Design And Research Of Mammography Assisted Classification System Based On DCCNN

With the pace of life and the environment constantly changing,breast cancer with high morbidity and mortality has become the most common type of cancer,which is seriously threatening women's physical health and life safety.Due to the uncertainty of the pathogenesis of breast cancer,it is difficult to detect breast cancer in the early stage of the disease.Mammography is the preferred imaging method for breast cancer screening,and also the simplest and most reliable method.Therefore,this paper takes mammography image data as the research object.The research on the detection and classification of lesions in breast images is designed to assist doctors in making faster and more accurate diagnosis,which is of great significance for the early diagnosis and treatment of breast cancer.Mass and calcification are the main positive signs in mammography,but most of the methods used to detect and classify them only focus on mass or calcification and do not take both into account.At the same time,the existing methods also have problems such as low detection accuracy and low classification accuracy of the lesions,which cannot provide doctors with information that is helpful for diagnosis.With the development of science and technology,artificial intelligence technology has become more and more widely used in breast cancer diagnosis.Not only can it help doctors deal with some boring and repetitive tasks,but it can also provide objective results for radiologists to assist in diagnosis.Therefore,this paper uses deep learning methods to detect and classify lesions in mammography images.Clinical breast lesions include not only masses,but also calcifications.In order to improve the accuracy of breast cancer screening,this paper considers both mass and calcification in the breast.First,a deep learning network Faster R-CNN is selected for target detection,and the loss function of the network is optimized,which could not only improve the detection accuracy of molybdenum target images,but also save the time of model training and testing.Then,according to the characteristics of mammography images,a Deep Cooperation Convolutional Neural Network was designed to classify the lesions.Multiple views are used as input to integrate the features of different layers,and in order to simultaneously consider the characteristics of channel and spatial dimensions,attention mechanism is introduced to divide the breast lesions into five more detailed categories,namely: normal breast,benign calcification,benign mass,malignant calcification and malignant mass.The design is able to balance the depth and width of the network.This paper use a dataset train the network,and the data set consists of 695 normal cases from the Digital Database for Screening Mammography(DDSM),753 calcification cases and 891 mass cases from the Curated Breast Imaging Subset of DDSM(CBIS-DDSM).Compared with the existing models,the test results show that the classification accuracy of the network model designed in this paper is higher.