SITReg: Multi-resolution architecture for symmetric, inverse consistent, and topology preserving image registration using deformation inversion layers

Deep learning based deformable medical image registration methods have emerged as a strong alternative for classical iterative registration methods. Since image registration is in general an ill-defined problem, the usefulness of inductive biases of symmetricity, inverse consistency and topology preservation has been widely accepted by the research community. However, while many deep learning registration methods enforce these properties via loss functions, no prior deep learning registration method fulfills all of these properties by construct. Here, we propose a novel multi-resolution registration architecture which is by construct symmetric, inverse consistent, and topology preserving. We also develop an implicit layer for memory efficient inversion of the deformation fields. The proposed method achieves state-of-the-art registration accuracy on two datasets.