Unsupervised Learning for Fast Probabilistic Diffeomorphic Registration

Adrian V. Dalca,Guha Balakrishnan,John Guttag,Mert R. Sabuncu
DOI: https://doi.org/10.1007/978-3-030-00928-1_82
2018-09-14
Abstract:Traditional deformable registration techniques achieve impressive results and offer a rigorous theoretical treatment, but are computationally intensive since they solve an optimization problem for each image pair. Recently, learning-based methods have facilitated fast registration by learning spatial deformation functions. However, these approaches use restricted deformation models, require supervised labels, or do not guarantee a diffeomorphic (topology-preserving) registration. Furthermore, learning-based registration tools have not been derived from a probabilistic framework that can offer uncertainty estimates. In this paper, we present a probabilistic generative model and derive an unsupervised learning-based inference algorithm that makes use of recent developments in convolutional neural networks (CNNs). We demonstrate our method on a 3D brain registration task, and provide an empirical analysis of the algorithm. Our approach results in state of the art accuracy and very fast runtimes, while providing diffeomorphic guarantees and uncertainty estimates. Our implementation is available online at <a class="link-external link-http" href="http://voxelmorph.csail.mit.edu" rel="external noopener nofollow">this http URL</a> .
Computer Vision and Pattern Recognition,Graphics
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to achieve a fast, probabilistic, and topology - preserving deformable registration method in medical image registration. Although traditional deformable registration techniques can provide impressive results and have a strict theoretical basis, they are computationally intensive because an optimization problem needs to be solved for each pair of images. In recent years, learning - based methods have accelerated the registration process by learning spatial deformation functions. However, these methods have problems such as model limitations, the need for supervised labels, and the inability to guarantee the topological invariance of the deformation. Moreover, most learning methods are not derived from a probabilistic framework that can provide uncertainty estimates. To solve these problems, the author proposes a probabilistic generative model and derives an unsupervised learning algorithm from it, which utilizes the recently developed convolutional neural network (CNN). Specifically, this method can learn image registration through the convolutional neural network without using supervised labels, while ensuring the topological invariance of the deformation and being able to provide uncertainty estimates. This method has been verified on 3D brain image registration tasks, and the results show that it has reached the state - of - the - art level in terms of accuracy, running speed, topology preservation, and uncertainty estimation.