%0 Journal Article
%T Breast Tissue 3D Segmentation and Visualization on MRI
%A Hong Song
%A Xiangfei Cui
%A Feifei Sun
%J International Journal of Biomedical Imaging
%D 2013
%I Hindawi Publishing Corporation
%R 10.1155/2013/859746
%X Tissue segmentation and visualization are useful for breast lesion detection and quantitative analysis. In this paper, a 3D segmentation algorithm based on Kernel-based Fuzzy C-Means (KFCM) is proposed to separate the breast MR images into different tissues. Then, an improved volume rendering algorithm based on a new transfer function model is applied to implement 3D breast visualization. Experimental results have been shown visually and have achieved reasonable consistency. 1. Introduction Recently, magnetic resonance imaging (MRI) technique has been widely used in diagnosing and detecting diseases. It provides an effective mean of noninvasively mapping the anatomy of a subject. It works better than X-ray computed tomography (CT) at soft tissue, such as breast. The three-dimensional segmentation and visualization of breast are useful for breast lesion detection and quantitative analysis. Segmentation is applied to extract the interesting tissues in the breast. Several algorithms have been developed for segmenting the breast tissues. Threshold-based method, the gradient method, polynomial approximation method, the active contour models, and classifier segmentation are used in breast skin segmentation. Raba et al. [1] summarized that threshold-based method, the gradient method, polynomial approximation method, the active contour models, and classifier segmentation are the main methods commonly used in breast skin segmentation. Chen et al. [2] introduced the fuzzy clustering algorithm to the tumor region segmentation which had achieved better results. Kannan et al. [3] made the breast region segmentation by introducing new objective function of fuzzy c-means with the help of hypertangent function, Lagrangian multipliers method, and kernel functions. However, these studies did not separate the fat and fibroglandular tissues. Pathmanathan [4] suggested a region-growing method, which required the user to manually choose one or more seed points. This method got satisfying results, but it is inefficient and time consuming. Nie [5] used two steps to segment the breast: firstly, locating the skin border and lungs region by standard FCM algorithm and secondly, extracting the fibroglandular tissue by an adaptive FCM algorithm. However, it is a semiautomated method. Two kinds of methods are mainly applied in volume visualization, which are surface rendering and volume rendering. For surface rendering, Marching Cubes (MC) algorithm [6] is usually used which was developed by Lorensen and Cline in 1987. MC represents 3D objects by surface representations such as
%U http://www.hindawi.com/journals/ijbi/2013/859746/