Fast Predictive Image Registration

TL;DR

This paper introduces a deep learning method for rapid image registration prediction that maintains the properties of LDDMM, significantly reducing computation time and enabling uncertainty estimation in deformations.

Contribution

It presents a patch-based deep encoder-decoder network predicting LDDMM parameters, achieving large speedups and incorporating Bayesian uncertainty estimation.

Findings

1500x speedup in 2D registration

66x speedup in 3D registration

Better accuracy than direct deformation prediction

Abstract

We present a method to predict image deformations based on patch-wise image appearance. Specifically, we design a patch-based deep encoder-decoder network which learns the pixel/voxel-wise mapping between image appearance and registration parameters. Our approach can predict general deformation parameterizations, however, we focus on the large deformation diffeomorphic metric mapping (LDDMM) registration model. By predicting the LDDMM momentum-parameterization we retain the desirable theoretical properties of LDDMM, while reducing computation time by orders of magnitude: combined with patch pruning, we achieve a 1500x/66x speed up compared to GPU-based optimization for 2D/3D image registration. Our approach has better prediction accuracy than predicting deformation or velocity fields and results in diffeomorphic transformations. Additionally, we create a Bayesian probabilistic version of our network, which allows evaluation of deformation field uncertainty through Monte Carlo sampling using dropout at test time. We show that deformation uncertainty highlights areas of ambiguous deformations. We test our method on the OASIS brain image dataset in 2D and 3D.