Deep Learning for Predicting Breast Cancer: A Systematic Review of Progress and Future Directions

Abstract

Abstract Breast cancer prediction is a critical area of research aimed at improving early detection and enhancing treatment strategies. Considering the fast development of Machine Learning techniques, the level of curiosity has increased dramatically in leveraging these algorithms for accurate and efficient breast cancer prediction. This survey paper comprehensively overviews the present condition of the art Machine Learning approaches employed in breast cancer prediction. This study analyzed a wide range of research studies, methodologies, and datasets to present a complete image of the state of the field, the problems it faces, and where it's going. Diverse techniques for Machine Learning, including deep learning models, SVMs, random forests, ANNs, and ensemble methods, are explored in terms of their strengths, weaknesses, and specific breast cancer prediction tasks they have been applied. Furthermore, the study also discussed the diverse input data modalities used, ranging from traditional mammograms and histopathological images to genomics and proteomics data. Challenges such as dataset imbalance, feature selection, interpretability, and generalizability are examined, along with proposed solutions and prospective directions for research. This survey paper aims to give a wealth of information for scientists, doctors, and others in the healthcare field to understand the advancements and potential of predicting breast cancer with Machine Learning, contributing to the development of improved precision and dependable predictive models for improved patient outcomes in the battle against breast cancer.