Long-Term Prediction Of Benign And Malignant Lung Lesions

Lung Diseases have become a serious threat to human health and life due to the large number of smokers,industrial waste gas pollution,crowded living environment and high living pressure,and Lung Cancer’s death rate has surpassed that of all other cancers,and is on the rise every year.The cure rate of patients with advanced lung cancer is zero,the current level of medical treatment can not reach the level of complete cure,the treatment of patients with advanced lung cancer can achieve a five-year survival rate of only about 16%.But the fiveyear survival rate of patients with early lung cancer can be increased by four to five times after surgery and targeted therapy,so it is important to explore the growth process and development of lung cancer.Lung nodules are pathological lesions in the lungs.When lung nodules are identified as malignant nodules,it is very likely that patients will be diagnosed as lung cancer.Predicting Benign and malignant lung nodules is important for radiologists to assess the staging of cancer and to plan individualized clinical treatment.After consulting literatures and reading related books,this paper focuses on the feature extraction and benign and malignant prediction of pulmonary nodules based on medical sequence images.(1)the method of extraction of the pulmonary nodules of pulmonary nodules based on multi-scale variable convolution neural networkBecause the structure of the lung lesion is complex,the shape changes are often changed,and the characteristics of the characteristics are not complete.Using the traditional convolution neural network to extract the characteristics of the single scale pulmonary nodules,or to increase the noise information in the neighboring environment,the extraction of the information is extracted,and the traditional convolution neural network is used to use the square sampling point,and the sampling point can not change with the shape of the lung nodules,and also causes the loss of some shape information.In this paper,the data from a single scale of pulmonary nodules is improved to three different scale of the lung nodules sequence image,which can not lose the edge information of large nodules,which can be used to keep rich and rich nearby information,but it will bring some noise to the knot,so that the comprehensive information of the sequence image of the lung nodules can be extracted.Secondly,the traditional convolution neural network is improved to be a variable convolution neural network,and the sampling points can be changed with the dynamic of the nodules,which can reduce the extraction of unused information,and refine the characteristics of the information.The network framework USES the progressive integrated framework,which is gradually extracted by the input order to extract the characteristic information of the different scale size nodules,and integrate the dual flow information into the comprehensive depth information of the lung nodules.The results of the experimental results of the nrow data set and the hospital data set show that the frame sensitivity,specificity,accuracy and auc value are the highest and the performance is optimal compared with the first method.(2)research on the malignant prediction of pulmonary nodules in the memory neural network based on the length of the long termIn this paper,a combination of quantitative image group learning characteristics and deep spatial and temporal learning characteristics of pulmonary nodules is presented in this paper.First,the quantitative clinical imaging group was mainly included in four categories,the characteristics of the pulmonary nodules,the size of the size of the lung nodules and the characteristics of the pulmonary nodules.Then,the multi-modal characteristics of the quantitative clinical imaging group and the depth space-time characteristics were combined by two layers of simple neural network.Finally,the characteristics of the pulmonary nodules of the lung nodules,which are characterized by time modulation,were used to predict the malignancy of the lung lesions.According to the experiment of the nrow data set and the hospital data set,this paper suggests that the AUROC of the frame is 0.92.71,which is higher than the current advanced method,which can be predicted effectively,the auxiliary physician is diagnosed and the follow-up treatment arrangement is made.