[FieldTrip] MEG source-model warp query

[Dr N.E. Adams]

Dear community experts

I hope you're all doing well! I am new to fieldtrip and I'm struggling 
to see how to get a sourcemodel that I can match directly to an atlas. 
I'm working with MEG data. In short, currently:

a) I generate a template grid
b) I constrain that grid by vertices found within an AAL atlas
c) I use this constrained template grid to create a sourcemodel for each 
individual's MEG-coregistered MRI using 'warpmni' in the config.

However, no warp process appears to affect the result - only a standard 
geometric transformation using an affine (translation and rotation), so 
I believe I'm either doing something wrong or missing a vital step. 
Debugging through ft_prepare_sourcemodel I find, from this line:...

sourcemodel.pos = ft_warp_apply(inv(normalise.initial), 
ft_warp_apply(normalise.params, mnigrid.pos, 'sn2individual'));

... that it first normalises (which I assume actually *is* a warp) the 
individual's mri (variable = normalise), transforms the template grid 
(variable = mnigrid) by the normed mri affine (note: doesn't use this to 
warp the grid, just transforms it), and then transforms that result by 
the non-normed mri.

1) I'm curious, why does it do this in two stages? i.e. why not just 
transform the template grid straight to the mri affine? What is the use 
of the normalised mri affine in the intermediary step, as it doesn't use 
the warp itself, just the affine?

2) The sourcemodel needs to be in the individual's native shape in order 
to do the source-reconstruction correctly. By changing my template 
construction method, I can create a grid that has complete coverage of 
the individual's brain (and is unrelated to the atlas). However, the 
atlas is in a standard shape so I can't match such source vertices to it 
unless I can either:
     - warp the standard shape atlas to the individual's brain or
     - make a version of the sourcemodel where it's vertices have been 
warped to the atlas shape, so that when I have used the original 
sourcemodel to source-reconstruct my data, I can use the warped 
sourcemodel to know where each vertex is in the atlas.

I have considered that this might be what ft_source_interpolate is for, 
but I've not managed to understand what it's doing, or even understand 
if I'm giving it the right information. For instance, using:...

interp = ft_sourceinterpolate(cfg, atlas, sourcemodel);

... produces a structure that looks a lot like the atlas, and contains 
what appears to be nonsense! :) When what I'd like to see from it is the 
sourcemodel grid warped to the shape of the atlas. - How would I get 
this?

I have attached some of my relevant code in the .txt file for you to 
peruse.

Thanks
Tallie
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sourcemodel to atlas - fieldtrip query.txt
Type: text/x-c++
Size: 6069 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200402/56d50e67/attachment.bin>