sharing EEG/ERP data on platform that supports MNE

✉️ Mailing List Archive (read-only)
1

Hi All,

We are trying a small scale EEG data sharing experiment by posting one
of our schizophrenia ERP study data sets on kaggle:
https://www.kaggle.com/broach/button-tone-sz

The idea is in part to see what novel analysis approaches users might
post online (both patient vs control and task condition classification
might be of interest), and part to make analyses transparent (which
hopefully makes them easier to reproduce!).? I know there are EEG
specific platforms for sharing, but this is on a small scale (<=500MB
could be shared on kaggle when I posted this, but they just upgraded
that to 10GB).? The kaggle platform also supports MNE python in jupyter
notebooks, so it may be of general interest to this group.? I'm
interested in any feedback you might have about this kind of approach
relative to other data sharing options for this type of data.

thanks,
Brian

2

Dear Brian,

thank you for uploading your data to Kaggle. It's really great that
researchers are sharing their datasets. Kaggle is a nice platform, since
you can quickly open a kernel and take a look at the data.

For bigger datasets, such as the raw EEG, one option is physionet.org
Also, there is currently a movement underway to standardize the
file/folder structure when sharing neuro datasets: BIDS. Here is for
example the BIDS-EEG proposal:
https://docs.google.com/document/d/1ArMZ9Y_quTKXC-jNXZksnedK2VHHoKP3HCeO5HPcgLE/edit?usp=sharing

Personally, I don't care that much where and how people share their
data, as long as it is out there :slight_smile: Thanks for doing it!

best,
Marijn.

3

Hi Brian,

I think that's great. I just started processing the IEEE Brain dataset so
you've left me only torn:) Hopefully, EMEG BIDS data can be hosted by
OpenNeuro too but if nothing else available from datalad where ever it is
stored.

Cheers,

-Morgan

4

Dear MNE experts,

for the analysis of some data for which we do not have individual MRI scans, we plan to use the Montreal data set colin27 instead. The dataset passed the freesurfer segmentation and although freesurfer has estimated the folded cortical surface for colin we would avoid using it as our participants most likely have different cortical foldings. Instead, we would prefer using a smooth surface positioned so that it covers the superficial parts of the cortex and ignores deeper structures. One solution is to take the inner skull surface, shrink it by e.g. 10 mm or so. Of course, we will not restrict the estimated activity to the surface normals of this non-folded surface. However, for the inferior parts of the cortex this solution is not optimal as it includes the cerebellum.
Hence, our question: Has somebody else used such a non-folded source space? Could you create a single surface source space which covers the outer volumes of the cerebrum excluding the cerebellum? Any suggestions are welcome.

best wishes and thanx in advance,
Burkhard (& Narjes)

5

hi Burkhard,

did you consider working in the volume with a grid source space ? It
should be way easier.

Alex