Research On Medical Assistant Diagnosis Of Breast Cancer Based On Machine Learning

The research on intelligent auxiliary diagnosis of breast cancer in clinical medicine is of great significance for further accurate diagnosis and treatment of cancer in clinical medicine.This paper proposes a new clinical diagnosis and evaluation system for breast cancer by combining clinical medical cooperation and taking the current clinical medical breast cancer slice as the research object.It uses computer vision,machine learning and deep learning to automatically detect and recognize the tumor area and internal cells on breast cancer slices,and on this basis,it has published and reviewed many papers and invention patents.This paper focuses on the following four aspects:(1)Aiming at the problem of insufficient and unbalanced data,we rely on data preprocessing technology to solve the impact of useless,damaged and redundant data on the whole data.The data enhancement technology is used to realize the data expansion of the public dataset and solve the problem of unbalanced data volume.Based on the real breast cancer slice data provided by the First Affiliated Hospital of China Medical University and the self-made data set,a lot of data annotation work was carried out when all types of breast cancer were known,and its data characteristics were analyzed.The principle of feature selection and extraction of deep learning network data was explained,so that the input data could better meet the data types required by the deep learning network model;Secondly,DC-GAN(Deep Convolution Generation Countermeasure Network)is used to synthesize or generate new in situ cancer images to expand the data set,and Siamese Network is used to measure the similarity between the real data set and the synthetic data set,which is better considered and studied from the perspective of protecting patient privacy.(2)In view of the difficulties in identifying tumor cells on breast cancer slices in current clinical medicine,the cells are dense and numerous,the cell count is difficult and the judgment is inaccurate,based on the real breast cancer slice data provided by the Pathology Department of the First Affiliated Hospital of China Medical University and the self-made data set,combined with the clinical experience and knowledge of pathologists,Computer vision and machine learning are used to automatically detect,recognize,classify and count tumor cells on breast cancer slices,realizing the goal of clinical medicine to identify breast cancer cells and breast cancer influencing factor ER.(3)To solve the problem that breast cancer is difficult to classify,the deep learning networks Faster-RCNN,SSD300,Retina Net,YOLOV3,YOLOV5,Swin-Transformer were applied to the diagnosis of breast cancer,to differentiate ductal carcinoma in situ(DCIS)and invasive ductal carcinoma(IDC),and to detect their targets.Through experiments,relevant experimental indicators and visual results were obtained.At the same time,the diagnostic results of various deep learning neural networks were compared,The optimal model is selected,and the YOLOV5 deep learning network model has the best target detection effect.(4)Aiming at the immunohistochemical staining of ER/PR of breast cancer in clinical medicine,the trained deep learning network model was called to conduct a complete and accurate diagnosis of breast cancer,and a clinical medicine three ring study was conducted to evaluate the immunostaining image and evaluate the role of breast cancer artificial intelligence detection system(DSBC).On this basis,the clinical medicine breast cancer diagnosis system interface(GUI)was developed.