Deep Learning Approach to Improve Breast Cancer Classification for Screening Mammography

Abstract

Identifying breast masses in breast cancer is a crucial step in the early diagnosis of breast cancer through mammography. However, the differential features between benign and cancerous masses in the initial stages of detection pose a persistent challenge. In response to the limitations faced by traditional convolutional neural networks, vision transformers are gaining importance as a promising approach. The transducer-based approach used in the study aims to overcome the inherent difficulties in distinguishing between benign and malignant masses by offering improved or comparable performance in classifying natural images. Within the framework of this study, a comprehensive comparison of vision transducer models was conducted to examine their potential to increase the accuracy and efficiency of breast cancer detection through mammography imaging.