Mixed Kernel SVM Algorithm Using In Lung Nodule Recognition

With the concern of computer-aided diagnosis,the feature extraction and classification of medical images become the focus of research in computer aided diagnosis.In order to judge whether the Computed Tomography(CT)image is normal,the feature data is extracted from the image,and then the feature data is analyzed,and the classifier is designed to classify the image directly and abnormally.The classification of the results of good and bad and the choice of feature information and the excellent design of the classifier has a great relationship.In order to improve the accuracy of classification,this paper proposes an improved feature extraction algorithm and classification algorithm based on the characteristics of medical image from the aspects of feature extraction algorithm and design of classification method.The main work of this paper is as follows:Firstly,the background and significance of lung Computer Aided Detection(CAD)are introduced,and the present research situation at home and abroad is described comprehensively.The theory of kernel function and support vector machine theory are systematically expounded.Secondly,the analysis of feature information is the basis of image classification.The selection of medical image feature extraction method is related to the accuracy of image classification.According to the characteristics of Region of Interesting(ROI)in pulmonary nodules and invariant moments,the morphological features,gray features,texture features and invariant moments of lung CT images were extracted.Thirdly,the hybrid nucleus Support Vector Machine(SVM)algorithm was used to identify the pulmonary nodules.By combining the polynomial kernel function with the sigmoid kernel function and combining with the advantages of SVM algorithm,the hybrid kernel SVM model is constructed and applied to the identification of pulmonary nodules.By comparing the identification results of pulmonary nodules by mixed kernel SVM and singlecore SVM,the AUC of mixed kernel SVM is higher than that of single-core SVM.Fourthly,the particle swarm optimization(PSO)algorithm is used to optimize the hybrid kernel SVM and applied to the identification of pulmonary nodules.By comparing the identification results of pulmonary nodules based on PSO-optimized hybrid kernel SVM and grid-searched hybrid kernel SVM,the training time of recognition of pulmonary nodules based on PSO-optimized hybrid kernel SVM is shorter than that based on grid-searched hybrid kernel SVM.