Current imaging techniques for histopathology-based breast cancer diagnosis: a review with comparative insights from machine learning and deep learning models, challenges, and future directions
Abstract
Breast Cancer is one of the leading causes of mortality among women worldwide, making early and precise diagnosis vital for enhancing patient survival rates. However, the manual interpretation of biopsy slides can be subjective, labor-intensive, and prone to observer differences. ML (Machine Learning) and DL (Deep Learning) approaches have become well-known in the study of histology images in recent years, as they provide strong new approaches to automatically enhance image analysis. Innovative ML and DL techniques used to identify breast cancer based on histopathological images are reviewed in this study. This study covers feature extraction, feature selection, classification, and medical image analysis. The performance assessment encompassed filter-based, wrapper-based, hybrid, and ensemble learning models. The benefits of AI-driven approaches, such as enhanced diagnostic accuracy, scalability, and repeatability, were investigated along with their drawbacks, such as data imbalance, lack of interpretability, and computational complexity. In addition, this analysis highlights publicly accessible datasets, benchmark models, and recent advancements in XAI (Explainable Artificial Intelligence), ViT (Vision Transformers), and Federated Learning. This highlights possible directions for future research and provides researchers, doctors, and developers with an understanding of how intelligent systems are changing the histopathological diagnosis of breast cancer.