Graph-Based Slice-to-Volume Deformable Registration

Abstract

Deformable image registration is a fundamental problem in computer vision and medical image computing. In this paper we investigate the use of graphical models in the context of a particular type of image registration problem, known as slice-to-volume registration. We introduce a scalable, modular and flexible formulation that can accommodate low-rank and high order terms, that simultaneously selects the plane and estimates the in-plane deformation through a single shot optimization approach. The proposed framework is instantiated into different variants seeking either a compromise between computational efficiency (soft plane selection constraints and approximate definition of the data similarity terms through pair-wise components) or exact definition of the data terms and the constraints on the plane selection. Simulated and real-data in the context of ultrasound and magnetic resonance registration (where both framework instantiations as well as different optimization strategies are considered) demonstrate the potentials of our method.

SymReg-GAN: Symmetric Image Registration with Generative Adversarial Networks

“…Image registration is of a fundamental problem in various fields including for example computer vision, medical imaging and remote sensing [2], [4], [7]. There exist many kinds of image registration techniques proposed in the past decades [1], [14], [15], [16], [17], [18] and they can be classified as directional registration and symmetric registration. Symmetric image registration estimates Fig.…”

“…SymReg-GAN is inspired by CycleGAN but not a simple extension of CycleGAN. CycleGAN is a pix2pix GAN for general image-to-image translation which involves generating a new synthetic version of a given image with a specific modification 1 . Extension of CycleGAN to resolving symmetric image registration is not straightforward considering the fact that the former outputs an image while the latter aims to solve transformation between images.…”

“…M ANY problems in medical image analysis, computer vision and remote sensing can be resolved by a process of estimating a mapping between physically corresponding points within the fields-of-view (FOVs) of a pair of images, which is called image registration and has continued to be an active research topic since 1960s [1], [2], [3], [4], [5], [6], [7]. Image registration is commonly posed as an estimation of the geometric transformation of one "moving" image into the space of another "fixed" image.…”

“…A large variety of techniques for both directional image registration [1], [14], [15], [16], [17], [18] and symmetric image registration [8], [9], [10], [19], [20] have been proposed (as reviewed in Sec. 2) .…”

Daubechies wavelet‐based local feature descriptor for multimodal medical image registration

“…This is due to the fact that medical images are more vulnerable to noise. A wide range of registration methods are published [10–14], covering all category of image registration schemes. Conventional feature extraction methods such as shift‐invariant feature transform [15], principal component analysis [16], and speed‐up robust features [17] provide strong points for registration using linear transformations.…”