Image registration techniques are used routinely in a variety of today’s medical imaging diagnosis. Since the problem is ill-posed, one may like to add additional information about distortions. This applies, for example, to the registration of contrast enhanced images, where variations of substructures are not related to patient motion but to contrast uptake. Here, one may only be interested in registrations which do not alter the volume of any substructure. In this paper we discuss image registration techniques with a focus on volume preserving constraints. These constraints can reduce the non-uniqueness of the registration problem significantly. Our implementation is based on a constrained optimization formulation. Upon discretization, we obtain a large, discrete, highly nonlinear optimization problem and the necessary conditions for the solution form a discretize nonlinear partial differential equation. To solve the problem we use a variant of Sequential Quadratic Programming method. Moreover, we present results on synthetic as well as on real life data. 1

Image registration is central to many challenges in medical imaging today. It has a vast range of applications. The purpose of this note is twofold. First, we review some of the most promising non-linear registration strategies currently used in medical imaging. We show that all these techniques may be phrased in terms of a variational problem and allow for a unified treatment. Second, we introduce, within the variational framework, a new nonlinear registration model based on a curvature type regularizer. We show that affine linear transformations belong to the kernel of this regularizer. This has the important consequence that an additional pre-registration step is no longer necessary. Furthermore, we develop a stable and fast implementation of the new scheme based on a real discrete cosine transformation. We demonstrate the advantages of the new technique for synthetic data sets and present an application of the algorithm for registering MR-mammography images.

Introduction The fundamental problem in brain warping is to define the class of admissible spatial transformations, which must be sufficiently broad to enable a reference anatomy to fit all subject anatomies, and to develop efficient, automated algorithms for the calculation of the appropriate transformation. In this chapter, we focus on numerical methods for inferring spatial warps that are very high in dimension in order to accommodate the complex ways in which the neuroanatomy of normal individuals can vary. Specifically, the elastic matching technique described in a previous chapter is implemented. The warps therefore correspond to deformations in the continuum mechanics, and we require methods for solving boundaryvalue problems. Two approaches are standard and each involves a different way of discretizing the problem. The finite difference method , which operates directly on the motion equations, is easy to code and computationally fast, but the fi

In this work, we describe an automated approach to morphometry based on spatial normalizations of the data, and demonstrate its application to the analysis of gender differences in the human corpus callosum. The purpose is to describe a population by a reduced and representative set of variables, from which a prior model can be constructed. Our approach is rooted in the assumption that individual anatomies can be considered as quantitative variations on a common underlying qualitative plan. We can therefore imagine that a given individual's anatomy is a warped version of some referential anatomy, also known as an atlas. The spatial warps which transform a labeled atlas into anatomic alignment with a population yield immediate knowledge about organ size and shape in the group. Furthermore, variation within the set of spatial warps is directly related to the anatomic variation among the subjects. Specifically, the shape statistics---mean and variance of the mappings---for the population ...

automated, feature matching approach to label cortical structures and activity anatomically in human brain MRI data. This approach does not assume that the existence of component structures and their relative spatial relationship is preserved from brain to brain, but instead disassembles a labeled atlas and reassembles its pieces to match corresponding pieces in an unlabeled subject brain before labeling. Mindboggle: (1) converts linearly coregistered subject and atlas MRI data into sulcus pieces, (2) matches each atlas piece with a combination of subject pieces by minimizing a cost function, (3) transforms atlas label boundaries to the matching subject pieces, (4) warps atlas labels to their transformed boundaries, and (5) propagates labels to fill remaining gaps in a mask derived from the subject brain. We compared Mindboggle with four registration methods: linear registration, and nonlinear registration using SPM2, AIR, and ANIMAL. Automated labeling by all of the nonlinear methods was found to be at least comparable with

We have developed a general-purpose registration algorithm for medical images and volumes. The transformation between images is modeled as locally affine but globally smooth, and explicitly accounts for local and global variations in image intensities. An explicit model of missing data is also incorporated, allowing us to simultaneously segment and register images with partial or missing data. The algorithm is built upon a differential multiscale framework and incorporates the expectation maximization algorithm. We show that this approach is highly effective in registering a range of synthetic and clinical medical images. 1

Abstract not found

A two-step parameter-free approach for non-rigid medical image registration is presented. Displacements of boundary structures are computed in the first step and then incorporated as hard constraints for elastic image deformation in the second step. In comparison to traditional non-parametric methods, no driving forces have to be computed from image data. The approach guarantees the exact correspondence of certain structures in the images and does not depend on parameters of the deformation model such as elastic constants. Numerical examples with synthetic and real images are presented. 1 Introduction Numerous applications in modern medical imaging deal with non-rigid image registration. Examples are image-atlas as well as multi-modality image registration in neurosurgery. There, a three-dimensional image (deformable template) has to be completely transformed onto another one (study). One group of methods dealing with non-rigid image registration is the so-called non-parametric metho...

This paper appeared in Medical Image Analysis, Volume 6, Number 4, 2002, pages 337--359

Abstract: Over recent years, non-rigid registration has become a major issue in medical imaging. It consists in recovering a dense point-to-point correspondence field between two images and usually takes a long time. This is in contrast to the needs of a clinical environment, where usability and speed are major constraints, leading to the necessity of reducing the computation time from slightly less than an hour to just a few minutes. Another constraint is the usual unwillingness of healthcare organizations to invest in expensive high-performance computing solutions. Cluster computing proved to be a convenient solution to our computation needs, offering a large processing power at a low cost. Bi-processor workstation can be used simultaneously for parallel computation and individual day-to-day use and they are already present in many labs and hospitals. Our goal in this article is to provide a more usable tool by taking advantage of the available computation power. Among the fast and efficient non-rigid registration algorithms, we chose the demons algorithm [1,2] for its simplicity and good performances. The parallel implementation decomposes the correspondence field into blocks, each block being assigned to one workstation of the cluster. We take advantage of the inherently regular structure of the algorithm that allows a nearly perfect static load balancing and also its locality, allowing to keep the amount of communication within a reasonable range. We obtained an acceleration of 11 by using 15 2GHz PC’s connected through a 1GB/s Ethernet network and reduced the computation time from 40min to 3min30.