A microcalcification detection using multiple support vector machines in digital mammogram

Abstract

A computer-aided diagnosis (CAD) system has been examined to reduce the effort of a radiologist. In the mammogram, it is helpful to improve the diagnostic accuracy of malignant microcalcifications in an early stage of detecting breast cancer. In this paper, we propose the algorithm using multi-layer Support Vector Machine (SVM) classifier to discriminate whether microcalcifications are malignant or benign tumors. SVM is a technique that is easier than the artificial neural networks (ANN) to use. Moreover, it can obtain good performance even though using a small number of training data sets. The proposed method to detect microcalcifications is composed of two-layer detection steps each of which uses SVM classifier. The first step is coarse detection. It can find out pixels of the regions considered as microcalcifications. Then, ROI (region of interest) is generated based on microcalcification characteristics. The next step is fine detection step to determine whether the found ROIs are microcalcifications or not by merging potential regions using ROI obtained from the former step and SVM classifier. The proposed method is specified on Korean mammogram database. The experimental result of the proposed algorithm is more robust in detecting microcalcifications compared with the previous method using ANN as classifier.