文献类型：会议论文

英文题名：An Effective Standardization Method for the Lab Indicators in Regional Medical Health Platform Using N-grams and Stacking

作者：Zhang, Jiaying[1];Wang, Qi[1];Zhang, Zhixing[1];Zhou, Yangming[1];Ye, Qi[1];Zhang, Huanhuan[1];Qiu, Jiahui[1];He, Ping[2]

机构：[1]East China Univ Sci & Technol, Sch Informat Sci & Engn, Shanghai 200237, Peoples R China;[2]Shanghai Hosp Dev Ctr, Shanghai 200041, Peoples R China

会议论文集：IEEE International Conference on Bioinformatics and Biomedicine (BIBM) - Human Genomics

会议日期：DEC 03-06, 2018

会议地点：Madrid, SPAIN

语种：英文

外文关键词：Regional medical health platform; lab indicator standardization; entity alignment; machine learning

摘要：Since 2008, a regional medical health platform has been built for managing electronic health records of top public hospitals in Shanghai. However, public hospitals often use different names to present a same laboratory examination item (or lab indicator) in this regional platform, which seriously hinders the interconnection and sharing of medical information among hospitals. In this paper, we propose an effective method to standardize the lab indicators using n-gram features and Stacking mechanism. Our proposed method sequentially combines a clustering model and a binary classification model. More specifically, we first cluster the lab indicators based on character uni-gram similarity distances to reduce the alignment scale, and then leverage a binary classification algorithm through Stacking mechanism based on character n-gram similarity features to generate candidate-standard indicator pairs iteratively. Experimental studies on the clinical data collected from eight top public hospitals in Shanghai show that our proposed method achieves a good performance 88.43% in terms of F-1-score in the final binary classification, which is highly competitive performance compared to baseline methods.