Learning Coupled Prior Shape and Appearance Models for Segmentation

Abstract

Abstract. We present a novel framework for learning a joint shape and appearance model from a large set of un-labelled training examples in arbitrary positions and orientations. The shape and intensity spaces are unified by implicitly representing shapes as "images" in the space of distance transforms. A stochastic chord-based matching algorithm is developed to align photo-realistic training examples under a common reference frame. Then dense local deformation fields, represented using the cubic B-spline based Free Form Deformations (FFD), are recovered to register the training examples in both shape and intensity spaces. Principal Component Analysis (PCA) is applied on the FFD control lattices to capture the variations in shape as well as on registered object interior textures. We show examples where we have built coupled shape and appearance prior models for the left ventricle and whole heart in short-axis cardiac tagged MR images, and used them to delineate the heart chambers in noisy, cluttered images. We also show quantitative validation on the automatic segmentation results by comparing to expert solutions.

Prior-based Segmentation and Shape Registration in the Presence of Perspective Distortion

“…The translation and rotation (R, t) between the image planes, and the scene structure (n, d), are recovered by decomposition of the known homography matrix (see [15,18,27] and references therein). In contrast, our novel approach calculates the homography directly in its explicit form (19). Rather than relying on point correspondence, we match two corresponding contours of the shape of interest using calculus of variations.…”

“…The statistical methodology [4,9,19,24,25,35,39] accounts for transformations beyond similarity and for small non-rigid deformations by using a comprehensive training set. It characterizes the probability distribution of the shapes and then measures the similarity between the evolving object boundary (or levelset function) and representatives of the training data.…”

FEM-based automatic segmentation of muscle and fat tissues from thoracic CT images

“…To better capture specific shape deformations, we extend the proposed FEM-based framework through the incorporation of a statistical deformation model (SDM) [8] learned from training data. There exist a few works on templatebased segmentation which follow a SDM constrained registration framework [9], [10], but these methods also suffer from the disadvantage of using a uniform mesh. The main contributions in this work are given below:…”

Shape-Based Mutual Segmentation

“…The statistical approaches, e.g. [3,5,12,18,29,33], capture possible shape variability by employing a set of similar but not identical shape priors. These methods, however, depend on the availability of a comprehensive set of priors or a segmented instance of the object of interest.…”