Back to the main page.

Bug 3127 - implement a pipeline for MEG source reconstruction if no MRI is available

Status NEW
Reported 2016-05-12 13:00:00 +0200
Modified 2018-07-12 20:11:43 +0200
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Mac OS
Importance: P5 normal
Assigned to: Robert Oostenveld
Depends on:
See also:

Robert Oostenveld - 2016-05-12 13:00:10 +0200

This will be needed for the ChildBrain ESRs visiting Nijmegen in July. They have data from a Neuromag scanner that includes headshape info, but not an anatomical MRI for each subject. On top of that, it pertains to children. Hence a small template might be desired. We need some way to spatially transform the template (any template, but desirably an age-specific template) to the MEG headshape.

Robert Oostenveld - 2016-05-12 13:03:17 +0200

Simon (and I) have access to the MR database from John Richards. That would be the best basis for the templates. Let me also CC John on this, as he might be interested in thinking and discussing along.

Simon - 2016-05-12 13:29:44 +0200

(In reply to Robert Oostenveld from comment #1) Is the purpose to generally provide with field trip these age specific template like we do with colin(from MRI to Volume conduction model?) or just for the purpose of the ESRs visiting?

Robert Oostenveld - 2016-05-12 13:43:40 +0200

(In reply to Simon from comment #2) the purpose would be general, i.e. this could address the FAQ "how to do MEG source reconstruction if I don't have an MRI". But the case with ChildBrain and the ESRs fro Jyvaskyla visiting Nijmegen makes it concrete. Also, this makes it concrete that we might not only want this to work for "a template" (like colin27), but for specific templates. PS Note that the old-fashioned (i.e. >20 year in use) solution would be to fit a single or localspheres model to the scalp surface and simply not plot the resulting source reconstruction on any MRI.

John Richards - 2016-05-12 15:02:32 +0200

Thanks for the cc. One can take measures of head shape (and presumably MEG or EEG placement positions) and register the head shape to the MRI (average, or other individual), translate the scalp placement positions into the MRI space, and then use the MRI for the head model for the source analysis. I think this is in a publication (might be Fuchs et al, 2002). We do this routinely with both FLIRT/ANTS for structural MRIs, and with point-set methods with just electrode positions or fiducial positions. We do this with EEG, ERP, and electrodes. I presume similar methods would work for MEG. I am not sure of the first question, but I am using both my average MRI templates and individual participants for individuals without a MRI. We have FT (and EMSE) head models and electrode placement locations (EGI GSN, HGSN, virtual 10-10) for every individual and for the average template. We also have average electrode placements for each average and, and individual electrode placement positions for the individual MRIs. My online database has electrode positions and segmented head models for the average MRI templates. One could use the average MRI templates alone and use other methods for sensor placements; or use the electrode placements for EEG, or use the segmented head models. I presume that one could take the MRIs from the database and use them to create head models on the FT site similar to that was done with the Colin brain. We have FT structures for all our individual MRIs and average MRIs, but these are somewhat specific to my own lab work and lab structures and I am not sure if they would be immediately translatable to a FT site; but the methods would certainly be applicable. We are using both the average MRI templates as stand-in models for infants without MRIs, and also using individual MRIs from an individual with a similar size head and close in age. I can send links to posters/papers using these methods. I have a poster presentation showing that the source analysis results done with the "close MRI' fits reasonably similar to an individual's own MRI; using the age-appropriate average MRI was a slightly lesser fit but still very close; and of course using older children or an adult MRI for infant ERP (or the MNI152) the fit was poorer. I plan to work this into a published study, probably in the coming year. McCleery, J.P., & Richards, J.E. (2012). Comparing realistic head models for cortical source locationization of infant event-related potentials. Poster presented at the International Conference on Infant Studies, Minneapolis, MN. June, 2012. The BESA people did a presentation last year using the MRI templates with BESA; they have head models based on these avgs integrated into BESA. Lanfer, B., Spangler, R. Richards J.E., & Paul-Jordanov, I. (23015). Age-specific template head models for EEG source analysis. Paper presented at the Organization for Human Brain Mapping Meeting, Honolulu HA, June, 2015. John

Robert Oostenveld - 2016-05-12 16:18:14 +0200

(In reply to John Richards from comment #4) Hi John, thanks for the helpful comments. BESA is a partner in the ChildBrain project, to which the concrete question also applies. As you point out with the posters, the methods have in principle been described and tried out, the challenge that I want to address with this bug is to make a practical solution that also works for less experienced researchers. I expect this "solution" to primarily consist of a piece of documentation on how to make and/or use a FT pipeline. It would be a piece of documentation to extend best Robert PS could you perhaps share a PDF of the 2012 poster with us? Perhaps not through a link in bugzilla (which is public); if you send me a private email I can forward it to the specific people that should see it.

Sarang Dalal - 2018-07-12 20:11:43 +0200

Thomas Hartmann and perhaps other OBOB lab members (Gaetan?) implemented FieldTrip-compatible functionality exactly like this: We're in the process of testing it for an adult for which we have a headshape but no MRI. My quick understanding of the code is that it warps a template brain with an affine transform; so perhaps a nonlinear warp would be on my own wishlist. :-)