A region and gradient based active contour model and its application in boundary tracking on anal canal ultrasound images

Abstract

A novel Gaussian mixture model (GMM)-based region and gradient active contour model is proposed for general object boundary tracking and the purpose of boundary tracking for anal muscle layers. Motion information is extracted from adjacent slice and is used to guide the first step of boundary tracking procedure. The idea is that GMM is introduced into the statistical feature computation for object region and background region, thereby providing an accurate model for regional pixel intensity description. Expectation-maximization algorithm and K-means algorithm are used for parameter solutions of GMM. Based on the available region information, gradient information and the self-constraints of the contour, a unifying active contour model is proposed. The proposed active contour models and tracking algorithm were tested on synthetic images and simulated ultrasound images to evaluate some generic features of the model for boundary tracking. Furthermore, it was applied to perform boundary tracking of anal muscle layers. The tracking results were evaluated by three experts. The results showed that the proposed method has a good performance for boundary tracking on anal wall ultrasound image. © 2007 Pattern Recognition Society.