[Mne_analysis] mne_analyze and mne coreg and transformation matricees

Nugent, Allison C. (NIH/NIMH) [E] nugenta at mail.nih.gov
Fri Jan 26 14:33:25 EST 2018
Search archives:

Hello,

     I am trying to understand the way that MNE transforms between coordinate systems.  I have CTF data, MRI data run through freesurfer, and a Polhemus fasttrack headshape file.  I'm running the graphical tools through python 2.7, although I tend to do more work through python3.  Everything in this message is referring to mne_analyze version 2.56, and mne coreg version 0.15.2.  First, I formatted the digitizer data properly, and changed it to a .fif file using mne_ctf_dig2fiff.  I decided to try out both the mne_analyze pipeline and the mne coreg pipeline to see which one fit better with my workflow.  I seem to get different results depending on which tool I use.

Firstly, it seems that mne_analyze needs the lh.seghead converted to .fif, it seems that mne coreg does not. In mne_analyze, when I load the surfaces, the digitizer data, open the viewer window, and open the coordinate alignment tool, it appears that the fiducial marks have default locations, I alter these to ones on the surface.  To facilitate comparison, I set the fiducials to the same locations in the mne coreg gui as well.  When I load the digitizer data in mne coreg, it appears that the data is automatically translated such that the nasion fiducial is set to be the same.  Sure enough, if I save the coordinate transform at that point, there is already a translation set (although no rotations).

Next, on mne_analyze I press "align using fiducials" and "fit fiducials" on mne coreg.  I would expect these to do the same things.  However, the transformation matrices are different.  In fact, the scaling is quite different, because in mne_analyze, many of the headshape points actually lie inside the brain, whereas in mne coreg, most points are outside the brain.  (I'd be happy to share screen shots here, the MRI is from my own brain/head).

Also, if I run through fitting the headshape, I get results that look qualitatively very similar.  However, the transformation matrices produced are quite different.  Why would these results be so discrepant?  Does one tool use the switch from CTF to neuromag, and the other doesn't? Are they different transforms? This website https://martinos.org/mne/stable/manual/source_localization/forward.html states that the transform saved from mne_analyze is T2, which is head to MRI.  But is it t_ctf_head_mri or t_head_mri?

Ultimately, part of what I'm trying to do is get these transformation matrices so they can be useful both within and outside MNE.  I have users that want to use other processing software, but still want the functionality of fitting digitizer headshapes to the freesurfer brain.  At some point, I want to get out the transformation from ctf_dev to the starting point of mne_analyze or mne coreg, so that I can compute the t_ctf_dev_mri transformation. I think I know how to do this, once I figure out what I'm getting out of mne coreg and mne_analyze.

Thank you!
Allison
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20180126/db98285b/attachment.html 


More information about the Mne_analysis mailing list