Histopathological Image Classification of Breast Cancer Using Kervolutional Neural Networks

Abstract

Histopathological Image Classification is a standard for diagnosing cancer. The classification helps in determining the best treatment among various treatment methods available. Breast Cancer classification are primarily constructed on histopathological photographs of the tissue in the tumor. In this project, we classify the histopathological images belonging to two major categories of tumor Benign and Malignant using KNN (Kervolutional Neural Network— Kernel Convolution Neural Network). Existing works using CNN mainly leverages activation layers as it only provides point-wise non-linearity, so we use KNN over CNN which provides indefinite complex actions of the human recognition system by making use of the kernel trick. It is a generalized version of convolution which can enhance the model’s extent and can capture higher order of traits using reinforcement kernel functions short of any added parameters.