PathEnsemble-XAI: An Ensemble Deep Learning Framework with Explainable AI for Breast Cancer Classification

Breast cancer is still one of the most common types of cancer that kills people around the world, so we need accurate and reliable diagnostic tools to find it early and trustworthy treatment. This paper represents PathEnsemble-XAI, a novel ensemble-based deep learning framework that combines multiple convolutional neural network (CNN) architectures for automated breast cancer classification from histopathological images. Our approach integrates EfficientNet-B3, MobileNet-V2, DenseNet-121, and ResNet-50 models through probability averaging to leverage diverse feature representations and enhance classification robustness. The framework incorporates Local Interpretable Model-agnostic Explanations (LIME) to provide visual interpretability, addressing the essential requirement of transparency in medical imaging. Extensive research on the BreaKHis dataset show that our method works well, with a classification accuracy of 98.48% and precision, recall, and F1-scores of 98.50%, 97.96%, and 98.23%, respectively. Through transfer learning and mixed-precision training, the proposed method maintains computational efficiency while achieving competitive performance. LIME visualizations confirm that the model focuses on clinically relevant histopathological features, including cellular organization patterns and nuclear morphology, thereby enhancing clinical trust and interpretability. Our results establish a new benchmark for breast cancer classification on the BreaKHis dataset and demonstrate the potential of ensemble methods combined with explainable AI for reliable medical image analysis.