Breast Cancer Analysis using Convolutional Neural Network

Abstract

Breast cancer is one of the crucial reasons for deaths in females. It affects one out of eight females worldwide. Breast cancer can be detected in early stages with the help of mammography. It is possible due to advancement of science and medical field; more reliable and accurate techniques have emerged to fight against this disease. Nowadays deep learning approach is being used by radiologists which help them make accurate diagnosis of breast cancer. This research contains a novel way of breast cancer detection using convolutional neural network and mammogram imaging system, to accurately classify mammogram image of tumor into benign (noncancerous) and malignant (cancerous). A proposed custom model is created which Is in resemblance with the VGG 16 model. Several mammogram images are used to carry out preprocessing. In order to get good results, we use preprocessing methods such as shearing, enlargement and equalizing image data are used. Feature extraction is done through CNN and classification is performed in fully connected network. The outcome described here demonstrates that the accuracy rate of the proposed automated method is better than other existing methods. Experimental results show the accuracy of the proposed method is 99.45% on training data. Classification report gives the prediction accuracy of 99% with good precision, recall and Fl score.