Deformable Registration of Brain MR Images via a Hybrid Loss

We learn a deformable registration model for T1-weighted MR images by considering multiple image characteristics via a hybrid loss. Our method registers the OASIS dataset with high accuracy while preserving deformation smoothness.