Research On Electronic Medical Record Auxiliary Diagnosis And Treatment Of Cerebrovascular Disease Based On Deep Learning

Cerebrovascular disease(CVD)is a common disease that jeopardizes the health of middle-aged and elderly people.Because of its long treatment cycle,long care cycle and high recurrence rate,it has become the main burden of various medical centers.At the same time,the data accumulated in hospitals,such as electronic medical records and patient profiles with the help of artificial intelligence technologies,have become a new idea of auxiliary diagnosis and treatment of CVD.Based on the above background,this paper starts with the analysis of the current research status of the auxiliary medical treatment of CVD electronic medical records.Based on the deep learning algorithms,this paper finds a breakthrough to make up for the incompleteness of electronic medical record data,the inconsistency of medical center data templates,and the lack of large-scale medical corpus,that is the fusion of named entity recognition and graph embedding.By using this technologies,the information have been extracted from massive electronic medical record data,constructs a similar medical record retrieval model,and based on the results of this model,this paper can expand the application scenarios of the auxiliary medical treatment of CVD electronic medical records.In this paper,the research include the following major contents:(1)Research on named entity recognition with attention:In view of the inconsistency of Chinese electronic medical record templates and the lack of large-scale corpus,the Bi-LSTM-ATT-CRF named entity recognition model with attention is introduced into the similar medical record retrieval research.By extracting key information of electronic medical records with attention scores,this methods can by pass the limitations of electronic medical record templates;and by using semi-supervised learning,this paper can solve the problem of corpus missing;(2)Research on similar medical record retrieval model based on CVD named entity network and graph embedding algorithm node2Vec:In view of the incompleteness of Chinese electronic medical record,this paper uses the graph embedding technology node2Vec to establish the network of CVD medical record data and explores the hidden relationship in the network through random walk technology,and obtains the chapter vector representation of the electronic medical record.Finally,the chapter vector representation of the electronic medical record is obtained,and the electronic medical record similarity is calculated by the cosine similarity;(3)Application scenarios of similar medical record retrieval model:Based on the status of information sharing of CVD collaborative prevention and control cloud platform in Tiantan Hospital,this paper realized the application of the auxiliary medical treatment of CVD electronic medical records based on similar medical record retrieval model under cloud platform.Through the study of the auxiliary medical treatment of CVD electronic medical records,this paper has reached the following conclusions:(1)Through comparison experiments with Conditional Random Fields,Bi-LSTM and Bi-LSTM-CRF,this paper demonstrated the applicability of the Bi-LSTM-ATT-CRF model with attention in the field of medical named entity identification;(2)Through comparison experiments with space vector model and LDA topic model,this paper proved that the effect of medical similar text detection can be effectively improved by the fusion of named entity recognition and graph embedding;(3)Through the results of practical application,this paper proved that the model results presented in this paper can extend the application scenarios of the auxiliary medical treatment of CVD electronic medical records.It can realize CVD auxiliary diagnosis and treatment from the similar medical record retrieval,disease distribution analysis and effect analysis of medical treatment methods,which can help doctors to quickly match the medical records with similar symptoms of CVD,find related cases,and improve doctors' work efficiency.