Application Of Convolutional Neural Network And Improved Algorithm In Ovarian Cancer Image

Ovarian cancer causes many female compatriots to be tortured every year.With the deterioration of modern living environment and many other factors,the probability of women getting ovarian cancer is increasing year by year.The traditional diagnosis process is that the doctor judges whether he has a malignant tumor according to the pathological section of the patient.Even experienced doctors may sometimes misdiagnose medical images,so hope to use computer technology to help doctors diagnose and minimize misdiagnosis.In the field of medical image processing,compared with traditional image classification methods,image classification based on convolutional neural networks not only enables autonomous feature learning of medical images,but also improves the accuracy of image recognition.This article aims to explore the use of convolutional neural network algorithm to improve ovarian cancer image recognition,the main work is as follows:1.Pre-process the ovarian cancer medical image:cut the pathological picture of ovarian cancer to select the required picture;enhance the data of the ovarian cancer picture to solve the problem of data imbalance.2.Using SIFT+SVM and LBP+SVM traditional image classification methods and convolutional neural networks to compare the classification experiments of ovarian cancer images,it was found that convolutional neural networks have obvious advantages over traditional methods;using convolutional neural network model AlexNet,VGGNet,and GoogLeNet compared the classification experiments of ovarian cancer images.3.In view of the shortcomings of the AlexNet model,the activation function,pooling method,network structure adjustment,classifier selection and other aspects were improved,and the FlexNet model was proposed.The experimental results show that the FlexNet model is accurately classified on the ovarian cancer data set.The degree is increased from 95.21%to 97.13%;Finally,on the MURA and MNIST data sets,the models before and after the improvement are compared experimentally,and the experimental results verify the reliability of the FlexNet model.