Hierarchical matching for automatic detection of masses in mammograms

The computer-aided detection (CAD) system plays a vital role in early cancer diagnosis and thus aids in reducing the mortality rate of breast cancer. In this study, a hierarchical template matching method is proposed to detect the breast mass in mammograms. First, three types of templates are designed and compared, and the sech template is selected to match the mass. Then, a hierarchical matching method, which makes full use of the high sensitivity of the local part-based template and the low false positive rate of the complete template, is used to detect the mass. At last, to eliminate some of flase positives, a set of rules are set based on various features. The performance of this proposed method is evaluated by 368 mammograms selected from the digital database of screening mammography. The sensitivity and the false positive rate are 96.2% and 5.2 per image, respectively. This preliminary study indicates that the proposed method is a promising strategy, and it has great potential in improving the accuracy of CAD systems in interpreting mammograms.