Design And Implementation Of Chinese Electronic Medical Record Information Extraction System

In tlhe wave of computer technology and information science development,medical information construction has become the trend of medical industry development.Electronic medical record data contains detailed records of patients'medical activities from admission to discharge.It contains abundant medical knowledge.Extraction of electronic medical record information is of great significance to clinical diagnosis and medical research.Therefore,it has become an important task to explore information extraction algorithms with high accuracy for medical text extraction.In addition,with the arrival of the era of big data medical treatment,the number of electronic medical records has increased dramatically in recent years.At present,the level of medical informationization in China is not high enough to meet the needs of large-scale data processing.Massive electronic medical record data has not been effectively managed and utilized.In this context,this paper designs and implements a Chinese electronic medical record information extraction system to support large-scale data processing,which is of great significance to the construction of medical informationization.The main work of this paper includes the following parts:(1)Based on the background and research status of the subject,the demand analysis of the electronic medical record information extraction system is carried out,and the overall framework and specific functions of the system are designed.The system is divided into three modules:storage,interface and service.(2)On the storage side,the storage system of HDFS-Redis+MySQL-HBase"three-level cache"structure is designed and deployed,and the performance of HBase is optimized.It meets the need of mass data storage,achieves the function of high-efficient addition,deletion and modification of large-scale data,and facilitates the use of SQL for data processing on local machines.For professional medical text data,the fields of the database are standardized,and the standardized management of metadata is realized,which lays a foundation for data exchange among different systems.(3)In the interface layer,two kinds of RESTful-based HTTP services are built to realize the requirement of remote real-time invocation of information extraction services,so that large-scale data updates,additions,deletions and queries can be completed in real-time through invoking the interface.The stability and efficiency of the interface are verified by pressure test.(4)On the server side,Hadoop-based information extraction and large data processing are designed and implemented to meet the system's demand for efficient information extraction of massive electronic medical record data.A long-tailed task recognition algorithm is proposed for the"long-tailed problem" which often occurs in Hadoop tasks.The performance and stability of the system are improved effectively by the comparison test.(5)In NLP information extraction module,Medical Improved Bi-LSTM+CRF algorithm model is proposed on the basis of CRF+Bi-LSTM.The cross-experiments show that the model achieves better results in medical text information extraction.In addition,medical dictionaries and annotated corpus have been updated by information extraction.At present,a variety of professional medical dictionaries and Chinese annotated corpus have been constructed,which lays a foundation for information extraction and further knowledge mining.