项目名称: 基于生存树的急性心肌梗死早期预警及其多生理参数建模
项目编号: No.61502472
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 计算机科学学科
项目作者: 苗芬
作者单位: 中国科学院深圳先进技术研究院
项目金额: 21万元
中文摘要: 近年来,我国急性心肌梗死死亡率呈快速上升趋势。最新研究表明,心肌梗死可防可控。本研究拟针对目前基于统计学生存模型的急性心肌梗死预测性能变异程度高的弱点,提出一种基于生存树的非线性非参数生存模型,以正确识别急性心肌梗死多模异构危险因素,准确构建其与疾病发作的内在关系,并提高预测模型的最优化程度。首先,研究生存模型中基于全局生存距离最大化及信息熵的生存树分裂准则及其终止条件,研究生存树集成学习中的参数最优化方法;其次,基于所提出的生存模型,建立基于海量临床检验数据的急性心肌梗死综合预测模型,研究预测模型建立过程中的多模异构数据融合方法;更进一步,针对预测模型的决策优化过程,研究面向院外监测的急性心肌梗死多生理参数实时预警模型及其最优化控制方法。本课题的研究将为前瞻性急性心肌梗死的预测提供新方法。
中文关键词: 预测方法;面向特定领域的预测技术;急性心肌梗死;生存模型
英文摘要: The mortality of acute myocardial infarction (AMI) is growing sharply in recent years. It has been demonstrated that AMI can be predicted and controlled before its occurrence. Existing AMI risk models based on statistical survival method are often with high variance. Our study will dedicate to develop an optimal and comprehensive risk model for predicting AMI using a non-linear non-parameter survival model, which is proposed aiming at identifying more important predictors and their inherent effects on the events. Firstly, we will focus on an effective split rule and stopping criterion based on distance maximization and information entropy in survival tree, and then investigate the parameter optimization method in ensemble learning for survival model. Secondly, a comprehensive clinical risk model for AMI will be developed based on the proposed survival model. We will also investigate an efficient integration method for heterogeneous clinical data. At last, we will explore a multi-parameter based real-time AMI risk model with the sequential parameters and its optimization method. Our research will provide a new kind of method for predicting AMI.
英文关键词: Prognosis method;Predictive technology for specific domain;Acute myocardial infarction ;Survival model