Breast Cancer Histopathology Image Classification Using Soft Voting Classifier

Abstract

Breast cancer is the reason of mortality rate among women worldwide. Breast cancer survival rate and mortality rate can be improved if it is identified at an initial stage. However, it is a very tough task to classify histopathology images. Therefore, computer-aided detection using various deep learning models is used to categorize the abnormalities in images. In this research, a soft computing classifier for 7-CNN model has been proposed for breast cancer histopathology image classification. The proposed methodology uses the basic CNN with four convolutional layers, basic CNN with five-layer CNN (with data augmentation), VGG 19 transfer learned model (with and without data augmentation), VGG 16 transfer learned model (without data augmentation) and Xception transfer learner model (with and without data augmentation). It uses seven models, and all the seven models have been used to make predictions. The datasets used in the research are hematoxylin–eosin (H&E) for experimentation. The performance of each model has been compared on the basis of accuracy, precision, recall and F1-score. The accuracy has been taken as the main evaluation criteria. The proposed methodology has achieved maximum accuracy of 96.91% on H&E dataset. The accuracy of proposed methodology has been compared with all the transfer learned models.