A Run-to-Run Algorithm for Glucose Regulation

葡萄糖调节的逐次运行算法

• 批准号: 6953163
• 负责人: LOIS JOVANOVIC
• 金额: $ 20.24万
• 依托单位: SANSUM DIABETES RESEARCH INSTITUTE
• 依托单位国家: 美国
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-09-30 至 2007-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6953163
• 关键词: artificial intelligence; bioengineering /biomedical engineering; biomedical automation; blood glucose; clinical research; clinical trials; computer assisted patient care; computer simulation; dietary carbohydrates; drug administration rate /duration; drug delivery systems; glucose metabolism; human subject; insulin dependent diabetes mellitus; insulin sensitivity /resistance; mathematical model; model design /development; patient oriented research

DESCRIPTION (provided by applicant): The long term objective of this work is to develop new algorithmic approaches to optimize the delivery of insulin in an automated fashion to people with type 1 diabetes. Specifically, we aim to develop a strategy, inspired by run-to-run control theory established by the chemical process industries, that "learns" from the previous sequence of glucose responses to insulin dosing (over the course of days), and optimally predicts the appropriate strategy for the forthcoming day. The notion of a "cycle" in engineering will be extended to manage the 24 hour routine of repeated meals, activities, and sleep cycles and the corresponding dosing of insulin. The algorithm will be tested in both simulation and clinical trials for robustness to sensor noise, uncertainty in the patient characterization, variability in the timing of the postprandial glucose peak, and variability in the carbohydrate content in the meals. The Specific Aims of this project are to: i) construct predictive patient sensitivity models for calculation of optimal insulin dosing from elevated (or depressed) glucose levels, ii) develop run-to-run algorithm for insulin bolus dosing to provide corrections in subsequent days based on previous history of glucose levels and insulin dosage, and iii) evaluate the robustness of the algorithm through meal challenges of varying carbohydrate content. The aims will blend prototype algorithms that are drawn from systems engineering with validation in a series of clinical tests. The proposed collaboration between systems engineers and renowned diabetes researchers in an established clinical research setting will allow a novel fusion of methods that can be truly characterized as "innovative". The medical collaborators in the proposal are located at the prestigious Sansum Medical Research Institute, which is located less than 10 miles from the campus of the University of California, Santa Barbara. The exchange of personnel will be facilitated, allowing the student and post-doc supported on this project to work at both the institute and the university over the span of the project

描述（由申请人提供）： 这项工作的长期目标是开发新的算法方法，以自动方式优化向 1 型糖尿病患者输送胰岛素。具体来说，我们的目标是制定一种策略，受到化学加工工业建立的逐次运行控制理论的启发，从先前的葡萄糖对胰岛素剂量的反应序列（几天内）中“学习”，并最佳地预测即将到来的一天的适当策略。工程中“周期”的概念将扩展到管理重复进餐、活动和睡眠周期的 24 小时例行程序以及相应的胰岛素剂量。该算法将在模拟和临床试验中进行测试，以测试传感器噪声的鲁棒性、患者特征的不确定性、餐后血糖峰值时间的变化以及膳食中碳水化合物含量的变化。该项目的具体目标是：i) 构建预测性患者敏感性模型，用于根据升高（或降低）的血糖水平计算最佳胰岛素剂量，ii) 开发胰岛素推注剂量的逐次运行算法，以便在随后的几天内提供校正基于之前的血糖水平和胰岛素剂量历史记录，以及 iii) 通过不同碳水化合物含量的膳食挑战来评估算法的稳健性。目标是将系统工程中得出的原型算法与一系列临床测试中的验证相结合。系统工程师和著名糖尿病研究人员在已建立的临床研究环境中拟议的合作将实现真正具有“创新”特征的新方法融合。该提案中的医疗合作者位于著名的 Sansum 医学研究所，该研究所距离加州大学圣巴巴拉分校校园不到 10 英里。将促进人员交流，允许该项目支持的学生和博士后在项目期间在研究所和大学工作