中医电子病历入院记录信息自动抽取方法研究

Research on the Automatic Extraction Method of Admission Record Information from Traditional Chinese Medicine of Electronic Medical Records

目的中医电子病历入院记录中包含了丰富的中医诊断信息,多用自然语言形式表达,为提高电子病历的利用以及深度促进知识加工,本文对电子病历入院记录展开信息自动抽取研究。方法本文选择主诉、刻下症等包含症状、舌象、脉象等中医术语的文本,以及现病史、既往病史等叙述性强的病史类文本作为主要研究对象,然后根据文本类型分别进行处理。对于主诉、刻下症等症状文本,直接采用命名实体识别技术,抽取症状实体;对于现病史、既往病史等病史文本,首先进行事件抽取,划分出诊疗事件,然后采用命名实体识别技术,抽取各类实体,最后将各类实体存入数据库。结果利用该方法对某医院骨伤科的电子病历进行了实验,其中对刻下症中症状实体的识别率达83.75%,对现病史中症状、疾病等实体的平均识别率达90.48%。结论本文提出的方法可对电子病历的入院记录信息实现自动抽取并以结构化的方式进行存储,为进一步分析和利用电子病历数据提供了便利,并为中医智能化奠定了坚实的基础。

Objective To improve the use of electronic medical records(EMR)and to facilitate in-depth knowledge processing,this paper conducts a research on the automatic extraction of information from the admission records of EMR,which contain rich information on Chinese medicine diagnosis and are mostly expressed in natural language.Methods In this paper,texts containing traditional Chinese medicine(TCM)terms such as symptoms,tongue signs and pulse signs,such as chief complaints and carving down symptoms,are selected as the main subjects of study,as well as texts with a strong narrative history,such as current medical history and past medical history.The texts are then treated separately according to their type.For symptom texts such as chief complaints and inscribed symptoms,the named entity recognition technique is used directly to extract symptom entities;for medical history texts such as current medical history and past medical history,event extraction is first carried out to classify the treatment events,and then the named entity recognition technique is used to extract various types of entities,and finally the various types of entities are stored in the database.Results The method is used to experiment with EMR from the orthopedic and trauma department of a hospital,in which the recognition rate of symptomatic entities in inscribed symptoms reaches 83.75%and the average recognition rate of entities such as symptoms and diseases in the current medical history reaches 90.48%.Conclusion The method proposed in this paper enables the automatic extraction and storage of admission record information from EMR in a structured manner,facilitating further analysis and utilization of EMR data and laying a solid foundation for the intelligence of TCM.