[WIP] Update ieeg data example for ECoG

[adam2392]

Reference issue

Addresses: mne-tools/mne-misc-data#3

What does this implement/fix?

Updates plot_ecog example to use EDF data from mne-misc-data.

The ECoG data that currently is used for mne-python demos only contains an ECoG montage, but no "demoable" raw EDF files. As such, most demos showing off ECoG/SEEG functionality sort of just hack together random data points generated (e.g. np.random.random(50, 100) for a 50 x 100 channel recording), whereas it would be nice to have a fleshed out demo w/ a real EDF file to show end-to-end MNE functionality.

Additional information

This also raised issues in mne-bids, where we wanted to make an ieeg example, but had to hack together a soln. since no EDF versions of iEEG data was available in mne.

mne-tools/mne-bids#404

[adam2392]

@larsoner I am trying to update the download path accordingly like you suggested, but I get this error:

Downloading archive https://github.com/adam2392/mne-misc-data/tarball/seeg to /Users/adam2392/mne_data
Downloading https://codeload.github.com/adam2392/mne-misc-data/legacy.tar.gz/seeg (0 bytes)
Error while fetching file https://github.com/adam2392/mne-misc-data/tarball/seeg. Dataset fetching aborted.
  0%|          | Downloading : 0.00/0.00 [00:00<?,        ?B/s]Traceback (most recent call last):
  File "/Users/adam2392/Documents/mne-python/tutorials/misc/plot_ecog.py", line 29, in <module>
    mat = loadmat(mne.datasets.misc.data_path() + '/ecog/sample_ecog.mat')
  File "<decorator-gen-405>", line 21, in data_path
  File "/Users/adam2392/Documents/mne-python/mne/datasets/misc/_misc.py", line 15, in data_path
    download=download)
  File "/Users/adam2392/Documents/mne-python/mne/datasets/utils.py", line 397, in _data_path
    remove_archive, full = _download(path, u, an, h)
  File "/Users/adam2392/Documents/mne-python/mne/datasets/utils.py", line 455, in _download
    hash_=hash_, hash_type=hash_type)
  File "<decorator-gen-3>", line 21, in _fetch_file
  File "/Users/adam2392/Documents/mne-python/mne/utils/fetching.py", line 117, in _fetch_file
    _get_http(url, temp_file_name, initial_size, timeout, verbose_bool)
  File "/Users/adam2392/Documents/mne-python/mne/utils/fetching.py", line 55, in _get_http
    with open(temp_file_name, mode) as local_file:
FileNotFoundError: [Errno 2] No such file or directory: '/Users/adam2392/mne_data/https://github.com/adam2392/mne-misc-data/tarball/seeg.part'
  0%|          | Downloading : 0.00/0.00 [00:00<?,        ?B/s]

Any suggestions?

[larsoner]

Fixed:

• paths for downloading
• colormap (should use one-sided for one-sided data)
• animation
• a few flake errors

Now things run smoothly for me:

I don't think that sphinx-gallery natively supports FuncAnimation objects, but it probably should.

[larsoner]

This example might give us a reason to look into sphinx-gallery/sphinx-gallery#150

[larsoner]

Renders okay, so data downloading etc. works. No animation for now but it at least captures the figure:

https://20034-1301584-gh.circle-artifacts.com/0/dev/auto_tutorials/misc/plot_ecog.html#sphx-glr-auto-tutorials-misc-plot-ecog-py

You should be good to keep developing now @adam2392

[adam2392]

I'm not going to do the animation actually. Don't feel very comfortable w/ my abilities there... But happy to take a look in the future of how this is possible.

The current state attempts to show users how to create some "snapshots" of the data and overlays them onto the FreeSurfer brain.

Some notes: For visualization purposes, I've found that MNE-python has a hard time for iEEG. However, the plot_ecog.py example has always been helpful to me because you can show ecog grids relatively simply, compared to say SEEG, or ECoG strips scattered over both hemispheres. I'm looking into blender to assist w/ my visualization needs on SEEG and more advanced cases where interactive (rotation, zoom, transparency, etc.) are necessary. If there is anything "Pythonable", then would like to make an issue to see if mne-python would like to use it.

[agramfort]

see https://20077-1301584-gh.circle-artifacts.com/0/dev/auto_tutorials/misc/plot_ecog.html#sphx-glr-auto-tutorials-misc-plot-ecog-py

[agramfort]

maybe @drammock want to have a look before merging?

[larsoner]

@adam2392 do you want to restore the animation I had put in? It looked pretty cool and sphinx-gallery/sphinx-gallery#687 is very close to being in (I think) so it should be possible to actually have the animation show up in the docs

[adam2392]

Sure, I think we need to figure out an API for this sort of thing. Really I think it should end up being built into the new 3D brain object that @GuillaumeFavelier is working on, there are already nice interactive controls in there that could be used.

Can you provide links to expand on what this 3D brain is? Sounds interesting and useful.

[larsoner]

#7162 links to a ton if PRs / issues / plans

[larsoner]

  File "/home/circleci/project/tutorials/misc/plot_ecog.py", line 142, in <module>
    epochs = mne.Epochs(raw, events, tmin=0, tmax=raw.times[-1],
  File "<decorator-gen-184>", line 21, in __init__
  File "/home/circleci/project/mne/epochs.py", line 2085, in __init__
    super(Epochs, self).__init__(
  File "<decorator-gen-175>", line 21, in __init__
  File "/home/circleci/project/mne/epochs.py", line 472, in __init__
    _check_baseline(baseline, tmin, tmax, info['sfreq'])
  File "/home/circleci/project/mne/epochs.py", line 1850, in _check_baseline
    raise ValueError('Baseline interval is only one sample. Use '
ValueError: Baseline interval is only one sample. Use `baseline=(0, 0)` if this is desired.

[larsoner]

generating gallery for auto_tutorials/misc... [ 33%] plot_ecog.py
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-circleci'
make: *** [Makefile:49: html_dev-pattern] Terminated

Too long with no output (exceeded 10m0s): context deadline exceeded

Probably need to set the number of frames and/or the interval

[larsoner]

I went to using the Hilbert transform to compute the power in the band, feel free to switch back to CWT if need be, but I doubt it matters much (maybe .filter(None, 10) the gamma-band one if you want it to be smoother?) and the Hilbert is much faster

[larsoner]

• tutorials page
• tutorial

[adam2392]

I went to using the Hilbert transform to compute the power in the band, feel free to switch back to CWT if need be, but I doubt it matters much (maybe .filter(None, 10) the gamma-band one if you want it to be smoother?) and the Hilbert is much faster

Sweet! Had dinner and came back to this sweet thumbnail. That's a neat sphinx trick... For purposes of an example, I think Hilbert is fine.

[agramfort]

@drammock merge if happy. LGTM

[agramfort]

rendered example https://20111-1301584-gh.circle-artifacts.com/0/dev/auto_tutorials/misc/plot_ecog.html#sphx-glr-auto-tutorials-misc-plot-ecog-py

…

[drammock]

A few suggestions to smooth things, but otherwise +1 for merge.

[drammock]

I'll cancel the Travis and Azure tests, we only need Circle CI for the last round of changes.

[larsoner]

I'll cancel the Travis and Azure tests, we only need Circle CI for the last round of changes.

Azure style we'll still want, too

[drammock]

thanks @adam2392 !