-
Notifications
You must be signed in to change notification settings - Fork 1.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ENH: plot_topo layout creation #3987
Comments
Yeah, that sounds like a good idea. I was already experimenting with this, but it seems that |
This would fix #3814 |
no opinion
|
You mean the horizontal offset? I think this could actually be correct. If
you set dev_head_t to identity does it go away? If so it indicates that the
head was shifted to the side relative to the sensors, making this a feature
not a bug.
…On Feb 22, 2017 5:07 AM, "Alexandre Gramfort" ***@***.***> wrote:
no opinion
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#3987 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ACQZXixYfX3GGL6AxBJrS2ABzmBav087ks5rfAjOgaJpZM4MAho3>
.
|
Why would it go away? These are just the sensor locations from the layout file. |
Then we should probably apply dev_head_t to get sensor locations in head
coordinates. Then the spatial relationship between the head and the
sensors, at least is terms of the center, will be correct.
|
But these are static locations from the layout file. Why would it help to transform them? |
If we are plotting sensors on top of a head, it would be best to plot the sensors with the same (or at |
... so maybe not for
This will give us much more realistic geometry than our current procedure, which is just to center based on extrema and scale everything to hit the boundaries. |
Even some of our built-in MEG layouts don't give head coordinates (like KIT-AD). I'm not sure it's worth the effort to change this. After all, in most cases our layouts work fine and the change would require overhaul of all the plotting functions that use layouts. Also, topomaps use layouts so it might be dangerous to change this. |
I agree it would be a big change. Such a process should make all of the
plots more accurately represent the head to sensor relationship, though.
Actually we'd really want to skip the layout entirely and use the channel
positions from info. In the case of MEG sensors, we'd just need to apply
dev_head_t.
Maybe we could add it as a separate mode?
|
The layouts should still be used at least for topo-plots since the sensor triplets have identical positions. |
Ahh right. Forgot about that :) And I guess there is some logic to the "unwrapping" transformation we normally do, since it's just one way (of many possibilities) to do a projection of spherical data. Maybe just disabling norm and/or centering is enough |
@jaeilepp I think you have more or less fixed this, right? |
Yeah, more or less. Maybe eventually it could be good to shift to convention you suggested, but I don't see a lot to be gained. Closing. |
This is essentially resurrected by #4239. My proposal is that:
This will fix issues like Edit: To clarify, |
... so the distinction would be that Layout is only ever used for interactive "sensor selection" stuff, and never shown on a head. Only sensor positions in head coordinates ever get shown on a head (which could easily tie in eventually to the proposed Digitization class). |
sounds reasonable to me !
|
This one won't happen for 0.15, moving to 0.16 |
I don't see this as a blocker for 0.16 and whatever solution we come up with should be tested thoroughly, so moving to 0.17 |
Not fixed yet, so reopening |
It seems like we keep hitting issues with topo scaling confusing people. I think we probably shouldn't change where (0., 0., 0.) is in any given channel layout, and instead ensure that our internal layouts are actually in head coordinates. I wonder if we could also just assume a head is e.g. 90 cm. That would make subselections of electrodes appear correct. A
normalize=True (default) | False
argument in functions that do such topologies could do it.@jaeilepp WDYT?
The text was updated successfully, but these errors were encountered: