Hidden Markov tree model applied to the detection of micro-calcification clusters in mammograms

Abstract

This paper is concerned to the application of a relatively new image texture segmentation algorithm named Hidden Markov Tree (HMT) to the detection of micro-calcification clusters in mammograms. The HMT is a wavelet-based tree-structured probabilistic graph that can capture the statistical properties of the coefficients of the wavelet transform. The aim of this approach is, on the one hand, to take advantage of the wavelet coefficients in the characterization of different textures, and on the other hand, to link these coefficients by a tree structure enabling texture change to be detected. The application of the method was evaluated using the Digital Database for Screening Mammography (DDSM) for training purposes and a sample of the Nijmegen database for testing purposes.