Histopathology Breast Cancer Classification Using CNN

Abstract

Breast cancer is the second leading cause of cancer mortality among women, early detection and treatment of breast cancer is very critical for the reduction of mortality rate. Cancer, known for its devastating impact, poses significant challenges in early detection within the medical field. In this study, breast cancer image samples were utilized for investigation purposes from BreakHis dataset. For accurate identification, current deep learning approaches relies on contour information. However, the cancer detection approach which is already existed retrieved only information which is semantic from the first layer and did not offer insights on the semantic levels which are not deep in breast cancer detection. Deep learning is a highly effective technique that possesses the capability to extract and structure significant insights from data, thus eliminating the necessity for domain experts to manually design feature extractors. Convolutional Neural Networks (CNNs), which belong to the category of deep, feed forward networks, have gained significant interest from researchers and industries. The recognition achieved by CNNs can be attributed to their integration in various domains such as transfer learning, speech recognition, natural language processing, signal processing, and object recognition. The proposed framework incorporates an effective and efficient strategy derived from a comprehensive review of commonly used approaches in each category. In the context of cancer screening, the process involves classifying image biopsies as either malignant or benign. Medical professionals like doctors and pathologists highly rely on the analysis of microscopic biopsy images to identify and categorize different abnormalities.