: Neural Deformation Fields for Approximately Diffeomorphic Medical Image Registration
This work proposes , a neural deformation model which results in approximately diffeomorphic transformations. In contrast to the predominant voxel-based approaches, represents deformations functionally which allows for memory-efficient training and inference. This is of particular importance for large volumetric registrations. Further, while medical image registration approaches representing transformation maps via multi-layer perceptrons have been proposed, facilitates both pairwise optimization-based registration as well as learning-based registration via predicted or optimized global and local latent codes. Lastly, as deformation regularity is a highly desirable property for most medical image registration tasks, makes use of gradient inverse consistency regularization which empirically results in approximately diffeomorphic transformations. We show the performance of on two 2D synthetic datasets as well as on real 3D lung registration. Our results show that can achieve similar accuracies as voxel-based representations in a single-resolution registration setting while using less memory and allowing for faster instance-optimization.
READ FULL TEXT