Study on the Method of Extracting Diabetes History from Unstructured Chinese Electronic Medical Record

Abstract

In this paper, based on the real electronic medical record data of the hospital, a customized method of rule-based learning and information extraction is designed, and three steps are adopted to realize the extraction of Chinese information: sampling and labeling. The medical history information of 600 electronic medical records (including current medical history, past history, personal history, family history, etc.) were randomly selected, and the information needed to be extracted (taking diabetes history as an example) was marked by the labeling platform developed in this study. According to the annotation results, the extraction template is summarized, and the extraction template can be directly used to extract the regular expression extraction rules, and these rules can be used to extract the actual information. The method of manual verification and automatic verification is used to verify the effectiveness of the method. By using the method of natural language processing and rule-based information extraction, an algorithm for extracting customized information from unstructured Chinese electronic medical record text data is designed and implemented. Aiming at the extraction of diabetes history in the hospital, the field verification of a single department has achieved good results.