Identification of Breast Cancer from Histopathological Images Using Deep Convolutional Neural Networks

Abstract

Breast cancer symbolizes the disease of uncontrolled growth of cells of the breast. There are 2 most common kinds of breast cancers we know about, (1) “Invasive lobular carcinoma” and (2) “Invasive ductal carcinoma”. There are almost 1.3-1.5millions of patients alone in Bangladesh, who are affected by breast cancer. Every year almost 0.2millions of patients, newly diagnosed with breast cancer. Among these two, 80% of cases are Invasive ductal carcinoma. In this work, a deep CNN approach is proposed to predict Invasive Ductal Carcinoma (IDC) from histopathological data using “Convolutional Neural Network” which is a state of the art machine learning algorithm. We use Breast Histopathology Images (198,738 IDC(-) image patches; 78,786 IDC(+) image patches) taken from Kaggle. We took 3 transfer learning approaches using VGG16, Inception V3, Inception ResNet V2 and one without transfer learning approach. Their final training accuracies are 77%, 89% , 88.5%, and 87% respectively