Convolution Neural Network based Approach for Breast Cancer Type Classification

Breast cancer is ranked high amongst the main causes of death that badly effected women worldwide. Cancer diagnosis is commonly carried out from Hematoxylin and Eison (H&E) stained breast histopathology images. However, possible varying conditions in slide preparation and stain color variations in H&E images could induce the chances of misdiagnoses. Therefore, in this paper, we have proposed a new deep learning based method for breast Hematoxylin and Eison (H&E) stained breast histopathology images. In pipeline of the designed approach color normalization method and sample extraction strategy are explained which are used for pre-processing of H&E histology images. Further deep feature computation with deep convolution neural network (CNN) & classification with various deep classifiers is also explained. The performance of designed approach is tested on publically available challenging breast histological BreakHis dataset. Results proved that the presented system has good potential for cancer type classification and outperforms the existing methods.