Breast Tumor Classification in Digital Tomosynthesis Based on Deep Learning Radiomics

Abstract

Breast cancer is the most frequently diagnosed cancer in women globally. Early and accurate detection and classification of breast tumors are critical in improving treatment strategies and increasing the patient survival rate. Digital breast tomosynthesis (DBT) is an advanced form of mammography that aids better in the early detection and diagnosis of breast disease. This paper proposes a breast tumor classification method based on analyzing and evaluating the performance of various of the most innovative deep learning classification models in cooperation with a support vector machine (SVM) classifier for a DBT dataset. Specifically, we study the ability to use transfer learning from non-medical images to classify tumors in unseen DBT medical images. In addition, we utilize the fine-tuning technique to improve classification accuracy.