Research On Prostate Disease Diagnosis Technology Based On Bayesian Belief Network

With the continuous progress of society,people’s health awareness is constantly improving.On the one hand,medical data is growing rapidly,showing the characteristics of large number,many types,low value density and complex correlation;On the other hand,there are higher requirements for the quality of disease diagnosis.The traditional method of relying solely on doctors’ subjective diagnosis shows certain limitations.With the help of data mining,big data,cloud computing and other new technologies to assist disease diagnosis,it has become a research trend and hotspot in the medical field and has broad development prospects.After many years of operation,the special disease database system of Qingshu health medical data research institute has accumulated a large number of original case data,which contains valuable diagnosis and treatment information.However,these valuable diagnosis and treatment information are diluted and hidden in the massive medical records in the form of text,which cannot be directly used.In view of this problem,the paper takes the diagnosis of prostate disease as the starting point and relies on the special disease database system platform to study the diagnosis technology of prostate disease based on Bayesian belief network,improve the processing and application ability of the platform for prostate disease related information,and better assist doctors in the diagnosis and treatment of prostate disease.The main research contents and innovations of the thesis are summarized as follows:1.A medical record text mining algorithm based on the improved CRF word segmentation algorithm is proposed.The algorithm improves the traditional CRF word segmentation algorithm by defining illegal paths,so that the accuracy of word segmentation is improved and the key information such as the type,etiology and symptom of prostate disease in the medical record text can be extracted more quickly and accurately.The TF-IDF algorithm is used to obtain the frequency of each node,integrate the results of text mining and the professional knowledge of disease diagnosis,clarify the causal relationship between disease types,causes and symptoms,build a data set and complete data preprocessing.2.A diagnosis algorithm of prostate disease based on Bayesian belief network is constructed.First,according to the causal relationship between the attributes in the data set,the topological structure of Bayesian belief network for prostate disease diagnosis is constructed.Then,the correlation probability of each node is calculated according to the Bayesian formula.Finally,the posterior probability of the disease category is calculated and the disease category label is obtained according to its relative size.The same dataset is input into decision tree,K-nearest neighbor,naive Bayes and Bayesian belief network,and their classification effect is compared.The classification accuracy of Bayesian belief network algorithm is higher than other three classification algorithms..3.Design and implement the whole process diagnosis simulation verification model.The simulation model firstly extracts the attributes of a large number of historical medical records and builds a training set,then builds a Bayesian belief network model and trains to obtain various probabilities of nodes to obtain a diagnosis model.Finally,it is assumed that the current diagnostician is diagnosed by the model,selects the disease inducement of the patient,automatically gets the diagnosis of the disease and gets the symptoms of related diseases,which provides an effective reference for the diagnosis of prostate diseases,The simulation results prove the effectiveness and accuracy of the system.