Discrimination between benign and malignant breast cancers in ultrasound images based on cost-sensitive boosting

Abstract

Breast cancer is one of the high-risk cancers, and breast ultrasound is routinely used as an adjunct to mammography for detection and diagnosis. Furthermore, the effective computer-aided diagnosis (CAD) system could improve the specificity of discriminating malignant from benign lesions on breast ultrasound images. This paper presents a method for discrimination between benign and malignant breast cancers in ultrasound images based on cost-sensitive boosting. Firstly, the image feature is extracted according to BI-RADS (Breast imaging report and data system), and a more simplified sub-feature set is obtained through minimal redundancy maximal relevance (mRMR) algorithm. Then three cost-sensitive Boosting models are trained and compared, and the optimal classification parameters are obtained by cross validation. Experiment shows that cost-sensitive AdaBoost performs the best, with AUC (area under receive operating characteristic curve) at 0.859 in the condition of controlled FNR (false negative rate) at 5%, better than CS-RealBoost and CS-LogitBoost. © 2012 Springer-Verlag.