Breast Tumors Multi-classification Study Based on Histopathological Images with Radiomics Approach

Abstract

Abstract Breast cancer is the most common malignant tumor in women. It has important clinical significance for the automatic classification of breast tumors, and the current research mainly focuses on the benign and malignant classification of breast tumors. In this paper, we proposed a radiomics method for multi-classification of breast cancer. By the radiomics method, 212 features were extracted for quantifying breast tumor images’ intensity, color and texture and a multi-classification diagnosis model of breast tumors was constructed by support vector machines (SVM). The breast tumors were divided into eight categories, these eight categories include adenosis, fibroadenoma, phyllodes tumor, tubular adenoma, ductal carcinoma, lobular carcinoma, mucinous carcinoma, and papillary carcinoma. Final, the classification accuracy reached 90.3%. The radiomics approach provides an auxiliary role for developing the best treatment plan for breast tumors.