Medical Diagnostic Information Fusion Method And Empirical Research Based On Evidence Theory

There are a lot of pain points in the traditional medical industry,such as uneven allocation of medical resources,complex and inefficient medical treatment processes,and strained doctor-patient relationship.The advance of big data and "Internet +" provides a new way to solve these problems.In the process of transforming traditional medical care to smart medical care,auxiliary medical diagnosis is an important aspect,which is of great significance for improving the efficiency of medical management.The technology of information fusion has important applications in auxiliary medical diagnosis.One of the effective methods is Dempster-Shafer theory,which has been widely used by virtue of its advantages in information expression and processing.Therefore,this thesis mainly researches the method of medical diagnostic information fusion based on Dempster-Shafer theory,and makes empirical analyses based on real data.In the Internet era,the massive amount of data generated in the medical field has played an important role in improving the efficiency of medical decision-making,providing personalized medical services,and improving the level of medical management.However,there are also problems such as diverse data sources,incomplete structure,and information redundancy.Therefore,the first research content of this research is to preprocess medical information to achieve effective management of medical data.Next,the basic probability assignment generation model of medical information is constructed to convert medical information into evidence representation and realize the expression and acquisition of medical knowledge.The third research content is to propose a fusion algorithm between different dimensions of medical data to achieve effective fusion of medical knowledge.Finally,a multi-classifier fusion method based on Dempster-Shafer theory is proposed to realize the auxiliary decision-making of medical diagnosis.First of all,the medical data is preprocessed to realize the effective management of medical information.This chapter describes the form of medical data studied in this research,and classifies medical data from two aspects:deterministic and uncertain.The formal management of medical data is researched from three perspectives,including outlier detection,default-value processing,and attribute reduction.The outlier detection method based on the law of gravitation,the default-value processing method based on ignore,delete and complement,and the attribute reduction algorithm based on R-type clustering are respectively proposed.Empirical analysis shows that the methods proposed in this chapter play an important role in improving the efficiency of medical data management.Secondly,for the preprocessed medical data,the basic probability assignment generation model is proposed to realize the acquisition and expression of medical diagnosis-related knowledge.The specific method is to select appropriate normalized medical data as the training set,and obtain the basic probability assignment generation model through statistical analysis.Based on the idea of K-nearest neighbor,for the sample to be tested,the basic probability assignment of its K nearest "neighbors" is investigated through the distance measurement,and the basic probability assignment of the sample is obtained accordingly.A numerical example is provided to demonstrate the usage of this method,and an empirical analysis is used to illustrate the importance of this method to the acquisition and expression of medical knowledge,which lays the foundation for the fusion of medical diagnosis information based on Dempster-Shafer theory.Thirdly,the fusion of multi-source medical diagnosis information is researched based on Dempster-Shafer theory.This research defines the concepts of clear evidence and fuzzy evidence from the perspective of reliability and entropy,and gives a method to measure the credibility of evidence for judging the reliability of the medical data.A credibility-based medical diagnosis information fusion rule is proposed,including the weight determination method and the weighted average fusion operation of medical diagnosis information.An empirical study of this method is carried out through numerical examples and experiments.It is found that the medical diagnosis information fusion method based on Dempster-Shafer theory can not only improves the accuracy of medical diagnosis,but also has a certain contribution in terms of interpretability.Finally,in order to achieve the purpose of auxiliary medical diagnosis,the problem of multi-classifier fusion of medical diagnosis information is further researched based on the information fusion rule defined above.The first step is to propose the basic probability assignment generation model for the output of the classifier to realize the knowledge expression of the medical diagnosis result of a single classifier.The second step is to propose a multi-classifier fusion algorithm to integrate the knowledge output by each classifier to improve the diagnosis accuracy.The third step is to build a medical diagnosis decision model,complete the comprehensibility of the classification result,and realize auxiliary decision support for medical diagnosis.The empirical analysis shows that this method has high accuracy and strong interpretability in auxiliary medical diagnosis.This thesis researches the medical diagnosis information fusion method based on Dempster-Shafer theory,and solves the key issues such as the standardized management of medical information,the acquisition and expression of medical diagnosis knowledge,and the fusion of multi-source medical information.It has important practical significance for reducing medical costs,improving the efficiency of medical diagnosis,and improving the level of medical management.