A Model Of Hepatitis B Virus Reactivation Based On Data Analysis After Liver Cancer Radiotherapy

The research establishes a model of hepatitis B virus(HBV)based on data analysis after liver cancer radiotherapy.In order to diagnose the illness,we separate hepatocellular carcinoma patient samples into two categories: HBV reactivation and HBV inactivation.The high dimensional vector data groups are composed by clinical factors,radiotherapy schedules and dose volume histograms(DVH).Unfortunately,the traditional medical statistics software cannot efficiently solve this problem due to the redundancy in data groups.For example,it is difficult to find the vital information concerning HBV reactivation and impossible to establish the prediction model correctly just by utilizing SPSS.Thus,we use intelligent computer technologies to extract features from high dimensional vector data groups in order to discover the key characteristic vector set which would be employed for forecasting.In this paper,we propose two methods: the first method uses Logistic Regression to find the optimal feature subset and the key factor,then the Support Vector Machine(SVM)would be used in pattern classification;the second method utilizes Genetic Algorithm to extract features,and then the SVM and Bayes Classifier would be used in classification recognition respectively.The experimental results of original data show that the accuracy rate of SVM classifier reaches 74.44%,and the accuracy rate of Bayes classifier is improved to75.76%.The accuracy rate of SVM classifier reaches 78.89% after feature extraction based on logistic regression analysis.Genetic algorithm is used to extract the optimal feature subset,the accuracy rate of SVM classifier reaches 82.22%,when the size of optimal feature subset is 5 and mutation operation is Gaussian.The accuracy rate of Bayes classifier is improved to 81.11%.Five risk factors are selected by genetic algorithm,including Child-Pugh,outer margin of RT,TNM,HBV baseline and GTV volume.The classification model based on feature extraction of genetic algorithm is suitable for HBV reactivation.