For the registration of embryo data, we use slightly different settings for MICebuildmodel compared to the registration of brain data. So far we have not been able to determine a workable set of parameters for mincANTS, and so we use minctracc. To allow for larger deformations (embryo data tends to be more dissimilar) a nonlinear stage with a larger blurring kernel is added at the start of the nonlinear stage, and the stiffness settings for the elastic grid for minctracc are set more loosely.
For input data at a resolution of 15micron, the following is a sample incantation of the MICebuildmodel.pl script :
Assumptions:

> MICebuildmodel.pl pipelinename some_embryo_pipeline initmodel /some/initial/model/prefix noinormalize noresampleatlas noregistrationaccuracy nonuc lsq6initialstep 0.1 lsq6identity nolsq6largerotations lsq6kernels 1.0_blur,0.8_blur,0.5_blur lsq12 lsq12initialstep 0.1 lsq12kernels 2.0_blur,1.0_blur lsq12maxpairs 0 nlin nlinstats nlinregistrationmethod minctracc nlinprotocol embryo_smasher.pl stiffness 0.5 similarity 0.5 weight 0.3 [ input embryo MINC files ] 
where the file embryo_smasher.pl contains:
> cat embryo_smasher.pl ( { memory => 4, generation => 1, kernel => 0.6, iterations => 40, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.3, }, { memory => 4, generation => 2, kernel => 0.4, iterations => 40, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.3, }, { memory => 4, generation => 3, kernel => 0.2, iterations => 20, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.2, }, { memory => 4, generation => 4, kernel => 0.15, iterations => 20, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.15, }, { memory => 4, generation => 5, kernel => 0.09, iterations => 10, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.09, }, { memory => 4, generation => 6, kernel => 0.05, iterations => 10, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.05, }, { memory => 4, generation => 7, kernel => 0.03, iterations => 10, simplex => 3, ncpus => 1, use_gradient => 0, optimization => "use_simplex", step => 0.03, }, ); 