Image registration is the process of aligning two or more images to achieve point-wise spatial correspondence.Typically, image registration is phrased as an optimization problem w.r.t. a spatial mapping that minimizesa suitable cost function and common approaches estimate solutions by applying iterative optimization schemessuch as gradient descent or Newton-type methods. This optimization is performed independently for each pairof images, which can be time consuming. In this paper we present an unsupervised learning-based approach fordeformable image registration of thoracic CT scans. Our experiments show that our method performs comparableto conventional image registration methods and in particular is able to deal with large motions. Registrationof a new unseen pair of images only requires a single forward pass through the network yielding the desireddeformation field in less than 0.2 seconds. Furthermore, as a novelty in the context of deep-learning-basedregistration, we use the edge-based normalized gradient fields distance measure together with the curvatureregularization as a loss function of the registration network.
展开▼
