Coregistering subjects: group template for mixed models?

Discussions
, ,
1

Is there some way of generating a mixed surface-volume group grid template and transforming it to subject space, for example doing the freesurfer-based cortical surface alignment and then the SPM based subcortical volume and then merging them for each subject before creating the head model? @Sylvain @John_Mosher

Best regards,

Emily

Ref: http://neuroimage.usc.edu/brainstorm/Tutorials/CoregisterSubjects

2

Hi Emily:

In principle, yes: you could indeed run the two procedures (surface and volume template alignments) described in the tutorial then create a mixed surface-volume source model from both, after pruning out the cortical grid points from the volume model, to avoid redundancy with the surface model. However, I think this is ambitious 'hacking'. Did you try to use just a volume model in the first place?

3

Hi Sylvain,

Thanks for your response and suggestions.

I am using a volume-only model for another study, which so far seems to work. But the one I’m currently working on is a follow-up of a certain previous work in which we used mixed models. Since we’d like to replicate and extend it, I think it would make sense to stick with the mixed models. However, it would be nice to be able to view averaged results, which perhaps are somewhat easier to interpret (and nicer looking)on the cortical surface, so I was hoping of finding a way of doing that while still taking into account sub-cortical sources.

Another consideration is that it’d be nice to extract data from volume ROIs of comparable size and position, which would be easier using a single atlas in standard space (right now I’m marking the centre in std space coordinates on subject brains and growing them until they are ~equal volume, which is okay but slow).

Do you have a sense of if/how well the mixed subject models project to mixed standard anatomy models?

Best,

Emily

4

I have never tried, and as I mentioned above, this is doable but I think labor intensive. A more direct alternative would be to use the same mixed model (ie a non-deformed template) for all subjects. You would loose in anatomical accuracy but that would make the group analyses straightforward.

1 Like