Research Of Pulmonary Nodule Auxiliary Diagnosis Technology Based On CT Image

Since there are no obvious symptoms in the early stage of lung cancer,when it is found,it is already in the middle and late stage,and the chance of cure is greatly reduced.Even if there is a chance of survival,the late effect is not satisfactory.In addition,the biopsy and puncture technique currently used for lung cancer diagnosis will cause great harm and pain to the patients psychologically and physically.The earlier Lung diseases are diagnosed,the earlier Lung tumors can be found,and timely treatment may improve the curative effect and survival rate.Therefore,it is of great significance to make accurate diagnosis and reasonable treatment plan in interventional therapy with computer diagnostic technology.To discover early-stage lung cancer patients and prompt treatment,the CT detected pulmonary nodules is needed to be distinguished between benign or malignantones.In view of the above problems,based on the analysis of the existing computer assisted diagnosis system technology,the database and shandong qilu hospital image data as the research object,the study of the computer assisted diagnosis system of pulmonary nodules.The main contents of this paper include the following:Firstly,in this paper,a series of image preprocessing algorithms for medical CT images and their advantages and disadvantages are summarized.This paper compares and analyzes the existing image preprocessing technologies,a series of image preprocessing algorithms based on image filtering,edge segmentation and image enhancement,and the advantages and disadvantages of various algorithms.Secondly,analyzes the extraction process of pulmonary nodule CT image texture features,analyzes the texture features of the image,analyzes the principle of texture features extraction,and builds a real-time extraction system of pulmonary nodule texture features based on the analysis of medical image data characteristics and Matlab simulation.Moreover,the extracted texture feature data were preprocessed to eliminate the negative impact of uneven samples on the classification and recognition of benign and malignant pulmonary nodules.The subjects were LIDC data from the American association for cancer research(NCD)and CT imaging data from qilu hospital in shandong province.Thirdly,in this paper,9 kinds of texture features of lung CT images including energy,entropy,correlation,contrast,sumaverage,variance and dissimilarity are extracted based on the gray level co-occurrence matrix.The samples were divided into 2 part,75% in test set,and the 25% in validation set.Using the improved W-Relief F feature weight calculation method,the original high-dimensional data is reduced dimensionally,and corresponding weight values of each texture feature are calculated.The weight value is applied to the improved Weighted k-means algorithm to construct a nodule classification model.The analysis results show that nine kinds of texture features have different effects on the classification of the classifier,and the contribution value is different.We can profit from the diagnosis of early stage carcinoma of the lung to some extent(the recognition rate of benign is 81.18%,malignantones is 91.48%),which utilizes the energy correlation contrast entropy as parameters.