BioHisCNet-A Lighter Classification Network for Prediction of Breast Cancer

Biopsy is a conclusive procedure to procure an accurate and detailed tissue diagnosis for breast cancer. Breast cancer detection and prevention are crucial due to its status as the most common cancer among women globally. This study uses deep learning (DL) to automate and enhance diagnostic accuracy by classifying biopsy images as benign or malignant. We utilized the Breakhis dataset for this analysis. Synthetic minority oversampling technique (SMOTE) has been applied to increase the performance of the model. A custom convolutional neural network (CNN) model with seven layers, including both convolutional and fully connected layers, a lighter network was developed from scratch. Our model achieves an efficacy of 92.54%, surpassing the performance of existing models such as ResNet-150, AlexNet, VGG-16, and Inception-V3 which are pretrained on the imagenet dataset.