Efficient Population Registration of 3D Data

Abstract

We present a population registration framework that acts on large collections or populations of data volumes. The data alignment procedure runs in a simultaneous fashion, with every member of the population approaching the central tendency of the collection at the same time. Such a mechanism eliminates the need for selecting a particular reference frame a priori, resulting in a non-biased estimate of a digital atlas. Our algorithm adopts an affine congealing framework with an information theoretic objective function and is optimized via a gradientbased stochastic approximation process embedded in a multi-resolution setting. We present experimental results on both synthetic and real images.

Attribute Vector Guided Groupwise Registration

“…Joshi et al [3] extends a large deformation diffeomorphic mapping algorithm [4] to work in a groupwise manner for unbiased atlas construction. A congealing framework [5] based groupwise registration scheme is proposed in [6], where intensity based entropy drives a gradient-based stochastic optimizer and pushes each image to the population center simultaneously. This method, which originally works for affine transformation only, is further extended by Balci et al [7] to incorporate B-Splines to model nonrigid deformation.…”

“…This method, which originally works for affine transformation only, is further extended by Balci et al [7] to incorporate B-Splines to model nonrigid deformation. The objective function, called stack entropy in [6] and [7], measures the compactness of the intensity distribution for a certain voxel location across different subjects. Similar to the popular entropy based metrics used in pairwise registration (e.g., mutual information [8][9]), the current formulation of stack entropy considers image intensity as the only feature, and discards local contextual information that can be provided by the voxel neighborhood.…”

Weakly Supervised Group-Wise Model Learning Based on Discrete Optimization

“…The approach proposed in this paper is related to two lines of previous work 1) Group-wise registration approaches: in [6] the authors establish a mapping on a spherical reference manifold, [7] which employes piece-wise affine deformations to map the entire data, [8] which uses congealing to obtain a model of appearance variation from a set of images, or [9] where correspondences between sets of interest points in a population of examples are obtained. 2) Work that integrates discrete optimization for the analysis of image populations: in [10] MRFs are used as an efficient way of encoding deformations for the registration of pairs of images, in [11] Sparse MRF Appearance Models (SAMs) localize objects and structures in images with shape and appearance models based on MRFs.…”

Groupwise Spectral Log-Demons Framework for Atlas Construction

“…The process of shape averaging [22,5,27] becomes particularly complex, and still remains unsolved, with organs undergoing large shape disparities. In the present state-of-the-art, the concept of geodesic shape averaging allows unbiased constructions of atlases through diffeomorphic methods [12,2,17], i.e., the transformation of a reference shape toward an average (the geometry of the atlas) follows a geodesic path on a Riemannian manifold (the space of diffeomorphic transformations).…”