MRG, VIZ, FIX: sEEG picking in _prepare_topomap_plot()

[drammock]

closes #8733

Not sure if this is a reasonable fix since I've never personally worked with sEEG, but based on general knowledge I'm assuming there's nothing special needed as far as combining sensor pairs? Can someone confirm?

cc @BarryLiu97. On this branch using your code from #8733 I get this figure:

[drammock]

All green, ready for review/merge.

[zhengliuer]

Great!

[agramfort]

given the extrapolation artifacts wouldn't it make sense to restrict the topomap to the convex hull of the sensors?

[larsoner]

Otherwise LGTM

And yes in the example something like extrapolate='local' (I think this is what does the convex + padding?) would look better

[larsoner]

Typically with these we just do:

Suggested change

	with warnings.catch_warnings(): # unit has changed from T/m to V
	warnings.simplefilter('ignore')
	evoked.set_channel_types(mapping)
	with pytest.warns(RuntimeWarning, match='unit changed')
	evoked.set_channel_types(mapping)

Does the same (plus a bit more, checking that we got the expected one) with fewer lines/imports

[drammock]

Here's what the sample code from #8733 generates now:

[agramfort]

I find it better

…

On Wed, Jan 13, 2021 at 6:13 PM Daniel McCloy ***@***.***> wrote: Here's what the sample code from #8733 <#8733> generates now: [image: Figure_1] <https://user-images.githubusercontent.com/1810515/104485615-5aaef880-5590-11eb-85cf-f49ecec37bfe.png> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#8736 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABHKHABCFYGFEISWBIMEGDSZXIETANCNFSM4V7RC6GQ> .

[jasmainak]

Late to the party, does plot_joint even make sense for sEEG data? What can you tell from this figure? I think you should just throw an error.

You need proper support for sEEG and a separate function altogether. Each "stick" plotted in a separate subplot.

[larsoner]

Even if suboptimal I think it should work. We can create other optimized functions for sEEG visualization that people can can prefer as well, but I don't see a good reason to raise an error here.

[zhengliuer]

Late to the party, does plot_joint even make sense for sEEG data? What can you tell from this figure? I think you should just throw an error.

You need proper support for sEEG and a separate function altogether. Each "stick" plotted in a separate subplot.

This sample data is not very good for many reasons, like the experimental paradigm is not good enough. So it is normal to see nothing useful, but to sEEG, it is useful if the experimental paradigm is good.
And it should be very good if every shaft has its own subplot

[jasmainak]

@BarryLiu97 do you have a screenshot or a figure from a publication to share which shows how topoplots are used in sEEG data analysis?

[zhengliuer]

@BarryLiu97 do you have a screenshot or a figure from a publication to share which shows how topoplots are used in sEEG data analysis?

Sadly, I haven't seen any publication using figures like plot_evoke_joint. But in sEEG, it may make sense if using brain regions as the standard to separate the subplots for the most important feature of sEEG is the ability to get the deep brain structures' information.

[larsoner]

Thanks for the fixes @drammock !