A Study of Machine Learning and DeepLearning Approaches for Breast Cancer Detection

The hallmark of cancer and related disorders is the uncontrolled growth of abnormal cellsthroughout the body, invading nearby tissues and eventually spreading to other organs. Unlikenormal cells, cancer cells have an unchecked growth cycle and can form tumors or metastasize(spread through the blood and lymph). The tumor's claw-like grip on tissues may have inspired theancient Greek term for "crab," which could explain the origin of the word "cancer." In breast cancer,a tumor develops when cells in the milk ducts or lobules of the breast multiply uncontrollably. Alump or tumor in the breast, along with other symptoms like skin or nipple changes, may resultfrom this uncontrolled growth. Breast cancer can stay confined within the breast or progress to aninvasive stage by spreading to other organs or tissues through the circulation. Although it primarilyaffects women, men can also develop breast cancer. This study focuses on diagnosing breast cancerusing ultrasound, mammography, and histopathological imaging, exploring current machinelearning and deep learning methods. We review the literature on various datasets (such as CBISDDSM, Breast Ultrasound Images dataset, and BreakHis), compare convolutional neural networkswith traditional machine learning classifiers, discuss some challenges in deployment and evaluation,and outline specific goals for future research aimed at enhancing dataset diversity, robustness,interpretability, and clinical validation.