CNN-Based Automated Detection of Metastatic Cancer in Histopathology Images
Omaia Mohammed Al-Omari, Omar Alkhatib, Tariq Al-Omari · Engineering Technology & Applied Science Research · 2025
Breast cancer remains one of the leading causes of cancer-related mortality, underscoring the critical need for accurate and efficient diagnostic solutions. This study presents an enhanced Deep Learning (DL) framework for the classification of breast cancer histopathological images, integrating both advanced Convolutional Neural Network (CNN) architectures and explainability techniques. The proposed approach utilizes the publicly available BreaKHis dataset, which contains over 7,900 histopathological images of benign and malignant breast tumors captured at varying magnification levels (40×, 100×, 200×, and 400×). An EfficientNetB3-based CNN is employed for automated feature extraction and classification, addressing the limitations of traditional Machine Learning (ML) methods that rely on handcrafted features and typically suffer from reduced generalizability. The proposed model significantly outperforms conventional classifiers, including Random Forest (RF) (74.56%), Support Vector Machines (SVM) (77.34%), and k-Nearest Neighbors (k-NN) (69.67%), by achieving a test accuracy of 92.33% on the BreaKHis dataset. To enhance model transparency and clinical relevance, Gradient-weighted Class Activation Mapping (Grad-CAM) is incorporated, which accurately localizes malignant regions in over 95% of test samples, offering visual interpretability of model predictions. Additionally, dimensionality reduction techniques, such as t-distributed Stochastic Neighbor Embedding (t-SNE) and Principal Component Analysis (PCA) are employed to analyze the feature space. These analyses reveal improved separability between benign and malignant clusters, further validating the effectiveness of the learned representations. The results of this work demonstrate the transformative potential of DL, particularly EfficientNet-based CNN architectures, in delivering both high diagnostic accuracy and interpretability, paving the way for more reliable and explainable Artificial Intelligence (AI)-assisted diagnostic systems in histopathology.