Optimizing Breast Cancer Diagnosis with VGG16 and CNN: An XAI Approach Using Saliency Maps

Breast cancer remains one of the most prevalent malignancies among women worldwide. Early detection is crucial for improving treatment outcomes, and deep learning techniques, particularly Convolutional Neural Networks (CNN), have demonstrated significant potential in medical imaging analysis. In this study, we utilized the MIAS Mammography Dataset, comprising 322 high-resolution mammographic images, to develop CNN-based models for breast cancer classification. The dataset includes 133 abnormal and 189 normal images, with abnormalities such as asymmetry (21 cases) and architectural distortion (22 cases). We implemented EfficientNetV2B0, EfficientNetB0, AlexNet, and VGG19 architectures to classify breast cancer into benign and malignant categories. To enhance model interpretability, we integrated Saliency Maps as an Explainable AI (XAI) tool, providing visual explanations for model predictions and improving transparency. Our approach achieved 98% accuracy using a modified VGG16 model, demonstrating its potential to advance medical imaging and support early detection strategies, ultimately benefiting patient care and treatment planning.