MODULAR BIO-BEHAVIORAL CLOSED-LOOP CONTROL OF TYPE 1 DIABETES

描述：从本质上讲，糖尿病患者控制血糖（BG）的闭环系统（称为人工胰腺）的发展是一个涉及生理学、行为科学和工程学的跨学科项目。因此，该项目代表了来自美国、意大利和法国的医生、心理学家、数学家和工程师的跨学科和国际努力，致力于了解1型糖尿病（T1DM）个性化闭环控制的行为和生物学先决条件。所提出的研究的主要思想是：为了取得成功，闭环控制必须适应个体的生理特征和每个人的行为特征。这种适应的关键是生物系统（患者）观察和模块化控制。因此，我们建议为一个模块化系统奠定基础，该系统由算法观察患者的行为和代谢状态，以及负责胰岛素输送和低血糖预防的控制模块组成。建立这个系统，我们将利用我们在人体代谢的硅建模和模拟方面的广泛专业知识，以及我们最近的闭环控制研究的经验。该系统的开发和测试将分四个阶段完成：第一阶段将调查与T1DM控制相关的行为事件模式。一项实地研究将记录饮食、胰岛素注射和锻炼，同时进行持续的血糖监测，旨在开发一种学习算法——行为观察者——它将随着时间的推移追踪一个人日常生活中关键的反复出现的元素。第二阶段将研究胰岛素敏感性和反调节受损与BG变异性的关系，旨在开发算法生理学观察者，跟踪葡萄糖变异性和复发性低血糖的特定参数，作为一个人胰岛素敏感性和反调节能力变化的标志。第三阶段将在现场测试一个咨询系统，为T1DM患者提供个性化的反馈。该系统将由行为和生理观察员提供信息，并将包括三个咨询模块：(i)提前24小时评估低血糖风险；（ii）丸量计算器建议餐前胰岛素剂量，（iii）基础速率顾问建议未来24小时的基础速率概况。第4阶段将侧重于自动化闭环控制，开展一系列研究（门诊和住院），依次测试三个控制模块，负责：(i)提前1-2小时检测和预防低血糖，（ii）餐前胰岛素丸的控制，（iii）基础速率和夜间稳定状态的控制。我们设想，根据患者或医生的选择，每个观察者或咨询/控制模块可以单独使用，也可以在集成的开环或闭环控制系统中使用。模块化方法将允许增量测试和系统特性的部署，这将构建和促进系统开发。

DESCRIPTION: By nature, the development of a closed-loop system (known as artificial pancreas) controlling blood glucose (BG) in diabetes is an interdisciplinary project involving physiology, behavioral science, and engineering. Consequently, this project represents an interdisciplinary and international effort of physicians, psychologists, mathematicians, and engineers from the United States, Italy, and France dedicated to the understanding of behavioral and biological prerequisites to individually-tailored closed-loop control of type 1 diabetes (T1DM). The principal idea of the proposed research is: in order to be successful, closed-loop control must adapt to individual physiologic characteristics and to the behavioral profile of each person. The keys to this adaptation are biosystem (patient) observation and modular control. Thus, we propose to lay the foundation for a modular system comprised of algorithmic observers of patients' behavior and metabolic state, and control modules responsible for insulin delivery and hypoglycemia prevention. Building this system, we will utilize our extensive expertise with in silico modeling and simulation of the human metabolism, and the experience from our recent closed-loop control studies. The development and testing of this system will be accomplished in four phases: Phase 1 will investigate patterns of behavioral events relevant to T1DM control. A field study will record meals, insulin injections, and exercise in parallel with continuous glucose monitoring, aiming to develop a learning algorithm - behavioral observer - which will track over time key recurrent elements of a person's routine. Phase 2 will investigate relationships of insulin sensitivity and impaired counterregulation with BG variability, aiming to develop algorithmic physiology observers, which will track specific parameters of glucose variability and recurrent hypoglycemia as markers of change in a person's insulin sensitivity and counterregulatory ability. Phase 3 will test in the field an advisory system providing personalized feedback to people with T1DM. The system will be informed by behavioral and physiology observers and will consist of three advisory modules: (i) evaluation of risk for hypoglycemia 24 hours ahead; (ii) bolus calculator suggesting pre-meal insulin doses, and (iii) basal rate advisor suggesting basal rate profiles for the next 24 hours. Phase 4 will focus on automated closed-loop control conducting a series of studies (outpatient & inpatient) testing sequentially three control modules responsible for: (i) detection and prevention of hypoglycemia 1-2 hours ahead, (ii) control of pre-meal insulin boluses, and (iii) control of basal rate and overnight steady state. We envision that, depending on patients' or physicians' choice, each observer or advisory/control module could be used separately, or within integrated open- or closed-loop control systems. A modular approach will permit incremental testing and deployment of system features, which will structure and facilitate system development. PUBLIC HEALTH RELEVANCE: With the advances in insulin delivery and continuous glucose monitoring, research must now focus on integrating these technologies into systems alleviating the burden of everyday diabetes maintenance and ensuring optimal diabetes control - a task that requires studies of physiology, behavior, and engineering. Thus, we propose an interdisciplinary project, which will lay the foundation for a modular diabetes control system using algorithmic observation of patients' behavior and metabolic state to inform control modules responsible for insulin delivery and hypoglycemia prevention. We envision that, depending on patients' or physicians' choices, each module could be used separately, or within integrated advisory or closed-loop control systems. A modular approach will also permit incremental testing and deployment of system features, which will structure and facilitate the progress towards the automated closed-loop control commonly known as artificial pancreas.