A comparative study of convolutional neural network features for detecting breast cancer

Breast cancer is the most deadly type of cancer among women survival rate of which can be increased with early diagnosis treatment. A computer-aided diagnostic as an auxiliary system in field can automatically detect abnormalities without wisdom of radiologists. The aim of this thesis is directly the detection of breast cancer by applying a deep learning due to their great achievements in the classification of picture-based objects.The data that is included 322 mammography images is used to classify the tumor by applying a new breed of deep learning algorithms- convolutional neural networks (CNNs).Compared with the CNNs in detection of breast cancer, the proposed methods out performs AlexNet by 48.2% , VGG16 by 72.2% ,ResNet50 by 65.3% , NasNet by 65.3% and Inception ResNetV2 by 60% in terms of accuracy rate. A well-designed convolutional neural network, as a promising technique for breast cancer systems, will certainly play more efficient role in existing systems.