LDDMM meets GANs: Generative Adversarial Networks for diffeomorphic registration

Ubaldo Ramon Julvez, Monica Hernandez, Elvira Mayordomo

Keywords: Large Deformation Diffeomorphic Metric Mapping, Generative Adversarial Networks, geodesic shooting, stationary velocity fields

TL;DR: We propose an adversarial learning LDDMM method for pairs of 3D mono-modal images based on Generative Adversarial Networks.

Abstract: The purpose of this work is to contribute to the state of the art of deep-learning methods for diffeomorphic registration. We propose an adversarial learning LDDMM method for pairs of 3D mono-modal images based on Generative Adversarial Networks. The method is inspired by the recent literature for deformable image registration with adversarial learning. We combine the best performing generative, discriminative, and adversarial ingredients from the state of the art within the LDDMM paradigm. We have successfully implemented two models with the stationary and the EPDiff-constrained non-stationary parameterizations of diffeomorphisms. Our unsupervised learning approach has shown competitive performance with respect to benchmark supervised learning and model-based methods.

Dataset Code:

Data used in the preparation of this article was obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database: https://adni.loni.usc.edu/ and from the Non-rigid Image Registration Evaluation Project (NIREP) database: https://github.com/andreasmang/nirep.

Source code sharing is not possible as per the request of one of the authors. Part of the code used for this article is also being used in a separate project and we do not wish to make that code public until its completion and publication.