The Artificial Pancreas: Current Situation and Future Directions

Cover
The Artificial Pancreas:

Current Situation and Future
Directions
Copyright
Dedication
Contributors
About the contributors
Foreword
	References
Preface
1 Feedback control algorithms for automated glucose management in T1DM: the state of the art
	1.1 Introduction
	1.2 Proportional-integral-derivative control
		1.2.1 Insulin delivery using PID control
		1.2.2 Glucagon delivery using PID control
	1.3 Logic-based control
	1.4 Model predictive control
		1.4.1 Unconstrained MPC with safety checks
		1.4.2 Multiple model adaptive MPC
		1.4.3 Zone MPC
		1.4.4 Set-point-based enhanced MPC design
		1.4.5 Multiple model probabilistic predictive control
		1.4.6 Adaptive generalized predictive control and MPC design
		1.4.7 Bihormone adaptive generalized predictive control
		1.4.8 Policy-based stochastic MPC
	1.5 Switched linear parameter varying control
	1.6 Personalization and adaptation
		1.6.1 Run-to-run approaches
		1.6.2 Iterative learning control
		1.6.3 Moving average approach
	1.7 Machine-learning-based control
		1.7.1 Reinforcement-learning-based approach
		1.7.2 Gaussian process MPC
		1.7.3 Deep learning-assisted control
	1.8 Summary
	Acknowledgments
	References
2 Getting IoT-ready
	2.1 Introduction
	2.2 IoT-enabled AP ecosystems
		2.2.1 The changing face of AP systems
		2.2.2 AP research platforms: past and present
	2.3 Interfacing with additional signals from the IoT
		2.3.1 Activity sensors
		2.3.2 Location sensors
		2.3.3 Multimedia
		2.3.4 Electronic health records
		2.3.5 Crowdsourcing
		2.3.6 Calendar data
	2.4 Rewiring controllers for an IoT-enabled AP
		2.4.1 PID
		2.4.2 MPC
	2.5 Case study: efficient resource utilization in an MPC-based embedded AP
	2.6 Case study: IoT-enabled autonomous bolus assist via deep learning based zone MPC
	2.7 High-level adaptation from big IoT data
		2.7.1 The interplay of cloud, fog, and edge computing
		2.7.2 An AP that learns from big data
	2.8 Conclusions
	Acknowledgments
	References
3 Multivariable AP with adaptive control
	3.1 Introduction
	3.2 Preliminaries
		3.2.1 Adaptive and personalized PIC estimator
		3.2.2 Recursive subspace-based system identification
	3.3 Adaptive PIC cognizant MPC algorithm
		3.3.1 Integrating insulin compartment models with subspace identification
			3.3.1.1 Set-point modification during exercise and recovery period
			3.3.1.2 Glycemic risk index
			3.3.1.3 Plasma insulin risk index
			3.3.1.4 Feature extraction for manipulating constraints
			3.3.1.5 Plasma insulin concentration bounds
			3.3.1.6 Hypoglycemia detection and carbohydrate suggestion
		3.3.2 Adaptive MPC formulation
	3.4 Results
	3.5 Conclusions
	3.A
	Acknowledgments
	References
4 The ARG algorithm: clinical trials in Argentina
	4.1 Introduction
	4.2 Control-oriented models
	4.3 ARG algorithm
		4.3.1 Switched LQG regulator
		4.3.2 SAFE layer
		4.3.3 Auxiliary modules
			Hypoglycemia-related module (Hypo-RM)
			Hyperglycemia-related module (Hyper-RM)
	4.4 Simulations
	4.5 Clinical trials
		4.5.1 Hardware and software
		4.5.2 Clinical procedures
		4.5.3 Results
	4.6 Conclusions
	Acknowledgments
	References
5 Use of intraperitoneal insulin delivery for artificial pancreas
	5.1 Bedside artificial pancreas: the birth of a concept
	5.2 Prioritization of subcutaneous insulin delivery in the development of a wearable artificial pancreas (AP)
	5.3 Rationale for using intraperitoneal insulin delivery
	5.4 Clinical experience with continuous intraperitoneal insulin infusion
	5.5 Closed-loop experience with IP insulin delivery
		5.5.1 AP with IV glucose sensing and IP insulin delivery
		5.5.2 AP with SC glucose sensing and IP insulin delivery
			5.5.2.1 IP insulin delivery: closed-loop vs. open-loop
			5.5.2.2 Closed-loop control with IP vs. SC insulin delivery
	5.6 Perspectives for IP insulin use in AP
	5.7 Declaration of interests
	References
6 Physiological models for artificial pancreas development
	6.1 Role of physiological models
	6.2 The University of Virginia/Padova T1D simulator
		6.2.1 A serendipitous beginning
		6.2.2 Accelerating AP research: the FDA-accepted T1D simulator
		6.2.3 Further developments of the UVA/Padova T1D simulator
	6.3 The oral glucose minimal model
		6.3.1 The model
		6.3.2 Insulin sensitivity: diurnal pattern
		6.3.3 Insulin sensitivity: simple vs. complex carbohydrates
	6.4 Models of new molecules
