BCNet-11: A Dilated Convolutional Neural Network for Breast Cancer Classification Using Histopathology Images
Sadman Sakib, Arefin Ittesafun Abian, Most. Marufatul Jannat Mim, Ripon Kumar Debnath, Azizul Hakim Zen, Md. Rayhan Ahmed · 2025
Breast cancer is currently one of the primary causes of death for women. Early breast cancer detection and treatment significantly improve the survival rate of patients and lowers mortality rates. In recent years, models for the precise and effective detection of breast cancer have been created utilizing deep learning and machine learning approaches. In this study, we conduct experiments on an imbalanced BreaKHis dataset containing breast cancer histopathological images. This research proposes a convolutional neural network (CNN) based architecture named BCNet-11 to detect breast cancer that efficiently completes classification tasks using fewer computational resources without sacrificing accuracy. We employed image preprocessing to eliminate the hidden noise from the histopathological images, then data augmentation to balance the images from both classes. In our proposed architecture, we utilized dilated convolution to expand its receptive field and capture broader spatial contexts and image patterns. Various performance measurement metrics were used to determine the overall efficacy of the proposed BCNet-11 architecture. The proposed model&s;s test accuracy of 97.56% demonstrated its effectiveness. We evaluate the performance of our models to that of other transfer learning (TL) models, and it demonstrates its robustness by surpassing other TL models&s; accuracy. In order to improve patient outcomes and survival rates, the suggested model will help radiologists and doctors in early breast cancer diagnoses.