DL-Driven Model for Accurate Detection of Breast Cancer from Histopathology Images: Clinically Validated on Algerian Medical Data

Accurate and early detection of high-risk breast cancers significantly affects treatment success and patient prognosis. Invasive Ductal Carcinoma (IDC) is the most prevalent and aggressive subtype, highlighting the need for improved diagnostic strategies. Motivated by this clinical challenge, we propose a deep learning-driven model for automatic IDC detection from histopathological images, aimed at assisting pathologists and enhancing diagnostic precision. The model extends the EfficientNetB0 architecture with additional domain-specific convolutional layers and is trained on the publicly available Breast Histopathology Images dataset. The model achieved strong performance, with an AUC of 96.13 %, accuracy of 89.26 %, recall of 90.89%, and F1-score of 82.54 %. Its clinical relevance was further assessed through validation on real Algerian histopathological slides, where the model reached 95.46\% accuracy under routine laboratory conditions. These results demonstrate the potential of our clinically validated deep learning model as a reliable tool for IDC detection and breast cancer diagnosis, bridging the gap between AI research and practical medical application.