Research On Computer Aided Diagnosis Of Breast Tumor Ultrasonic Image Based On Convolutional Neural Network

As a mature detection method,ultrasound image has become an important method for clinical disease detection due to its non-invasive,real-time,non-radioactive damage and low price,providing important auxiliary information for clinical disease diagnosis.The application of convolutional neural networks in ultrasound images has brought new changes to the analysis and diagnosis of breast tumor ultrasound images.This article first reviews the research status of ultrasound image-assisted diagnosis,then analyzes the key technologies of breast tumor diagnosis,and then proves through experiments that data enhancement is beneficial to improve the generalization ability and robustness of the model,combined with data enhancement and improvement The Alex Net network model can maximize the accuracy of breast tumor ultrasound image classification.Finally,the classification of breast tumor ultrasound images based on convolutional neural network is summarized,and future research directions are prospected.The main research work of this paper is as follows:(1)Aiming at the problem of insufficient data set of breast tumor ultrasound images,a method of data enhancement is proposed to expand the data set.Data enhancement is achieved through methods such as horizontal mirroring,vertical mirroring,rotation,and gray-scale transformation of the original data set.The experimental results show that data enhancement based on the original data set is beneficial to improve the sample quality of training data,thereby enhancing the generalization ability and robustness of the model.Therefore,in the subsequent model improvement experiments of breast tumor ultrasound images,the data set after data enhancement will be selected for the experiment.(2)Aiming at the problem of low accuracy of convolutional neural network in breast tumor ultrasound image classification,an improved Alex Net model aimed at enhancing feature extraction is proposed.On the basic Alex Net model,maximum pooling and batching are added.One operation.In the case of ensuring that the data set samples remain unchanged,Vgg16 and Le Net5 are selected as the network structure of the comparison experiment.Experimental results show that the combined data enhancement and improved Alex Net network model can maximize the accuracy of breast tumor ultrasound image classification,and its classification accuracy rate is 84.9%.