Segmentation of mammograms into distinct morphological texture regions

Abstract

This paper presents a comprehensive discussion on the segmentation of mammograms using morphological texture features. These features are derived from morphological granulometries with various structuring elements. Each structuring element captures a speci c texture content. The segmentation is carried out in an unsupervised manner by applying the KL transform feature reduction and Voronoi clustering on the extracted morphological texture features. The evaluation of the segmentation outcome by a trained radiologist is provided.

A Constant-time Algorithm for Erosions/Dilations with Applications to Morphological Texture Feature Computation

“…They form the computation engine of an``electronic second opinion'' system to allow computer prompts of textural abnormalities in digitized or digital mammograms [19,20]. …”

Classification of breast mass abnormalities using denseness and architectural distortion

“…Various mathematical tools have been used to define texture features. Co-occurrence matrix [13], fractal [14], wavelet [15], and granulometry [16], [17] based methods have been used to define texture features. Using such texture features, we found that there is a fair amount of overlap between the distributions of benign and malignant categories for a relatively large database considered in this work.…”

Texture analysis to characterize strain distribution in elastograms

“…In essence, morphological granulometry [11] consists of a sequence of morphological openings by a series of structuring elements given by kB = (k-1) B ⊕ B, where ⊕ indicates the morphological opening operator of a structuring element B. As a result, a sequence of opened images with decreasing volumes is obtained.…”

“…As a result, a sequence of opened images with decreasing volumes is obtained. The graph of the volume underneath opened images versus the increasing size of the structuring element is known as size distribution [11]. Moments of the normalized size distribution denote morphological texture features.…”

“…These inclusions simulate tissue abnormalities considering that many cancerous abnormalities are denser than benign ones. Image texture features [6] have been extracted and evaluated for a number of prevalent techniques including sum and difference histograms [7,8], fractals [9], wavelets [10], and more recently granulometric moments [11]. Texture features are computed globally over the entire inclusion to produce one feature value per inclusion.…”