Research On Computer Aided Design Of Pulmonary Nodules By Content-based Medical Image Retrieval

Lung cancer is one of the most morbidity and mortality rates in the world,and the incidence and fatality rate of lung cancer are increasing year by year,which poses a great threat to human health.Lung cancer CAD plays an important role in early detection of lung cancer.In this paper,based on the content of image retrieval related technology to lung cancer CAD,the research and implementation of the medical image retrieval technology for lung cancer CAD.The technology can be targeted for lung cancer is detected in the various types of lesions,retrieve the suspected lesions from confirmed image library and the medical records with the same pathological features of medical images,and to retrieve target according to the results of the retrieved image given the corresponding diagnosis,which helps doctors in medical image evaluation,improve the efficiency of diagnosis,reduce the financial burden on the doctor.This article first analyzes the key technologies of content-based medical image retrieval,and then the main CAD system to the lungs in the feature extraction,feature selection and similarity measure algorithm is studied.In feature extraction algorithm research,first from the perspectives of medical isolation sign of pulmonary nodules,and based on this extract can fully express its medical signs of feature space,mainly extract including gray features,shape features and texture features,etc.In the study of feature selection algorithm,in order to reduce unnecessary computation and improve the performance of the system,the feature space optimization was carried out with probability density selection before identification.In terms of key similarity measure,this paper divided into two layers,the first layer according to the points of lung nodules Lobulation and spicule rating close match in the database and the target image screened in semantic similar part of samples,and then the second layer on the part of the screen using the improved KNN algorithm of sample space,selected as the target image classification from the end of the vision is the most similar sample for reference.Finally by the parallel comparative study of three methods of CBIR in the lung CAD demonstrated that the method in this paper improved the retrieval efficiency,There are significant improvement on the single check rate and overall percentage.provides reference for other researchers,has certain scientific significanceand research value.