基于支持向量机的急性肾损伤早期诊断预测模型研究

项目名称： 基于支持向量机的急性肾损伤早期诊断预测模型研究

项目编号： No.61201093

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 崔丽艳

作者单位： 北京大学

项目金额： 27万元

中文摘要： 急性肾损伤（Acute kidney injury，AKI）是临床上一种常见且严重的并发症，研究显示患者的发病率和死亡率都在逐年增加，其主要原因是AKI的诊断主要依靠肌酐和尿量的变化，而且肌酐和尿量只有在肾功能明显受损时才有可能检测出变化，目前尚无合适的分子标志物可以对AKI进行早期预测，如果能找到AKI早期诊断标志物将能够使患者得到及时有效的治疗。尿液中性粒细胞明胶酶相关脂质运载蛋白（NGAL）等分子标志物与肾功能损伤有关。支持向量机比传统的统计学分析在早期诊断预测方面的优势是能够处理复杂、多维以及非线性变量。 本申请拟建立急性肾损伤尿液分子标志物数据库，在此基础上构建包括NGAL等8项尿液标志物影响在内的急性肾损伤的早期诊断支持向量机预测模型，以期能够在肾功能发生变化前为临床的AKI诊断提供依据，为急性肾损伤的危险分层奠定基础。同时本研究也为AKI的定义及诊断标准提供了新途径。

中文关键词： 急性肾损伤；机器学习；早期诊断；；

英文摘要： Acute kidney injury (AKI) is common and critical complications in the hospital settings. The incidence and mortality has been increasing significantly. Criteria for the diagnosis of AKI rely heavily on measurements of serum creatinine and urine output. Critically ill patients may already have significant AKI even if their serum creatinine has not changed yet. Unfortunately no reliable diagnostic markers which can predict the AKI are currently available. Identification of markers that predict which of AKI is present would allow earlier and more appropriate therapy. Eight biomarkers, NGAL, KIM-1, IL-18, L-FABP, et al have been tested to various degrees in patients who already have kidney injury. Support vector machine (SVM) analysis is potentially more successful than the conventional statistical techniques in predicting early diagnosis when the relationship between variables that determine the diagnosis is complex, multidimensional and non-linear . The purpose of the current study was to develop databases of urine biomarker and prediction model of early diagnosis in AKI by using supprot vector machine analysis. The model aids the diagnosis of AKI prior to changes in kidney function and risk stratification. We may be on the verge of a new, more targeted definition and diagnostic criteria of AKI based on urine biom

英文关键词： acute kidney injury；machine learning；early diagnosis；；