Research And Implementation Of Hepatocellular Carcinoma Pathological Image Recognition

HCC is concealed and its progress is rapid,as a result it has a high mortality and poor prognosis.But early diagnosis and treatment can improve the prognosis and survival levels.Pathological images are the gold standard in HCC diagnosis and Computer Aided Diagnosis(CAD)can detect the small changes in HCC images which will help improve the accuracy of pathological diagnosis.Finally,CAD has an important significance and application value in improving the clinical diagnosis of HCC.At first,the research status of HCC in CAD and the medical image recognition technology are analyzed,especially feature selection methods and classification methods.Then different features are extracted on liver pathological images after preprocessing,a new hybrid feature selection method based on Maximum Minimum Backward Selection(MMBS)search strategy is proposed,and the classification model of Voting Optimization Random Forests(VORF)is built.Finally,the proposed classification methods are applied to the development of HCC pathological image recognition system.In order to verify the effectiveness of the algorithms,the thesis has made a serial of experiments.In the validation experiments of feature selection methods,comparative experiments are carried out in different search strategies and different feature selection methods.Experimental results show that MMBS search strategy is excellent and the hybrid feature selection method based MMBS is effective.In the validation experiments of classifier designing methods,this thesis conduct the experiments to optimize the parameters in VORF classifer,also compare the performance of classification algorithms.In addition,comparation exprements of defferent classification methods,which are VORF,KNN,back-propagation neural network(BPNN),SVM,SVM based on Improved Fruit fly Optimization Algrothm(IFOA-SVM)and Random Forests,are done.Experimental results show that the proposed classification method can achieve a higher classification accuracy and sensitivity.What's more,it can reduce the rate of misdiagnosis and ensure the recognition time to meet the needs of physicians.In order to verify the practicality and effectiveness of the system,a lot of tests are conducted after the system development has been completed.The tests include testing the order of all the oprations and their absent situation during HCC pathology image classification.After testing,the system is perfected again.Finally,using a large number of images to test the system,the test results show that:the proposed method has a higher accuracy and recognition speed.