Mass Detection in Digital Mammogram Image using Convolutional Neural Network (CNN)

The implementation of convolutional neural networks (CNN) in medical imaging has become very favorable nowadays especially in mammography. CNN is capable of evolving an automatic mass detection system that plays an important role in aiding the radiologist to makes an accurate diagnosis as well as increase recalls back the patient to further being investigated. Thus, in this paper, a revolutionary computer-aided system with entirely automated detection scheme in digital mammogram is proposed. This proposed CAD framework consist of three fundamental phases such as preprocessing of mammogram images, mass detection, as well as classification of mass into three category such as benign, malignant, and normal. We utilized the authentic version of 322 mammograms images from MIAS database and its augmented mammograms image in testing and training the proposed system using CNN. At first, the CNN is trained using the large augmented database. After that, the model is transferred and tested onto the smaller database which is the original database. Three usually used CNNs such as VGG19, InceptionV3, and MatConvNet is evaluated in this study. As a result, the proposed CAD system able to detects the mass position with overall accuracy of 97.04%. This proved that the use of CNN in this study is applicable and feasible to be used by the radiologist in helping them detecting and classifying breast mass in digital mammogram image.