The Research On Identification Technologies Of Benign And Malignant Breast Tumor Based On Convolutional Neural Network

World Health Organization surveys and statistics show that the incidence of breast cancer ranks first in all female malignant tumors,which seriously threatens the health of women.At present,the clinical manifestations of breast cancer are various,which makes it more difficult for doctors to diagnose.To reduce the mortality rate of breast cancer,the paper designs a benign and malignant mammary tumor classification model based on the current research of computer-aided breast cancer diagnosis system.The diagnosis provides a reliable basis and effectively improves diagnostic efficiency.The main works of this paper include:(1)Design a breast mass segmentation method based on kernel genetic fuzzy clustering algorithm to achieve accurate segmentation of breast mass area.The relevant data is obtained from the cooperative hospital.We construct the database and mark the benign and malignant of breast tumors under the guidance of doctors.This paper uses wavelet transform to remove the noise in the image,uses the genetic algorithm to optimize the clustering center of kernel fuzzy clustering algorithm and extracts the segmentation contour of breast mass.(2)A method for classification of malignant and benign breast tumors using convolutional neural networks based on migration learning is proposed to identify the benign and malignant of breast tumors.This method introduces migration learning into deep learning and optimizes the initial parameters of the network by using an open source big data set ImageNet to pre-train deep convolutional neural network,which solves the problem of the difficulty of overfitting of large-scale convolutional neural networks because of insufficient labeled mammographic X-ray images.Experimental simulations show that the accuracy of segmentation,over-segmentation and under-segmentation of breast X-ray image mass based on kernel fuzzy c-means clustering algorithm are 94.75%,0.27%,and 3.56%,respectively.Finally,the classification accuracy,sensitivity,specificity,and F1-Measure of benign and malignant breast tumors were 90.8%,89.8%,89.9%and 88.4%,respectively.The algorithm can provide a certain reference for the classification and diagnosis of breast tumors.