A Data-Driven Hybrid ML–XAI Approach for Early stage Breast Cancer Screening

Abstract

Breast cancer continues to be the most common type of cancer in women globally and poses a serious threat to global health. Early detection remains a crucial factor in determining survival rates, even with the advancements in imaging technologies. Recent years have seen impressive advancements in the detection, diagnosis, and prognosis of breast cancer thanks to machine learning (ML) and deep learning (DL) techniques. However, as the majority of ML models function as "black boxes" with little transparency, a significant obstacle to clinical application is the lack of interpretability. A promising strategy to close this gap is Explainable Artificial Intelligence (XAI), which provides patients and physicians with understandable insights into algorithms that make decisions. A thorough summary of hybrid ML–XAI architectures for early breast tumor detection is given in this review. We talk about XAI techniques like LIME, SHAP, and Grad-CAM, investigate how they are integrated into hybrid frameworks, and look at the function of ML in mammography, ultrasound, and histology. The study also critically assesses the body of research, identifies present issues, and suggests future paths for the application of reliable, interpretable, and high-performing AI in breast cancer screening. Hybrid ML–XAI frameworks offer a revolutionary route toward dependable, moral, and clinically applicable AI-driven healthcare by fusing interpretability and accuracy.

Keywords: Breast Cancer, Machine Learning, Explainable Artificial Intelligence, Early Screening, Hybrid Frameworks, Medical Imaging.