Benign And Malignant Classification Method Of Pulmonary Nodules Based On Semi-supervised Learning

The morbidity and mortality of lung cancer rank first,in China and even the world,have been seriously affected the lives and health of the people.The medical field is to diagnose the benign and malignant conditions of pulmonary nodules through CT images of the lungs,as an important evidence for the diagnosis of lung cancer at present.Although the medicine has developed quite well,the current diagnosis of lung cancer in hospitals mainly depends on experienced doctors reading through the naked eye.Traditional diagnosis methods have misdiagnosis and wrong diagnosis.Due to the high production cost of CT image data sets,this paper proposes a method based on semi-supervised learning to classify benign and malignant pulmonary nodules in CT images.This method which combined by an unsupervised learning’s generative adversarial networks and a supervised learning’s convolutional neural network.The research content of this article is divided into two parts: the augmentation of lung CT image data based on the generative adversarial networks and the classification of benign and malignant pulmonary nodules based on the convolutional neural network.On the one hand,it is to study the data augmentation based on the generative adversarial networks model.The traditional data augmentation operation is to use the traditional image processing methods to reverse,rotate,and translate the original image.However,these type of operations only increases the number of images,and the information in the picture has not changed,which cannot achieve the effect of real data amplification.This paper proposes a generative adversarial network based on the fusion of DCGAN and WGAN-GP.During the model training process,a progressive training mode is proposed.The model can generate clear images to solve the data augmentation of lung CT images.On the other hand,it studies the classification model based on convolutional neural network.The structure of the convolutional neural network model is improved by bring the residual blocks,and the network depth is improved to 42 layers.In terms of data sets,combined with the augmented samples by generative adversarial networks for training,and related reference experiments were designed to analyze the classification results and model performance.The accuracy,sensitivity,specificity and AUC value of the experimental results were 96.5%,96.67%,96.33% and 0.953.These results are better than other existing methods in various indicators by comparison.Meanwhile,it demonstrates the feasibility of semi-supervised learning,and provides a new method for small data to participate in deep learning networks.The main innovation of this paper is the introduction of an augmented model based on generative adversarial networks,which is integrated with supervised learning and a semi-supervised learning method,which is applied to the classification of benign and malignant pulmonary nodules,and achieves ideal results.