@agramfort use transparent skin but filled colors for fiducials and digitized points so we get a sense of depth

add centroid for the nasion and for the ear fiducials as those are typically recorded with good precision by the researchers

add centroid for the nasion and for the ear fiducials as those are typically recorded with good precision by the researchers

I think these were originally large because @christianbrodbeck had the opposite sense, that there was some ambiguity in it (from a digitizer accuracy standpoint, probably). But I agree that with well behaved subjects you can more precise than we show here, and it shouldn't hurt to make them the same size as the MRI ones. I'd rather just shrink the markers rather than make a large translucent one and a small opaque one.

use transparent skin

We tried this default in #8446 and decided against it. So I think if people want this they should use the new dict support for surfaces by doing something like surfaces=dict(head=0.7).

filled colors for fiducials

Part of the reason the fids are translucent is because in theory the bem/subject-fiducials.fif that you mark in MRI space should align perfectly / overlap with the fiducials recorded using the digitizer, so we need some way to see two sets of points in the same space. This is also why the MRI fids are opaque and the digitizer fids are translucent (and much larger, but I agree this isn't really necessary, as above). One option would be to make the MRI fiducials some other shape like a square, and the digitized ones stay a circle. That way you can see both even if they line up. (I think it's okay to use squares even though they have sidedness / orientation -- we can align the sides with the XYZ/RAS of the MRI, so actually get a bit more information out of it.)

If I recall correctly the reason for making the digitizer fiducials large was mainly to be able to see them together with the MRI fiducials, and to see the even when, for example, they're sunk slightly under the MRI head surface. With template brains the expected alignment is less perfect.

+1 for being consistent, and I think the yellowish head is more pleasant / easier to see than the gray one. Regarding fiducial size, etc, I think making them somewhat smaller would be OK, and I like the idea of MRI fids and digitizer fids being different geometric solids... actually I think a cube for one and a regular octahedron for the other would be ideal --- if you pick the relative sizes right it would make it really easy to tell if they line up:

@drammock do you want to lead this following my gist above?

+1 for cube plus something else. I'm not sure VTK has an octahedron glyph, if not then cube+sphere seems fine (cube for MRI, sphere for dig) and can also have a nice clean overlap if their volumes are identical. This is also consistent with how dig points show up as spheres anyway...

I would start by trying to make them equal-volume, leading to a cube-side-to-sphere-radius ratio of:

>>> (3 / (4 * pi)) ** (1/3)
0.6203504908994001

This way perfect alignment means some parts of the sphere will protrude, and the corners of the cube will, too.

@drammock do you want to lead this following my gist above?

how urgent is this? it might take me some time to get up to speed on vtk etc.

Not that urgent, but I think I can probably do it quickly -- let me take a look

Okay I took a try with the cube. I agree octahedron would be better but it will be a pain to implement.

$ mne coreg -d subjects -s sample --fif MEG/sample/sample_audvis_raw.fif --trans MEG/sample/sample_audvis_raw-trans.fif

Gives:

And with "fit fid" you get a different alignment where the nasion is aligned, you can see they are on top of each other at least:

And tutorials/source-modeling/plot_source_alignment.py gives:

Okay got octahedron working, ready for review/merge from my end:

thx @larsoner !