Active-Contour-Based Image Segmentation Using Machine Learning Techniques

Abstract

Abstract. We introduce a non-linear shape prior for the deformable model framework that we learn from a set of shape samples using recent manifold learning techniques. We model a category of shapes as a finite dimensional manifold which we approximate using Diffusion maps. Our method computes a Delaunay triangulation of the reduced space, considered as Euclidean, and uses the resulting space partition to identify the closest neighbors of any given shape based on its Nyström extension. We derive a non-linear shape prior term designed to attract a shape towards the shape prior manifold at given constant embedding. Results on shapes of ventricle nuclei demonstrate the potential of our method for segmentation tasks.

Investigating existing medical CT segmentation techniques within automated baggage and package inspection

“…CT, MRI), and the anatomical structures reputed to be segmented. These methods can be broadly classified into two categories, namely (a) region-based methods which perform the segmentation by finding coherent regions according to some criteria [25][26][27][28][29][30][31], and (b) boundary-based methods that find the boundaries of the object of interest [32][33][34][35][36][37][38][39][40].…”

“…Boundary-based segmentation approaches range from the earliest and simplest threshold-based techniques [28,41,42] to the more sophisticated modelbased techniques such as active contours [32,33] and level sets [34][35][36][37][38] based on local gradients. The watershed segmentation method [39,40] can also be considered under this category since it uses the boundary edge of the object of interest.…”

Medical Image Segmentation using an Extended Active Shape Model

“…Machine learning algorithms have been used to improve the performance of image segmentation algorithms [1], [7], [8], [11], [15]. Li etal describes a machine learning approach for improving active shape model segmentation, which can achieve high detection rates [11].…”

“…Plath etal [7], proposed a machine learning method for multiclass image segmentation in which local and global evidences are coupled. They first formulate a Conditional Random Field that couples local segmentation labels in a scale hierarchy.…”

Contour-HOG: A Stub Feature based Level Set Method for Learning Object Contour

“…Thus, the lower dimensional representation can be learned and the energy function is minimized in the learned latent space. Etyngier et al [6] proposed a non-linear manifold learning method for learning shape prior. Specifically, a diffusion map is constructed using Nyström extension to learn shapes in a low-dimensional subspace.…”