Cross Modality Deformable Segmentation Using Hierarchical Clustering and Learning

Abstract

Abstract. Segmentation of anatomical objects is always a fundamental task for various clinical applications. Although many automatic segmentation methods have been designed to segment specific anatomical objects in a given imaging modality, a more generic solution that is directly applicable to different imaging modalities and different deformable surfaces is desired, if attainable. In this paper, we propose such a framework, which learns from examples the spatially adaptive appearance and shape of a 3D surface (either open or closed). The application to a new object/surface in a new modality requires only the annotation of training examples. Key contributions of our method include: (1) an automatic clustering and learning algorithm to capture the spatial distribution of appearance similarities/variations on the 3D surface. More specifically, the model vertices are hierarchically clustered into a set of anatomical primitives (sub-surfaces) using both geometric and appearance features. The appearance characteristics of each learned anatomical primitive are then captured through a cascaded boosting learning method. (2) To effectively incorporate non-Gaussian shape priors, we cluster the training shapes in order to build multiple statistical shape models. (3) To our best knowledge, this is the first time the same segmentation algorithm has been directly employed in two very diverse applications: a. Liver segmentation (closed surface) in PET-CT, in which CT has very low-resolution and low-contrast; b. Distal femur (condyle) surface (open surface) segmentation in MRI.

Deformable Segmentation via Sparse Shape Representation

“…In offline learning, 3D volume images along with the manually labeled ground truths are employed to learn the appearance and shape characteristics of the organ under study. More specifically, methods proposed in [10,9] are used to learn landmark detectors and a set of spatially adaptive boundary detectors. Meanwhile, organ surfaces are stored in a shape repository, which will be exploited to derive shape priors during runtime.…”

“…As discussed before, although learning-based landmark/boundary detectors can tackle reasonable appearance variations [10,9], they might generate wrong responses in the presence of severe imaging artifacts/diseases, and hence mislead the deformable model. In this scenario, shape prior is the only information source to initialize/correct the deformable surface.…”

“…Since one-to-one correspondences are already constructed among all shapes, affinity propagation only needs to perform once for the model shape. The similarity used in the affinity propagation is defined as the combination of the image similarity and geodesic distances between vertices [9].…”

Deformable segmentation via sparse representation and dictionary learning

“…This will be the major topic in the remainder of this paper. (For more details of appearance modeling, refer to Zhan et al (2008Zhan et al ( , 2009Zhan et al ( , 2011. )…”

“…Although learning-based landmark/boundary detectors can tackle reasonable appearance variations (Zhan et al, 2008(Zhan et al, , 2009, they might generate wrong responses in the presence of severe imaging artifacts/diseases, and hence mislead the deformable model. In this scenario, shape prior is the only information source to initialize/correct the deformable surface.…”