Breast Cancer Classification Using Machine Learning

Abstract

In the pursuit of precise forecasts in machine learning-based breast cancer categorization, a plethora of algorithms and optimizers have been explored. Convolutional Neural Networks (CNNs) have emerged as a prominent choice, excelling in discerning hierarchical representations in image data. This attribute renders them apt for tasks such as detecting malignant lesions in mammograms. Furthermore, the adaptability of CNN architectures enables customization tailored to specific datasets and objectives, enhancing early detection and treatment strategies. Despite the efficacy of screening mammography, the persistence of false positives and negatives poses challenges. Computer-Aided Design (CAD) software has shown promise, albeit early systems exhibited limited improvements. Recent strides in deep learning offer optimism for heightened accuracy, with studies demonstrating comparable performance to radiologists. Nonetheless, the detection of sub-clinical cancer remains arduous, primarily due to small tumor sizes. The amalgamation of fully annotated datasets with larger ones lacking Region of Interest (ROI) annotations is pivotal for training robust deep learning models. This review delves into recent high-throughput analyses of breast cancers, elucidating their implications for refining classification methodologies through deep learning. Furthermore, this research facilitates the prediction of whether cancer is benign or malignant, fostering advancements in diagnostic accuracy and patient care.