Morphological filtering and stochastic modeling-based segmentation of masses on mammographic images

Abstract

The objective of this study is to develop an efficient method to highlight the geometric characteristics of mass patterns, and isolate the suspicious regions which in turn provide the improved segmentation of suspected masses. In this work, a combined method of using morphological operations, finite generalized Gaussian mixture modeling, and contextual Bayesian relaxation labeling was developed to enhance and segment various mammographic contexts and textures. This method was applied to segment suspicious masses on mammographic images. The testing results showed that the proposed method can detect all suspected masses as well as high contrast objects and can be used as an effective pre-processing step of mass detection in computer-aided diagnosis systems.

Quantification and segmentation of brain tissues from MR images: a probabilistic neural network approach

“…In the first one, the maximum likelihood quantification scheme, tissue types are first quantified using maximum likelihood principle, where only soft classification of the pixel images is required [1]. Further classification of a sample is then performed by placing it into the class for which the posterior probability or the support function is maximum, i.e., by Bayesian consistent labeling [45], [46]. The quantities obtained by sample averages after imperfect pixel classification may not be consistent with the previous quantification result [23].…”

“…It operates on an initial segmented image, preferably one with uniformly distributed classification errors, such as the one segmented by the classification-maximization (CM) algorithm [29]. PCRN uses stochastic discrete gradient descent procedure where each pixel is randomly visited and its label is updated [16], [45], i.e., pixel i is classified into the kth region if…”