基于FAM-CART的ICU患者生死预测研究

发布时间：2018-05-15 01:06

本文选题：重症监护室 + 生死预测　；参考：《北京交通大学》2017年硕士论文

【摘要】：重症监护室(Intensive Care Unit,ICU)是现代医院中对抢救患有危重病情病人的重要单元,ICU患者死亡率则是衡量ICU救治水平和服务质量的一个重要指标。目前临床上已经有多种评分系统用于患者的病情评估和生死预测,但这些评估系统均需要耗费大量的人力和财力。因此在人工智能高速发展的背景下,许多学者尝试使用数据挖掘和机器学习方法研究ICU患者生死预测问题,并取得了一些进展,但是目前仅限于实验室的学术研究,距离临床应用仍有距离,同时使用机器学习方法进行预测使得预测结果的解释性较差,很难被临床医护人员接受。因此本文提出了一种基于FAM-CART模型的ICU患者生死预测研究方法。本文主要介绍了基于FAM-CART模型的ICU患者生死预测方法。在分析了现有ICU患者病情评估和生死预测方法的特点基础上,首先对患者的ICU监护信息进行整理分析,分别采用正常值、均值和二值数据填充方法进行数据预处理,并根据生理指标的临床特性对其进行特征提取,然后采用Fuzzy ARTMAP神经网络进行ICU患者的生死预测,并将基于三种数据预处理方法的预测结果进行对比。最后采用预测结果最优的数据预处理方法,利用FAM-CART模型对ICU患者的生死进行预测,最后将预测结果与临床评分系统和逻辑回归、人工神经网络、支持向量机、Adaboost等算法的预测结果进行比较和分析。本文主要开展了以下研究工作:(1)总结和分析临床ICU患者生死预测方法的现状和不足,从而提出基于FAM-CART模型的ICU患者生死预测的方法;(2)提出了基于混合FAM-CART模型的ICU患者生死预测方法,通过使用数据集训练Fuzzy ARTMAP神经网络,并利用其得到的原型节点的质心和置信因子与CART相结合,从而构建FAM-CART模型用于ICU患者的生死预测研究;(3)通过分析ICU患者数据集的特点和缺失程度,设计三种数据预处理方法,并采用Fuzzy ARTMAP神经网络对数据预处理方法进行验证,确定能获得最好预测结果的数据预处理方法;(4)采用FAM-CART模型实现ICU患者生死预测,并将预测结果与基于Fuzzy ARTMAP神经网络得到的预测结果,以及其它经典的机器学习方法的预测结果进行对比分析,验证本研究方法的预测效果。本文研究旨在根据临床ICU监护数据,设计一种既具有良好的预测性能,又能被临床医护人员理解和接受的ICU患者生死预测方法,研究结果表明论文中提出的方法能取得较好的预测性能,可以为临床应用提供理论参考。
[Abstract]:Intensive Care Unit (ICU) is an important unit in modern hospitals for rescuing critically ill patients. The mortality rate of ICU patients is an important index to measure the level of ICU treatment and the quality of service. At present, there are a variety of clinical scoring systems for patients to assess the disease and life and death prediction, but these assessment systems require a lot of human and financial resources. Therefore, under the background of the rapid development of artificial intelligence, many scholars try to use data mining and machine learning methods to study the problem of life and death prediction of ICU patients, and have made some progress, but only in the laboratory academic research. There is still a distance from clinical application and machine learning method is used to predict the result which is difficult to be accepted by medical staff. Therefore, this paper presents a method for predicting the life and death of ICU patients based on FAM-CART model. This paper mainly introduces the method of predicting the life and death of ICU patients based on FAM-CART model. On the basis of analyzing the characteristics of the existing methods of ICU patients' condition evaluation and life and death prediction, the information of patients' ICU monitoring is analyzed firstly, and the normal value, mean value and binary data filling method are used to preprocess the data, respectively. Then Fuzzy ARTMAP neural network was used to predict the life and death of ICU patients, and the prediction results based on three data preprocessing methods were compared. Finally, the optimal data preprocessing method is used to predict the life and death of ICU patients with FAM-CART model. Finally, the prediction results are combined with clinical scoring system and logic regression, artificial neural network. The prediction results of support vector machine and Adaboost algorithms are compared and analyzed. This article mainly carried out the following research work: 1) summarizing and analyzing the present situation and deficiency of the methods of predicting the life and death of clinical ICU patients. The method of predicting the life and death of ICU patients based on FAM-CART model is put forward. The method of predicting the life and death of ICU patients based on mixed FAM-CART model is presented. The Fuzzy ARTMAP neural network is trained by using data sets. The centroid and confidence factor of the prototype node are combined with CART to construct FAM-CART model to predict the life and death of ICU patients. By analyzing the characteristics and missing degree of ICU patient data set, three data preprocessing methods are designed. The Fuzzy ARTMAP neural network is used to verify the data preprocessing method and the data preprocessing method which can obtain the best prediction results is determined. The FAM-CART model is used to predict the life and death of ICU patients. The prediction results are compared with those based on Fuzzy ARTMAP neural network and other classical machine learning methods to verify the effectiveness of the proposed method. The purpose of this study is to design a method for predicting the life and death of ICU patients, which has good predictive performance and can be understood and accepted by medical staff according to clinical ICU monitoring data. The results show that the proposed method can achieve good predictive performance and provide theoretical reference for clinical application.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R459.7

【参考文献】