Research On Disease-symptom Semantic Net Construction And Application

Misdiagnosis is a common phenomenon in clinical diagnosis.The consequences of misdiagnosis are different,some people increase their physical and psychological suffer,and delay the rehabilitation time;some people even lose their lives.Misdiagnosis is one of the main causes of medical malpractice and medical disputes.In clinical diagnosis,due to the limitations of people's understanding and the complexity of the disease,the phenomenons that the doctor's preliminary diagnosis does not accord with the essence of the disease occur from time to time.With the advancement of technology and the development of modern medicine,a variety of modern inspections equipment in clinical practice make the diagnostic methods greatly improved,however,the clinical misdiagnosis rate does not decline.According to rough statistics,the misdiagnosis rate is still 10%-15%.The main reason for misdiagnosis is the confusion of similar symptoms.Symptoms are the primary bases for clinical diagnosis.The knowledge of symptoms and misdiagnosis is massively stored in various books?literatures and open source databases.Therefore,it is of great significance for improving clinical diagnosis rate to integrate relevant knowledge sources and construct a "disease-symptom" knowledge system.The system can prompt the misdiagnosis in the process of disease diagnosis.In recent years,much progress has been made in the field of biomedical knowledge representation:(1)Structured biomedical knowledge representation and discovery.Ontology is an important method of structured knowledge representation and a clear formal specification of shared conceptual models.Its main function is to realize knowledge sharing and knowledge reuse.Some ontologies of the main areas have been established,such as gene ontology,disease ontology,human phenotype ontology.(2)Unstructured biomedical knowledge representation and discovery.In recent years,a large number of biomedical information and knowledge have been published on the Internet,in the forms of semi-structured and unstructured texts,such as academic papers,medical textbooks and case reports.Liu et al.integrated the field of semantic bioinformatics and solved the problem of data resource linking.Mohammed et al.linked the disease ontology with the symptom ontology,by connecting the relationship between the disease and the symptom.Cheng et al.integrated various sources of knowledge about human diseases,by establishing semantic relationships of disease-related databases;Huang et al.designed a web-based algorithm to extract the relationship between diseases and genes from a variety of biomedical corpora.Bai et al.constructed a hybrid biomedical knowledge network,by connecting multiple biomedical ontology and knowledge sources.However,there are still some unresolved problems in the field of the support for misdiagnosis prompts.First,the existing symptom ontology is based on anatomy,there is no semantic relationship between concepts,so the similarities between symptoms are not reflected in the ontology.Second,the relationships between symptoms and diseases are stored in unstructured text and not extracted for structured representation.Moreover,the symptoms and diseases are not simple one-to-one relationships,there are also common relationships and rare relationships.Most importantly,none of the existing medical knowledge representation systems contain the knowledge of the differential diagnosis(misdiagnosis).The knowledge of differential diagnosis is usually stored in medical manuals and other literatures,and is not expressed structurally in computer systems,it limits the direct use of the knowledge of misdiagnosis between diseases.In summary,a Disease-Symptom Semantic Net(DSSN)was constructed in this article.Firstly,the corpora describing the symptom texts were obtained from several knowledge bases of medical domain,symptom words were recognized from medical corpora,and the rich symptom words candidate sets are obtained.Then,the semantic similarities between the symptom words in the candidate set were calculated,synonyms were merged according to the semantic similarities,a symptom ontology based on semantic relations between symptom words was established.And then,through the natural language processing and text mining of knowledge base in multiple medical fields,disease-symptom relationships and misdiagnosis relationships were obtained.Finally,based on the established symptom ontology,these relationships were added to construct a Disease-Symptom Semantic Net(DSSN).In addition,as part of research on DSSN application,a tool for misdiagnosis prompt based on Protégé were developed.By using this tool,users can directly obtain the symptoms of a disease and its misdiagnosis,and can clearly obtain the same symptoms and different symptoms of the disease.So,the tool can prompt misdiagnosis information in the process of clinical medical diagnosis and reduce the probability of misdiagnosis.