MRG Butterfly epoch

[jaeilepp]

Butterfly plot can be toggled with a 'b' key.
closes #2205

[mainakjas]

@jaeilepp can we have a help pop-up when you press ? a la gmail? I think @choldgraf @teonlamont and myself all agree that this would be useful.

[agramfort]

can you share screenshots to it's easy to have a feeling of what you're working on?

[jaeilepp]

Currently working on y axis...

[agramfort]

you're going to make 3 butterflies? one per channel type?

[jaeilepp]

There's a button for changing which channel type to view. Default view is the one with all channels.

[mainakjas]

hmm ... do we really need a button? can't we just have it all there as three subplots?

[jaeilepp]

can't we just have it all there as three subplots?

Good point. What about the other types (EOG etc.). Should they be shown too?

[agramfort]

EOG should be displayed separately

[mainakjas]

I was thinking more along the lines of evoked.plot() which has grad, mag and EEG nothing else. If we start looking at EOG, ECG, we have tons of other channel types ...

[jaeilepp]

So what's the verdict? Three axes with mag, grad and eeg, fourth with eog or one for all existing types?

[mainakjas]

I'll let @agramfort decide

[dengemann]

Three axes with mag, grad and eeg, fourth with eog or one for all existing types?

Yes, this should be it!

[mainakjas]

I don't quite like the toggle key to be 'b' because it doesn't make sense when you want to go back from the butterfly plot. How about something like spacebar (or any other better suggestions ...) ?

[choldgraf]

+1 to a pop up infobox with possible key presses. Personally I think 'b' would be fine if the option were given as "turn butterfly plot on/off". Then it's just a toggle and semantically makes sense to be 'b' for me.

[jaeilepp]

Ready for review, I guess. Unless we want the infobox for this PR.

[dengemann]

Unless we want the infobox for this PR.

infobox later.

[dengemann]

@jaeilepp trying to run the example on your branch, it seems there are regressions,

/Users/dengemann/github/mne-python/examples/plot_from_raw_to_epochs_to_evoked.py in <module>()
     51 
     52 # Plot epochs.
---> 53 epochs.plot(trellis=False, block=True, fig_title='Auditory left/right')
     54 
     55 # Look at channels that caused dropped events, showing that the subject's

TypeError: plot() got an unexpected keyword argument 'fig_title'

Also the figure looks weird now:
https://www.dropbox.com/s/r9sl6c655bn3osc/Screenshot%202015-06-17%2012.34.38.png?dl=0

[dengemann]

Butterfly view looks awesome!

I really much more liked the event ids on the top x axis, it was much cleaner.

[jaeilepp]

I really much more liked the event ids on the top x axis, it was much cleaner.

Yeah, especially when there are a lot of epochs/view. I should probably change it back and deal with the layout problem of the title on another PR.

[teonbrooks]

it'll be nice to have the infobox, I honestly have no clue what key does what other than the '-', '=', and '+'.

[mainakjas]

yeah +100 on infobox

[agramfort]

screenshot? :)

[jaeilepp]

[mainakjas]

The channel bar shouldn't be there any more no?

[mainakjas]

how do I know which is grad, mag or EEG?

[mainakjas]

no units on y axis?

[mainakjas]

can we have it interactive so that you can click and know the channel name, just like in evoked.plot() ?

[teonbrooks]

for the title, does that require a separate pr to fix? it's reverted to the clustering in the top. @dengemann my two concerns about the event id at the top is that it is clutters the title area at the moment, and it's not labeled. i know that @jaeilepp mentioned that to get the layout right with the title and the event ids at the top, he would need to rework the shared layout for both raw.plot and epochs.plot

[jaeilepp]

can we have it interactive so that you can click and know the channel name, just like in evoked.plot() ?

What would be the button? Left click selects a bad epoch and right click draws a line.

[jaeilepp]

We are running out of buttons. Left click is for bad epochs and right click for vertical lines.

[mainakjas]

Can't we use left click on channel name?

[jaeilepp]

Yeah, that's possible.

[mainakjas]

ok let's do it then :)

[jaeilepp]

Ready for review.

[teonbrooks]

clicking to select bad channels is great! we will need that feature in raw.plot but that can be a separate pr.
the butterfly plot is slow for me :( what are people's opinions on just plot the max and min of the good channels with fill and leave the bads as it is.

[teonbrooks]

just noticed this: dropping bads in the epochs viewer should update the info['bads']. this is the behavior we have in raw.plot.

[teonbrooks]

sorry, my last comment was mistaken. the channel amplitudes changed from butterfly to channel view which made the view look different. I think we would want to keep the scaling of the two plots independent.

[agramfort]

+1 for merge when Teon is happy

[teonbrooks]

@agramfort what are your thoughts on the butterfly plot just being the envelope of the signal. it might be time for me to upgrade my hardware but the browsing very sluggish b/c I think it's plotting all the lines in this view but that doesn't really give more information.

[agramfort]

I agree that it's dense but it's a quick way to see if one channels goes wrong compared to others.

[dengemann]

@teonlamont I'd rather suggest to collect and implement a low performance config profile that can be selected optionally, e.g. for SSH or old computers.

[dengemann]

with mne.viz.low_performance(**kwargs):
    epochs.plot()

[teonbrooks]

ok cool. i will look into doing so. will be switching back to my desktop soon so this should become less of a problem.
only concern now is independent scaling for the plot. i'm actually ambivalent now so everyone likes it as is then i'm +1 then

[teonbrooks]

oh, maybe a command line message to remind/notify the user of the drop channel similar to the drop epochs message. would be very useful.

[dengemann]

only concern now is independent scaling for the plot.

What do you mean?

[jaeilepp]

Just fixed it.

[teonbrooks]

What do you mean?

scaling butterfly plot doesn't affect channel view and vice versa

[teonbrooks]

for plotting the bad channels, can you make it so that it is in the background for all the channels. in the channel view, it is currently overlapping the previous channel.

[agramfort]

Let's merge then

On 24 juin 2015, at 11:23, Teon L Brooks notifications@github.com wrote:

oh, maybe a command line message to remind/notify the user of the drop channel similar to the drop epochs message. would be very useful.

—
Reply to this email directly or view it on GitHub.

[jaeilepp]

Yeah, ready from my part.

[dengemann]

Thanks @jaeilepp !!!