Request for Automagic EEG Pipeline with BIDS Dataset

[Magicdjez]

Edit : I just change the quality of the images
Hello,

I'm writing to you because I want to preprocess a large EEG dataset from children in an automated way, and your tool seems to be a good candidate for this. I’m currently running some tests with it, but I’m encountering a few issues and wanted to get your input.

The project is in BIDS format and includes four tasks per participant that I’m trying to preprocess automatically. Below are the parameters I set for the project during my tests.

Here is what the EEG folder for a participant looks like.

ran tests on a small subset of participants to see what happens, and I’m currently facing the following situation. Here is the output Automagic

Looking at this, I have a few questions. I should mention that the same behavior occurs for all participants — I’m just using this one as an example. The first thing I noticed is that all the EEG files are roughly the same size, whereas in my case, the resting state file should be significantly larger than the others. Is this normal?

My first intuition was that the pipeline might be preprocessing the wrong files or looping over the same one. So I checked the outputs, and two things surprised me. First, the .png files are all identical across the different tasks, which I found odd. Since all four tasks were completed in a single session, I can imagine the bad channels being the same, but the fact that the filtered EEG data and ICA-corrected clean data are also identical was unexpected.

After seeing this, I checked the .txt files and noticed that the correct files are being processed and that the sanity check scores are different.

So I was wondering if all this behavior is expected, or if the issue might be due to misconfiguration of my dataset or the input parameters for the pipeline.

Best regards,
Djeser

[ksgfan]

Hi Djeser,

That's a bit odd, it looks like the same file is being processed multiple times. It's hard to say why that's happening without checking it directly. Would you mind sending me the files of one subject (e.g., via Dropbox) so I can try it out and debug on my end?

Also, just a quick note: the quality scores might differ slightly even for the same file, because ICA decomposition involves some randomness, so the results aren't always exactly 1-to-1 the same.

Best,
Dawid

[Magicdjez]

Yes, that’s exactly what I’m thinking.Here is the link to one participant. Let me know if you need anything else.

Thanks for your help!

https://www.dropbox.com/scl/fi/qn5ceztf15t2zvttlxqut/sub-NDARAC904DMU.zip?rlkey=b1y67kzxxtohvwg71xkjmw5qc&st=3g6jq5f4&dl=0

[ksgfan]

I managed to preprocessed the files correctly.

1. Please uncheck the 'Data in BIDS format'
2. Provide channel location file in .sfp format (you can find the file in automagic/matlab_scripts/GSN-HydroCel-129.sfp)
3. Add the channel location file type (.sfp)

I dont know, what is the issue with BIDS yet, but I will investigate in the next days.

Best,
Dawid

[Magicdjez]

I will try this thank you very much !!

have a good day

[Magicdjez]

Hello,

I just tried using your parameters, but I encountered the following issues. I'm not exactly sure what they mean:

Here is the path structure for my project:

Please let me know if I’m missing something or if you need more information.

Thank you for your help!

Best regards,
Djeser

[ksgfan]

Please go to automagic-master/mastlab_scripts/bva-io-master/readbvconf.m
Set a break-point at line 59
Run automagic and wait until it reaches this line of code
Check, what values the following variables have:

**fid** = fopen(fullfile(**pathname**, **filename**), 'r');
**raw** = textscan(fid, '%s', 'delimiter', '');

thanks, Dawid

[Magicdjez]

Hello, sorry for the delay — I had to switch to running things locally because my server is undergoing maintenance for several days.

I ran the script with the breakpoint, and I think I’ve identified where the problem comes from. Here is the folder structure I’m using to access my data:

And here are the paths the script is trying to use to open the file:

As a result, fopen returns -1:

However, when creating the project, Automagic does detect my participants correctly. Is there a way to handle this without modifying the structure of my BIDS-formatted data?

Edit : I try to change my path and automagic doesn't see my file

Thanks again for your help,
Best
Djeser

[ksgfan]

The paths are created in automagic-master/src/block.m. Go to line 1312 and investigate:

                parts = strsplit(self.sourceAddress, filesep);
                hdr = parts{end};
                parts = parts(1:end-1);
                path = strjoin(parts, filesep);
                [~, data] = evalc('pop_loadbv(path, hdr)');

[Magicdjez]

I checked, and right from the beginning of the script — in the object itself — the eeg folder doesn’t appear:

and the path doesn't appear when I create the projet.

I also noticed that the rest of the file paths are built based on this missing information.
Here is the dataset I'm using locally — it includes the same participant as yesterday, in case you want to test any parameters that could help avoid this bug.

https://www.dropbox.com/scl/fo/2fupv6vhjvdnkuud6buwf/ALHsPfnWs9Le-Io76q4am00?rlkey=9z70uy6s81vcyqbp1rlocyjto&st=fztl9q0b&dl=0

Let me know if you need anything else.

Best regards,
Djeser

[ksgfan]

i think i found the bug.

go to automagic-master/src/block.m to line 1416 and replace it with self.sourceAddress = [self.project.dataFolder, self.subject.name, slash, relAdd, self.fileName, self.fileExtension];

The relAdd was missing.

Let me know if it solves your problem.

[Magicdjez]

I added the relAdd and set a breakpoint — here are the results. I think we’re on the right track. relAdd is empty

[ksgfan]

weird

please add the restoredefaultpath in the runAutomagic script. Then close the GUI and rerun the script.

%% start Automagic
restoredefaultpath;   % <--------
mainGuiFile = 'mainGUI.m';

addAutomagicPaths();
[hFig] = loadingScreen();
run(mainGuiFile);
close(hFig)

[Magicdjez]

The restoredefaultpath was already there

[ksgfan]

I have the same folder structure as you have, following params and it runs smoothly:

Do you notice any difference to your setup? Are you on mac or windows? I am on mac.