Research On The Construction And Retrieval Of Literature Type Case Database

Case report is a representative form of case type literature.As a common medical document,its main role is to report new cases found in clinical diagnosis and treatment（such as AIDS and New Coronavirus pneumonia are reported for the first time through case reports）,and the special clinical manifestations and clinical diagnosis and treatment experience of the disease.The number of published case reports exceeds 3million,which is a valuable medical knowledge resource accumulated by mankind in clinical diagnosis and treatment.However,because the literature cases represented by case reports are unstructured,and there are differences between different case reports,before doctors need to refer to case reports for auxiliary diagnosis,they often do not know what disease the current patient has,and the existing literature retrieval tools can not meet the needs of doctors to retrieve case reports directly through the patient’s clinical symptoms and auxiliary examination images.In order to overcome or at least significantly reduce the above limitations,this paper constructs a structured literature case database and designs a image and text retrieval system for it.In addition to the simple string matching retrieval function,the system further realizes the complex retrieval function: Retrieve the information of case type literature by inputting medical entity or medical image.Taking the case report as the object,this paper uses ontology construction to study the construction method of document type case database,and uses deep learning to study the retrieval method of document type case medical image.Its basic principles and methods are applicable to other similar types of medical documents.The main research work of this paper is as follows:1.Design a literature case database for case report.Firstly,collect a large number of case reports,mark and extract the key information.Secondly,through the reuse of the existing medical ontology,aiming at the entities in the case report,reconstruct the three ontologies of disease,symptom and body system.Extract the entities contained in the key information in the case report,map with the reconstructed medical ontology after standardization,and complete the instantiation of the ontology.The information and ontology extracted from the case report are stored in a structured form,so as to realize the structured construction method of document type case database.2.A medical image retrieval method based on convolution neural network feature fusion is designed.For the demand of doctors to retrieve similar images by inputting medical images,the retrieval model designed in this paper is divided into two stages:extracting image eigenvalues and calculating image similarity.In this paper,a network model based on CNN feature fusion is proposed for medical image feature extraction.This model can effectively extract the feature representation of traditional Chinese medicine images of literature cases.The Euclidean distance formula is used to calculate the distance between feature vectors.The smaller the Euclidean distance of the two image eigenvalues,it shows that the more similar the two images are,the better the effect of medical image retrieval.Experimental results show that the proposed method has good performance.3.Design and implement an image and text retrieval system for case report.Based on the demand analysis and overall design of the system,combined with the document type case database and the medical image retrieval model,the system realizes the function of "searching text by text and searching map by map",and assists doctors to retrieve the structured information of case reports in various forms.The results show that the system can effectively help doctors make full use of the medical resource of case report.