MULTIPLE INSTANCE LEARNING FOR HISTOPATHOLOGICAL BREAST CANCER IMAGE CLASSIFICATION

Abstract

Cancer is a disease in which some of the body’s cells grow uncontrollably and spread to other parts of the body. The availability of proper screening methods are important for detecting initial symptoms. Also, A tumor can be malignant (cancerous) or benign (not cancerous). A benign tumor is usually not a serious problem unless it presses on a nearby structure or causes other symptoms. A benign tumor can become quite large, but it will not invade nearby tissue or spread to other parts of your body. A malignant tumor can spread to other parts of your body. Cancer can be in many types and forms. breast cancer forms in breast cells and its very common type of cancer in women. The kind of breast cancer depends on which cells in the breast turn into cancer. The main factors that influence your risk include being a woman and getting older. Most breast cancers are found in women who are 50 years old or older. Need of best screening method is very important to identify benign and malignant tumors. Histopathological images have very important place for identifying breast cancer. A weakly supervised learning called multiple instance learning is used for the computer aided diagnosis of cancer. Without having to label every instance, multiple instance learning involves grouping instances (pictures) into bags (patients). more modern ones, such a deep learning-based approach and a non-parametric approach (MIL-CNN) is used here. The non-parametric technique, which is one of the MIL methods, offers the best overall results and, in some situations, enables the achievement of classification rates that are not possible with traditional (single instance) classification frameworks. The tests are performed on the publicly available BreaKHis dataset, which consists of 82 patients’ microscopic biopsy images of 82 benign and malignant breast cancers. Above 95% accuracy lead to a better result