Research On Cardiovascular Disease Assistant Decision Model Based On The Fusion Of Knowledge Driven And Data Driven

With the rapid development of artificial intelligence,knowledge graph has become a popular research technology of artificial intelligence.Intelligent medicine,which is the intersection of knowledge graph and medical health,has always been a hot research topic.Cardiovascular disease is one of the diseases with the highest mortality rate in the world today,Cardiovascular disease deaths account for the first cause of death among urban and rural residents,45.91% in rural areas and 43.56% in urban areas.Cardiovascular disease-related data is highly specialized and has a complex structure.An effective cardiovascular disease knowledge graph is constructed to manage these.Data and assisting doctors in initial diagnosis are problems that need to be solved urgently.The construction of the cardiovascular disease knowledge graph is divided into two modes: knowledge-driven and data-driven.The knowledge graph constructed using knowledge-driven lacks real data support,medical concepts/relationships are complex,and simulation is difficult,so it is difficult to form a scale.The problem of using datadriven knowledge graph is also very obvious.Medical data sources are complex,various medical data standards are inconsistent,and medical data is heterogeneous and diversified;most of the data are general data sets open on the Internet,and the reliability is not high.Structured and unstructured data are mixed and do not combine the doctor’s clinical experience and medical standards.Based on the data-driven and data-driven fusion of the cardiovascular disease domain knowledge graph,combined with PDNet which is proposed in this article.this study constructed a cardiovascular disease auxiliary decision model(CDMIKG-PDNet).The main work is as follows:(1)First,use the open source knowledge graph Wikidata,open KG,etc.,the medical language specification defined by the open source knowledge base UMLS,and the prior medical knowledge of doctors in partner hospitals to construct a medical concept graph for cardiovascular diseases,which meets the needs of experts and comprehensively covers cardiovascular diseases.Disease medical knowledge,the concept graph has a cardiovascular disease concept node.The experimental data comes from the real structured data and unstructured data(electronic medical records)in the cardiovascular disease database of the cooperative hospital(Anzhen Hospital).For the structured data,the RDF/OWL mapping rules are used to import the above-mentioned cardiovascular disease medical concepts graph;the method of manually labeling unstructured data(doctors in cooperative hospitals)is proposed,and a Bi LSTM-CRFKB model is proposed.This model combines cardiovascular disease knowledge related to the cardiovascular disease concept knowledge base.In this paper,the cooperative hospital On the data set,the accuracy of the extracted data reached 90.35%.The above two kinds of data are imported into the concept graph,thereby constructing the cardiovascular disease medical instance graph,and the instance graph is visualized using Neo4 j.(2)On the basis of the established cardiovascular disease medical instance graph,tools and doctors are provided to interactively select features,and according to the clinical rules formulated by the doctors,the feature combination nodes that also guarantee the prediction accuracy and the realistic clinical performance are selected and these The real cardiovascular disease data contained in the node is used to train the cardiovascular disease medical instance knowledge graph model(CDMIKG-PDNet).Comparing the data selected by the constructed cardiovascular disease medical instance knowledge graph model(CDMIKG-PDNet)with PDNet(data without graph screening),PNN,Deep FM,Din and DBN,the experimental results prove that,This method has high accuracy in predicting cardiovascular disease on the data set of cooperative hospitals and other algorithms can be used to assist decision-making in cardiovascular disease.