Customized Convolutional Neural Network for Breast Cancer Classification

The deep convolutional neural networks are most trustable and reliable approach to solve any kind of problems in these days. Especially, in recent days’ mammography image analysis uses deep neural network to identify the early stage cancer and it became most acceptable over other machine learning algorithms. These approaches are supporting radiologists to detect suspicious mass variability and other key characteristics in mammography images with good accuracy through greater performance. However, the performance and accuracy of these computer-aided breast cancer detection systems depends on factors, such as quality mammography images, selections of DNN architecture with appropriate hyper parameters. In this work, proposed a new customized convolutional neural network architecture to identify cancerous and healthy breast using mammography images obtained from digital database of screening mammography. The proposed model performed efficiently when compared with other previously proposed models in detection. Cancerous and healthy measured in terms of accuracy and area under curve (AUC) with optimized parameters and cost.