A Classifier Ensemble Method for Breast Tumor Classification Based on the BI-RADS Lexicon for Masses in Mammography

Abstract

A computer-aided diagnosis (CAD) system is a tool to assist clinicians in interpreting medical images. In mammography, CADs provide a classification of tumors to distinguish between benign and malignant cases, aiming to support the clinical conduct. Nevertheless, CADs disregard informing about the internal criteria utilized to classify breast tumors, particularly, compatible with the Breast Imaging-Reporting and Data System (BI-RADS). In this context, we propose a new scheme of tumor classification based on the BI-RADS lexicon for masses. The terms of shape, margin, and density are modeled using specific feature sets to provide different perspectives of the tumor in terms of benign and malignant findings. The outcomes of the three models are further used for the final histopathological classification of the tumor. The proposed method is compared with two conventional CAD systems that classify tumors using a single feature set. The results show that the proposed method obtains 90% accuracy, whereas the two conventional CADs reach an accuracy of 89% and 76%. Therefore, the proposed method is suitable for the histopathological classification of tumors by using the information provided by the three models of the BI-RADS lexicon for masses.