A high-sensitivity computer-aided system for detecting microcalcifications in digital mammograms using curvelet fractal texture features

Abstract

A high-sensitivity computer-aided system for detecting microcalcifications in mammographic images requires an efficient feature extraction stage and an enhanced pattern recognition technique for classification. In this paper, a new and an efficient texture feature extraction method, curvelet-based fractal texture analysis is proposed. The system consists of two main stages. In the first stage, the suspicious microcalcification regions are separated from the normal tissues using curvelet layers from which the fractal dimensions are computed to describe the decomposed and oriented texture patterns. The decomposition of the input image is done using the curvelet layers. In the second stage, an ensembled fully complex-valued relaxation network classifier is used for classifying mammograms. The proposed system exhibits superior performance in terms of high true positive rate and low false positive rate, in comparison with the existing techniques. The experimental results yielded a classification accuracy of 98.18%, which indicates that curvelet fractal is a promising tool for analysis and classification of digital mammograms.