Research On Recognition Of Medical Time Expressions And Events In Chinese Electronic Medical Records

The rapid development of the Internet has accelerated the pace of hospital information system construction.As a part of the construction of medical information system,the number of electronic medical records is increasing day by day.The electronic medical records contain a large amount of medical information that is closely related to the patient's health status,and the electronic medical records are statement of the medical events of the patient at the medical institution.How to mine some key information contained in the electronic medical record texts through technical means,which has important research value in clinical decision support and auxiliary diagnosis.With the development of information technology,the structure of medical information based on natural language processing has become a research hotspot.Medical time expressions and events are an important basis for understanding the occurrence and development of patients' diseases.It has great significance for doctors and researchers to understand dynamic medical phenomena.Since existing research has mainly focused on English electronic medical records,there are few corpora of Chinese electronic medical records.Therefore,the research conducted the following studies:(1)Designed medical time expressions and events annotation rules in Chinese electronic medical records.Based on the definition and annotation rules of medical time expressions and events given in English electronic medical record in the I2B2 evaluation,the research built the medical time expressions and events knowledge rules after analyzing and researching a large number of Chinese electronic medical records.Formulated Chinese electronic medical records time expressions and medical events annotation rules,constructed 1,500 Chinese electronic medical records time expressions annotation corpora and 2000 Chinese electronic medical records medical events annotation corpora.(2)Research on time expressions recognition of Chinese electronic medical records.The research proposes a method that leveraging CNN and BiLSTM to automatically recognize time expressions in Chinese electronic medical records.comparing with several methods,the method of leveraging CNN and BiLSTM has the highest performance in recognizing Chinese electronic medical records time expressions.The overall recognition F1-score reaches 95.38%,and it also performs well on each subclass of time expressions.(3)Research on medical events recognition of Chinese electronic medical records.The research proposes a method based on convolutional attention mechanism and multi-graph neural network to identify medical events in Chinese electronic medical records.comparing it with several methods,it shows that the proposed medical events recognition method has better performance,the F1-score is 92.21%,and it also has a certain degree of improvement in the recognition of each category of medical events.