AUTOMATED METHODS IN CLUSTERED MICROCALCIFICATIONS DETECTION MODULE OF A CAD SYSTEM

Abstract

Defining region of interests (ROIs) containing abnormal lesions on digital mammograms is the first step in many Computer-Aided-Diagnosis (CAD) systems for the classification of early signs of breast cancer as malignant or benign. The motivation of this paper is to study the efficiency of automated methods used in clustered microcalcifications (MCCs) detection module of a proposed CAD system. The proposed methods are based on several image processing concepts, such as morphological processing, fractal analysis, adaptive wavelet transform, local maxima detection and high-order statistics (HOS) tests. We applied these methods on a set of MIAS database mammograms. The mammograms consisted of two groups, which were cancerous (clustered MCCs) and non-cancerous (normal) and they were digitized at a size of 1024 by 1024 with 256 gray levels. The results showed that the efficiency of HOS test, fractal analysis and morphological approach were 99%, 92% and 74%, respectively. It was proven that the HOS test was the most efficient, and gave reliable results for every mammogram tested.