重症监护病人的数据驱动模型预测血糖控制

发布时间：2018-03-22 00:03

本文选题：闭环血糖控制　切入点：强化胰岛素治疗　出处：《北京化工大学》2013年硕士论文　论文类型：学位论文

【摘要】：重症监护患者经常遭受应激性高血糖，，应激性高血糖会导致严重的负面效果甚至死亡。通过降低血糖水平，可以减少负面效果甚至死亡率。此外，闭环血糖控制被认为是处理这类问题最理想的方式。模型预测控制作为一种闭环控制方法，由于其处理约束和时滞的优越性，在血糖控制领域表现出了良好的控制效果。实际上，人体的“胰岛素—血糖动态”带有明显的物理约束条件和时滞，因此，模型预测控制被认为是血糖领域中最有希望的选择之一。然而，对于新来的重症监护病人，由于缺少其个体化模型，传统的模型预测控制遇到了极大的挑战。因此，本文提出了数据驱动模型预测控制，旨在实现ICU强化血糖管理。为了验证数据驱动模型预测控制策略的有效性，将其在两组不同的虚拟病人身上做仿真测试，其中一组基于Cobelli等提出的糖尿病代谢模型，另一组基于Hovorka等提出的ICU血糖管理模型。仿真结果表明数据驱动模型控制策略不仅实现了高效地血糖控制，而且对病人个体化差异以及测量噪声有很好的鲁棒性。同时，本文选取了经典的胰岛素输注协议—Yale协议作为对比方法。本文选取了以下一些评价指标：血糖浓度在三个区间（180mg/dL、70mg/dL以及在70-180mg/dL之间）的百分比、血糖危险指数、控制变化性网格分析、血糖和胰岛素的平均值和方差。每项指标都表明数据驱动模型预测控制方法的控制性能优于Yale协议。
[Abstract]:Intensive care patients often suffer from stress hyperglycemia, hyperglycemia will cause serious adverse effect or even death. By lowering the blood glucose level, can reduce the negative effect and even mortality. In addition, the closed-loop control of blood glucose is considered to deal with this problem. The best way to model predictive control as a closed-loop control method, because of its the superiority of dealing with constraints and delays, showing good control effect in blood glucose control field. In fact, the body's insulin glucose dynamics "with obvious physical constraints and delays, therefore, the model predictive control is considered to be one of the most promising fields in the selection of blood glucose. However, patients in intensive care for the new here, due to the lack of the individual model, traditional model predictive control has encountered great challenges. Therefore, this paper proposes a data driven model predictive control, purpose In the implementation of intensive glucose management. In order to verify the ICU data driven model to predict the effectiveness of the control strategy, the simulation test in two groups of patients with different virtual body, one group based on the metabolic model proposed by Cobelli, another group of ICU blood sugar management model proposed by Hovorka based on the simulation results show that the data. Model driven control strategy not only realizes the efficient control of blood glucose, but also has good robustness for individual patients and the differences of measurement noise. At the same time, this paper selects the classic insulin infusion protocol Yale protocol as the example of law. This paper chooses the following index: blood glucose concentration in the interval (180mg/dL, three 70mg/dL and 70-180mg/dL in between) the percentage of blood glucose variability control risk index, grid analysis, mean and variance of blood glucose and insulin. Each index showed that according to the number of flooding The control performance of the dynamic model predictive control method is superior to the Yale protocol.

【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：R459.7;TP273

【参考文献】