Breast Cancer Classification and Predicting Class Labels Using ResNet50

Abstract

Numerous studies have been conducted using Deep Learning paradigms to detect Breast Cancer. Breast cancer is a medical condition where abnormal cells in the breast grow uncontrollably, forming tumors. If not treated, these tumors can metastasize and spread to other parts of the body, potentially leading to life-threatening consequences. In the year 2020, there were approximately 2.3 million cases of breast cancer diagnosed in women, leading to around 685,000 deaths worldwide. By the close of 2020, there existed 7.8 million women who had been diagnosed with breast cancer within the preceding five years, solidifying it as the most widespread form of cancer globally. Breast cancer is observed across all countries, affecting women at various ages post-puberty, with incidence rates tending to rise in older age groups. The aim of this paper is to classify and predict the class labels of breast cancer. To achieve this, a ResNet50 model is utilized and mammography images are employed to locate cancer within the image and classify it to emphasize the affected area. The ResNet50 identifies mass regions and classifies them as either ductal carcinoma, inflammatory, triple negative or invasive cancer. The experimentation is carried out on breast cancer dataset and achieved 90.6% accuracy both for classification as well as prediction