any topographical information. It just remains gray, regardless of what time point I select. 2D sensor caps function normally. Any thoughts on this? Thank you so much.
Best,
Jared
Hi Francois,
Looks like the 3D sensor cap issue was due to the OpenGL hardware acceleration. I turned it off and now the 3D cap displays correctly. Thank you.
Best,
Jared
Hi Jared,
In this case, there is a problem with the support of your graphics card.
You can try to update the drivers for your graphics card, or install a new Matlab version.
Francois
