基于AR模型和支持向量机的急性低血压预测  

Prediction for Acute Hypotensive Episodes Based on AR Model and Support Vector Machine

作　　者：王志刚[1] 赖丽娟[1] 熊冬生[1] 吴效明[1] 

机构地区：[1]华南理工大学生物医学工程系,广州510006

出　　处：《中国生物医学工程学报》2011年第2期250-255,共6页Chinese Journal of Biomedical Engineering

基　　金：广东省科技计划项目(2009B030801004)

摘　　要：ICU中,急性低血压的发生严重威胁着患者的生命安全,临床上对其预测性判断主要依靠医生经验。为实现急性低血压预测,利用PhsioNet的MIMIC II数据库ICU监护中的患者临床记录,对发生与未发生急性低血压两者间的平均动脉压信号进行AR模型的功率谱估计,运用医学信息学理论,选取功率谱幅度的中位数、平均值、最大值、标准偏差和极差用于支持向量机分类预测器的学习和训练,建立分类预测模型。预测模型对测试集进行分类预测,得到预测正确率为87.5%,表明相对于直接提取患者平均动脉压信号的统计特征参数作为预测特征,本方法具有更好的预测效果,有利于实现急性低血压提前预测。The occurrence of acute hypotensive episodes （AHE） in intensive care units （ICU） seriously endanger the lives of patients, and mainly depended on diagnosis in advance by experienced doctor. Based on the records of patients in ICU from the MIMIC II dataset in Physionet,this article aims to predict AHE occurrence and applys the theory of medical informaties to achieve the prediction of occurrence of AHE. The proposed approach obtains the power spectrum estimation according to the mean arterial blood pressure between AHE patients and healthy persons by means of the AR spectral estimation. Then we extracts the median, mean, maximum,standard deviation and range for learning and training based on support vector machine （SVM） and establishes a predicting model with the predictive accuracy of 87.5%. The experiment demonstrates that this approach has better prediction accuracy comparing with the one which uses statistical parameters of the mean arterial blood pressure directly, and is beneficial to early prediction of AHE.

关 键 词：急性低血压 AR模型 功率谱 支持向量机 预测 

分 类 号：R318.04[医药卫生—生物医学工程]