Research On Computer Aided Diagnosis And Recognition Of Liver Ultrasound Images

As an important method for the examination of liver diseases,liver ultrasound imaging technology has been widely used in China.However,due to the limitations of clinicians' visual fatigue,negligence and diagnostic level,the diagnosis result of liver diseases is easy to be affected.In order to enhance the objectivity of the diagnosis result of liver disease,the computer aided diagnosis technique of liver disease has been paid attention by many experts and scholars.This technique has very important clinical practical value,on the one hand,it can improve the accuracy of diagnosis results.On the other hand,it can alert the possibility of the existence of liver disease,remind clinicians and patients to make further diagnosis and examination,reduce the rate of missed diagnosis and misdiagnosis.In this thesis,the computer aided diagnosis and recognition method of liver disease has been studied in depth.The research content of this thesis mainly includes the following aspects:(1)The feature extraction algorithm of multi-feature fusion was studied and proposed.Aiming at the problem that any single feature cannot accurately describe the characteristics of liver ultrasound image,multiple features were combined in parallel.The results show that the multi-feature fusion feature extraction algorithm fully takes into account the advantages of single feature,extracts more distinguishing feature vectors,and describes the texture features of liver ultrasound image more accurately.(2)Through the dictionary learning theory,after repeated iterations and updates,the optimal dictionary is obtained,and then the optimal coding sparse matrix is obtained.The experimental results show that the features of sparse expression after dictionary learning are more robust than the features before sparse expression.(3)The PSO-ELM classifier algorithm is studied and proposed.Particle swarm optimization algorithm(PSO)algorithm is introduced on the basis of conventional ELM classifier.PSO algorithm is used to optimize the input weight and hidden layer node threshold of ELM classifier,and pso-elm classifier is constructed.The classification effect of the classifier before and after optimization is analyzed and compared.The experimental results show that the recognition ability of ELM classifier optimized by PS0 algorithm is further improved compared with conventional ELM and SVM classifiers.