		6.4.1 Inhaled insulin
		6.4.2 Subcutaneous UltraFast acting insulin analog
		6.4.3 Modeling of pramlintide: in silico assessment of optimal pramlintide to insulin ratio
	6.5 Modeling subcutaneous glucose sensor delay
	6.6 The UVA/Padova T1D simulator for nonadjunctive use of glucose sensors
	6.7 Adaptive AP algorithms
		6.7.1 Run-to-Run strategy for adaptive MPC tuning
		6.7.2 In silico testing
	6.8 Conclusions
	6.A
		6.A.1 UVA/Padova T1D simulator model equation
			Glucose subsystem
			Insulin subsystem
			Glucose rate of appearance
			Endogenous glucose production
			Glucose utilization
			Renal excretion
			External insulin rate of appearance
			Subcutaneous insulin kinetics
			Intradermal insulin kinetics
			Inhaled insulin kinetics
			Subcutaneous glucose kinetics
			Glucagon kinetics and secretion
			Subcutaneous glucagon kinetics
	Acknowledgments
	References
7 Deployment of modular MPC for type 1 diabetes control: the Italian experience 2008-2016
	7.1 Introduction
	7.2 AP hardware
		7.2.1 APS: the in-patient hardware platform
		7.2.2 DiAs: the out-patient hardware platform
			Config. A: Omnipod and Dexcom SEVEN PLUS
			Config. B: t:slim and G4 Platinum
			Config. C: Accu-Chek Combo and G4 Platinum
	7.3 Telemedicine
	7.4 AP control algorithm
		7.4.1 Safety layer
		7.4.2 Control layer
			7.4.2.1 Hyperglycaemia Mitigation System
			7.4.2.2 Model Predictive Control
				Model
				Cost function
				Constraints
				Meal announcement and feed-forward action
		7.4.3 Adaptation layer
			7.4.3.1 Adaption rules
			7.4.3.2 Real-life algorithm
	7.5 In-patient studies
		7.5.1 Study design
		7.5.2 Data analysis
		7.5.3 Results
		7.5.4 H2MS and mMPC
		7.5.5 Further inpatient studies testing the mMPC
		7.5.6 Concluding remarks
	7.6 Out-patient studies
		7.6.1 Study design
		7.6.2 Data analysis
		7.6.3 Results
			7.6.3.1 H2MS, DiAs configuration A [28]
			7.6.3.2 mMPC, DiAs configuration A [29]
			7.6.3.3 mMPC, DiAs configuration C [21]
		7.6.4 Concluding remarks
	7.7 Real-life testing
		7.7.1 Evening & night use of mMPC
		7.7.2 Day & night use of mMPC
		7.7.3 Adaptive mMPC
	7.8 Concluding remarks
	7.9 Conclusions
	Acknowledgments
	References
8 Integrating the clinical and engineering aspects of closed-loop control: the Virginia experience
	8.1 Introduction
	8.2 Overview of the technology
	8.3 Early outpatient AP studies (2011-16)
	8.4 Pediatric studies
	8.5 Ongoing clinical trials (2016-)
	8.6 Conclusion
	References
9 Strategies to mitigate hypoglycaemia in the artificial pancreas
	9.1 Introduction
	9.2 Safety monitoring systems and hypoglycaemia prediction
	9.3 Control strategies for insulin infusion limitation
	9.4 Feedforward actions in exercise-informed systems
	9.5 Carbohydrate intake suggestions as counterregulatory control action
	9.6 Glucagon as counterregulatory control action
	9.7 Conclusions
	Acknowledgments
	References
10 Multiple-signal artificial pancreas systems
	10.1 Introduction
	10.2 Improving control of physical activity and exercise
		10.2.1 Quantitative models of exercise and T1D
		10.2.2 Physical activity as a cue to modified artificial pancreas operation
		10.2.3 Adapting AP operation by quantifying the effect of physical activity and exercise inputs
	10.3 Improving control of meals and snacks
	10.4 Improving control of blood glucose overnight
		10.4.1 Detecting sleep
		10.4.2 Detected sleep as a cue to modified artificial pancreas operation
	10.5 Enhancing patient safety and security
		10.5.1 Remote patient monitoring for emergency services
		10.5.2 Context awareness: pros and cons
	10.6 Conclusions
	References
11 Artificial pancreas in pediatrics
	11.1 Introduction
	11.2 Effect on children of hypoglycaemia and hyperglycaemia
	11.3 Overview of AP development in pediatrics
	11.4 Threshold suspend and predictive low glucose suspend in pediatrics
	11.5 Managing expectations about AP use among families and diabetes providers
	11.6 Hybrid closed-loop systems in pediatrics
	11.7 Fully closed-loop systems
	11.8 Understanding differences in AP tuning in pediatrics
	11.9 Use of AP systems in young children
	11.10 Skin care issues impacting device use
	11.11 Psychological and social considerations for AP in pediatrics
	11.12 Role of automated decision support in pediatrics and short-term AP use in the developing world
	11.13 Long-term needs and future improvements
	11.14 Conclusions and summary
	References
Index
Back Cover