Ambulatory Artificial Pancreas: merging physiology, behavior, and control design

描述（由申请人提供）：当代研究越来越关注人工胰腺（AP）的发展-一种被称为闭环控制（CLC）的工程系统。最终目标是研制一种动态AP，它有可能对1型糖尿病患者的健康和生活产生巨大影响。我们的跨学科国际团队一直处于CLC开发的前沿，创建模型、硅测试平台、安全和控制算法，代表了当今AP开发的最先进水平。通过这个项目，我们将对动态CLC的探索提升到一个新的水平，首次提出合并1型糖尿病最优控制的三个关键方面：人类行为、生理学和工程学。我们的主要目标是建立、测试和验证一个新的动态CLC系统，该系统可以根据行为和生理学的实时变化进行通知并自适应。我们的基本假设是：行为事件的速率和随后的代谢反应可以分为分层的时间尺度，这可以转化为模块化的工程层次结构，在每个时间尺度上具有清晰可识别和可处理的控制目标。第一阶段-建立评估算法和控制模块（主要时间尺度为分钟-小时）：我们将首先描述心理-行为标记和急性行为事件（例如饮食，运动）与生理反应强度和CLC实时适应需求的关系。工程工具的设计将负责患者安全和预防低血糖，以及本地和远程AP系统的“健康”。我们将开发一个框架来解决由行为挑战引发的系统转变，并将进行创新的生理实验来评估“黎明”现象、复合碳水化合物餐后的葡萄糖通量和肝胰高血糖素敏感性。阶段2 -判断控制成分的效应大小（主要时间尺度为天-周）：我们将设计一种自适应学习算法，识别患者的生物行为模式，如饮食和运动时间，以及胰岛素敏感性的日变化。协调的临床研究和大规模的计算机实验将评估初始化和实时适应控制成分纳入CLC的效应大小。具体来说，我们将评估使用生理和行为的效果：(i)标记物初始化CLC和（ii）调整胰岛素剂量和基础率。第三阶段-长期动态CLC的系统验证和试验：最终的多中心试验将在患者的自然环境中验证我们的系统，为其最终转化为临床实践做准备。驱动第3阶段的主要假设是：与最先进的传感器增强开环治疗相比，闭环控制将减少低血糖的频率，并将增加在70-180 mg/dl目标范围内度过的时间，而不会影响HbA1c测量的平均血糖控制。

DESCRIPTION (provided by applicant): Contemporary studies focus increasingly on the development of artificial pancreas (AP) - an engineering system known as closed-loop control (CLC). The final goal - an ambulatory AP - has the potential to make a tremendous impact on the health and lives of people with type 1 diabetes. Our interdisciplinary international team has been at the forefront of CLC development, creating models, in silico testing platform, safety and control algorithms that represent the state of the art in AP development today. With this project, we bring the quest for ambulatory CLC to a new level, proposing to merge for the first time three key aspects of the optimal control in type 1 diabetes: human behavior, physiology and engineering Our primary goal is to build, test, and validate a new ambulatory CLC system that is informed by, and is adaptive to, real-time changes in behavior and physiology. Our underlying hypothesis is: the rate of behavioral events and the ensuing metabolic responses can be divided into hierarchical time scales, which can be translated into a modular engineering hierarchy with clearly identifiable and tractable control goals at each time scale. Phase 1 - Building assessment algorithms and control modules (primary time scale minutes-hours): We will first characterize the relationships of psycho-behavioral markers and acute behavioral events (e.g. meals, exercise) with the magnitude of physiological response and the need for real-time adaptation of CLC. Engineering tools will be designed responsible for the patient safety and prevention of hypoglycemia and for the 'health' of the AP system on both local and remote levels. We will develop a framework to address system transitions instigated by behavioral challenges and will conduct innovative physiological experiments to assess "dawn" phenomenon, glucose fluxes following complex carbohydrate meal, and hepatic glucagon sensitivity. Phase 2 - Judging the effect size of control components (primary time scale days-weeks): We will engineer an adaptive learning algorithm that recognizes patients' bio-behavioral patterns, such as meal and exercise timing, and diurnal variation in insulin sensitivity. Coordinated clinical studies and large-scale in silico experiments will estimate the effect size of inclusion into CLC of initialization and real-time adaptation control components. Specifically, we will assess the effect of using physiological and behavioral: (i) markers to initialize CLC and (ii) profiles to adjust insulin boluses and basal rate. Phase 3 - System validation and trial of long-term ambulatory CLC: A final multi-center trial will validate our system in patients' natural environment in preparation for its ultimate translation into clinical practice. The primary hypothesis driving Phase 3 is: compared to state-of-the-art sensor augmented open loop therapy, closed-loop control will reduce the frequency of hypoglycemia and will increase the time spent within the target range of 70-180 mg/dl, without compromising average glycemic control as measured by HbA1c. PUBLIC HEALTH RELEVANCE: The artificial pancreas based on closed-loop control, has the potential to make a tremendous impact on the health and lives of people with type 1 diabetes. The development of this technology has made significant strides over the last five years; however, it is still in infancy, currently being tested in inpatient clinical-research center setting. As the transition is made from the clinic to outpatient trials and then to approved ambulatory devices, additional strategies will need to be developed to optimize control and individualize treatment, requiring creative, medically-inspired engineering design and safety monitoring.