Learning Registration Models with Differentiable Gauss-Newton Optimisation

Mattias P Heinrich

Keywords: Learn to optimise, medical image registration, second order descent

TL;DR: Learn to predict displacement Jacobian and residual vectors for improving second-order least-squares solution for image registration.

Abstract: We propose to capture large deformations in few iterations by learning a registration model with differentiable Gauss-Newton and compact CNNs that predict displacement gradients and a suitable residual function. By incorporating a sparse Laplacian regulariser, structural / semantic representations and weak label-supervision we achieve state-of-the-art performance for abdominal CT registration.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Image Registration

Secondary Subject Area: Meta Learning

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.