CSD might break montage

✉️ Mailing List Archive (read-only)
1

External Email - Use Caution

Hello all,

I am applying CSD to my epoched data and want to compare the
distributions of the ERPs on the scalp before and after surface laplacian.

epochs_csd = mne.preprocessing.compute_current_source_density(epochs)

I use : mne.viz.plot_compare_evokeds([epochs.average()], picks='eeg',
axes='topo')
To plot before CSD, everything works fine but then I use :
mne.viz.plot_compare_evokeds([epochs_csd.average()], picks='csd',
axes='topo')
And then the channel order is preserved but the cartoon head is shrinked
to the lower left of the figure (see attached) making it impossible to read.

I am assuming that the montage is lost in the csd transform as
epoch.info prints me :
<Info | 11 non-empty values
  bads: []
  ch_names: Fp1, AF7, AF3, F1, F3, F5, F7, FT7, FC5, FC3, FC1, C1, C3,
C5, ...
  chs: 64 EEG
  custom_ref_applied: True
  dig: 712 items (3 Cardinal, 64 EEG, 645 Extra)
  file_id: 4 items (dict)
  highpass: 0.0 Hz
  lowpass: 40.0 Hz
  meas_date: 2019-06-06 11:46:59 UTC
  meas_id: 4 items (dict)
  nchan: 64
  projs: []
  sfreq: 1000.0 Hz
>

and epochs_csd.info :

<Info | 11 non-empty values
  bads: []
  ch_names: Fp1, AF7, AF3, F1, F3, F5, F7, FT7, FC5, FC3, FC1, C1, C3,
C5, ...
  chs: 64 CSD
  custom_ref_applied: True
  dig: 712 items (3 Cardinal, 64 EEG, 645 Extra)
  file_id: 4 items (dict)
  highpass: 0.0 Hz
  lowpass: 40.0 Hz
  meas_date: 2019-06-06 11:46:59 UTC
  meas_id: 4 items (dict)
  nchan: 64
  projs: []
  sfreq: 1000.0 Hz
>

Is there a quick workaround ?

Best,
Gabriel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo_csd.png
Type: image/png
Size: 68233 bytes
Desc: not available
Url : http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200624/cf4e1a22/attachment-0001.png

2

External Email - Use Caution

The topomaps in the example don't show this problem, at least for
plot_topomap:

https://mne.tools/dev/auto_examples/preprocessing/plot_eeg_csd.html#sphx-glr-auto-examples-preprocessing-plot-eeg-csd-py

First, it's worth trying latest
<https://mne.tools/dev/install/advanced.html#using-the-development-version-of-mne-python-latest-master>
`master` branch rather than 0.20.x to see if we have fixed it already.

If that doesn't work, can you see if you can reproduce on the `sample`
dataset by modifying the example above? If not, there might be something
about your data in particular -- if you can turn this into some tiny script
(probably gist <https://gist.github.com/>) to demonstrate the problem and
share your data someone can take a look.

Eric

3

External Email - Use Caution

I was on the dev branch after the commit #7863 on CSD on github, but
even when updating to the master branch with the link you provided I
still have the same error.

I tried with the example in
https://mne.tools/dev/auto_examples/preprocessing/plot_eeg_csd.html#sphx-glr-auto-examples-preprocessing-plot-eeg-csd-py
Topomap works (as for my data actually) but when I use
mne.viz.plot_compare_evokeds([evoked_csd], picks='csd', axes='topo')
I have the same error. So it seems to be specific to plot_compare_evokeds

I posted a gist in case it's spares you or others some time
https://gist.github.com/GWeindel/2af99b1a3b07d2d75d8e27cbacdf741e

Best,
Gabriel

4

External Email - Use Caution

i had the same problem with evoked.plot_joint when i tried using the CSD
routine on epoched data then averaging into an Evoked object -- topography
wouldn't plot, just got a black circle in the bottom left of the panel
where a topomap should be. the evoked plot at the bottom plotted fine though

5

External Email - Use Caution

This looks to be a genuine bug, which I think I've narrowed down to mne.channels.layout.generate_2d_layout. I'll work on a fix. Gabriel, Sammi, if you have GitHub accounts you can subscribe to this issue to track progress: https://github.com/mne-tools/mne-python/issues/7933

-- dan
Daniel McCloy
https://dan.mccloy.info
Research Scientist
Institute for Learning and Brain Sciences
University of Washington

??? Original Message ???

6

External Email - Use Caution

Hi Dan,

I saw that a fix for this was merged - what's the best way of getting this
into the version of mne that i'm using at the moment? im running 0.20.5 at
the moment

sorry to bug about this!
Sammi

7

External Email - Use Caution

hi Sammi,

try :

pip install -U https://github.com/mne-tools/mne-python/archive/master.zip

Alex

8

External Email - Use Caution

Thanks a lot! Out of interest, is there a (theoretical?) reason that
interactive topomaps aren't permitted for CSD type channels, only EEG/MEG ?
can't select a time-period from a joint_plot to get a topo, but unsure why
this might not be the case (unless it just hasn't been implemented yet?)

9

External Email - Use Caution

no particular reason AFAIK

it was never suggested

A

10

External Email - Use Caution

Hi Sammi,
pip install -U https://github.com/mne-tools/mne-python/archive/master.zip
will get you the latest development version, which includes the fix you want.

-- dan
Daniel McCloy
https://dan.mccloy.info
Research Scientist
Institute for Learning and Brain Sciences
University of Washington

??? Original Message ???