Classification And Prediction Of Hepatitis B Virus(HBV)Reactivation After Precise Radiotherapy For Primary Liver Cancer

Reactivation of hepatitis B virus in patients with primary liver cancer after precise radiotherapy is a common complication.At present,most of the researches on the risk factors of HBV reactivation after precise radiotherapy in PLC patients at home and abroad are only conducted through medical statistics.This paper mainly proposes a series of computer intelligence algorithms for such problems.These algorithms fully consider the risk factors of hepatitis B virus reactivation,including radiotherapy clinical factors,metrology parameters and other mainstream medical factors.The established HBV reactivation classification prediction model can help doctors to take targeted individualized treatment measures for liver cancer patients after precise radiotherapy,It is important to prevent HBV reactivation by taking anti-virus and anti-tumor methods,reduce the probability of onset,improve the treatment effect of patients,and prolong the patients' life cycle.This article mainly provides two solutions to solve the problem of hepatitis B virus reactivation after liver cancer radiotherapy.Scheme 1: The feature selection method is used to analyze the risk factors of HBV reactivation,and then the intelligent classification prediction model of HBV reactivation is established.The NCA is used to select the original data set of PLC,find the feature subset of different risk factors combination,and establish the support vector machine and the linear discriminant classification model to carry out HBV reactivation classification prediction before and after Bayesian global optimization.Scheme 2: The original feature is reduced by using the feature extraction method based on sparse autoencoder.Finally,the SVM and Softmax classification model are established for HBV reactivation classification prediction.Both ideas have effectively solved the problem of predicting the reactivation of hepatitis B virus after precise radiotherapy of primary liver cancer.The risk factors found after NCA feature selection mainly include KPS score,tumor stage TNM,Child-Pugh,HBV DNA level,segmentation mode,external release boundary,V25.The V25 is the first risk factor proposed on the basis of previous studies.When KPS score,tumor stage TNM,HBV DNA level,segmentation mode,and external release boundary feature subsets were selected for classification prediction,the accuracy of SVM classification model after Bayesian global optimization adjustment was as high as 87.33%,which compared with the original SVM,the prediction accuracy is improved by 5.11 percentage points in the 10-fold cross validation.The method of feature extraction by sparse autoencoder can effectively reduce the data dimension and improve the prediction accuracy.Among them,the SVM classification model has the best classification performance for the two-layer sparse autoencoder,and the average accuracy rate is 78.52% under the 10-fold cross validation.