Extracting the LDDMM geodesic from ANTs registration

[ghost]

Hello,

I understand that the 'fwdtransforms' entry of the dictionary returned by ants.registration() contains the matrix representation of the diffeomorphism mapping the moving image to the fixed image. This diffeomorphism [phi] is the endpoint at t = 1 of the full geodesic path of diffeomorphisms going from the identity diffeomorphism (t = 0) to the full registration [phi]. How can I extract this geodesic path using ANTs? This would ultimately boil down to finding the time-varying velocity field between the two images, from which I can compute the flow via integration to obtain a point on the desired geodesic at any time 0 < t < 1. Any help or suggestions would be really appreciated.

Many thanks, Gianfranco Cortes

[ntustison]

If you want to use one of the TimeVaryingVelocityField transforms and write out the velocity field, you're going to need to use the ANTs antsRegistration program directly. ants.registration() in ANTsPy doesn't provide user access to that output.

[stnava]

@ntustison - quick q: do you think it would be easy to add this option to ants.registration and antsRegistration given our recent various discussions on this topic?

[ntustison]

Yeah, I'll see if I can do that.

[ntustison]

https://github.com/ANTsX/ANTsPy/pull/207

This should work. Once we get it checked in, I'll propagate the enhancements to ANTsR.

[stnava]

here is an example from a recent checkin building on commits by @ntustison

python
import ants
fi = ants.image_read(ants.get_ants_data('r16'))
mi = ants.image_read(ants.get_ants_data('r64'))
syn = ants.registration(fixed=fi, moving=mi, type_of_transform = 'SyNCC' )
mytx = ants.registration(fixed=fi, moving=mi, type_of_transform = 'TV[6]', grad_step=10, syn_metric='CC', syn_sampling=4, verbose=True )
ants.image_mutual_information( fi, mi )
ants.image_mutual_information( fi,  mytx['warpedmovout']  )
ants.image_mutual_information( fi,  syn['warpedmovout']  )

however, we have not added post-hoc velocity field integration to antsr or antspy just yet.

[stnava]

ok - we added ants.integrate_velocity which should make this complete.