基于无创血压构建MIMIC-III波形数据库匹配子集-《中国医学物理学杂志》

文章信息/Info

Title:: Construction of a MIMIC-III waveform database matched subset based on non-invasive blood pressure

作者:: 陈智恒1; 尹存芳2; 罗俊卿3; 1.南方医科大学公共卫生学院，广东广州 510515； 2.南方医科大学南方医院急诊科，广东广州 510515； 3.南方医科大学生物医学工程学院，广东广州 510515

Author(s):: CHEN Zhiheng1; YIN Cunfang2; LUO Junqing3; 1. School of Public Health, Southern Medical University, Guangzhou 510515, China 2. Department of Emergency, Nanfang Hospital, Southern Medical University, Guangzhou 510515, China 3. School of Biomedical Engineering, Southern Medical University, Guangzhou 510515, China

Keywords:: Keywords: critical care medicine non-invasive blood pressure clinical database waveform database

摘要:: 为解决MIMIC-III临床数据库与波形数据库匹配度低的问题，提出一种基于相似度匹配的算法以构建全新的MIMIC-III波形数据库匹配子集。首先利用BP神经网络，对波形数据库缺失的ICU位置信息进行填补。然后利用匹配算法计算两个数据库无创血压的相似度。最后检验并选取相似度高的匹配数据，得出新的波形数据库匹配子集。本算法共匹配了1 133条波形数据库数据。匹配结果表明，该方法匹配的平均正确率为89.03%，具有较高的准确度和效率。这一方法为研究人员构建MIMIC数据库提供了新的思路，并为基于MIMIC数据库的临床研究提供更多的信息。

Abstract:: Abstract: A similarity matching algorithm is proposed to construct a MIMIC-III waveform database matched subset for solving the problem of low matching degree between MIMIC-III clinical database and waveform database. After filling in the missing ICU location information using BP neural network, the matching algorithm is used to calculate the similarity of non-invasive blood pressure between the two databases, and the matching records with high similarity are examined and selected to construct a new matched subset. A total of 1 133 waveform database records are matched. The matching results show that the proposed method has high accuracy and efficiency, with an average matching accuracy of 89.03%. The method provides new ideas for researchers to construct MIMIC database and provides more information for clinical research based on MIMIC database.