This brief introduces recursive modeling techniques that take account of variations in blood glucose concentration within and between individuals. It describes their use in developing multivariable models in early-warning systems for hypo- and hyperglycemia; these models are more accurate than those solely reliant on glucose and insulin concentrations because they can accommodate other relevant influences like physical activity, stress and sleep.
Such factors also contribute to the accuracy of the adaptive control systems present in the artificial pancreas which is the focus of the brief, as their presence is indicated before they have an apparent effect on the glucose concentration and so can be more easily compensated. The adaptive controller is based on generalized predictive control techniques and also includes rules for changing controller parameters or structure based on the values of physiological variables. Simulation studies and clinical studies are reported to illustrate the performance of the techniques presented.
|
|
|
1 | (8) |
|
|
|
1 | (1) |
|
|
|
2 | (1) |
|
|
|
3 | (6) |
|
2 Components of an Artificial Pancreas System |
|
|
9 | (14) |
|
|
|
10 | (4) |
|
2.2 Sensors for Physiological (Biometric) Variables |
|
|
14 | (3) |
|
|
|
17 | (3) |
|
|
|
20 | (3) |
|
3 Factors Affecting Blood Glucose Concentration and Challenges to AP Systems |
|
|
23 | (10) |
|
|
|
24 | (1) |
|
3.2 Exercise and Physical Activities |
|
|
25 | (2) |
|
|
|
27 | (1) |
|
|
|
28 | (1) |
|
|
|
29 | (1) |
|
|
|
29 | (1) |
|
|
|
30 | (1) |
|
3.8 Glucose Sensor Signal Accuracy and Delay |
|
|
31 | (2) |
|
4 Modeling Glucose and Insulin Concentration Dynamics |
|
|
33 | (18) |
|
|
|
33 | (4) |
|
4.2 Time Series Models and System Identification |
|
|
37 | (11) |
|
4.2.1 Experiment Planning for Data Collection |
|
|
38 | (1) |
|
4.2.2 Selection of Model Structure |
|
|
38 | (3) |
|
4.2.3 Model Performance Criteria |
|
|
41 | (1) |
|
4.2.4 Parameter Estimation |
|
|
41 | (4) |
|
|
|
45 | (3) |
|
4.3 Recursive Time Series Models |
|
|
48 | (1) |
|
|
|
49 | (2) |
|
|
|
51 | (4) |
|
6 Various Control Philosophies for AP Systems |
|
|
55 | (10) |
|
6.1 Proportional-Integral-Derivative Control |
|
|
55 | (1) |
|
6.2 Model Predictive Control |
|
|
56 | (3) |
|
|
|
59 | (2) |
|
6.4 Knowledge-Based Fuzzy Logic Control |
|
|
61 | (4) |
|
7 Multivariable Control of Glucose Concentration |
|
|
65 | (18) |
|
7.1 Recursive Model of Glucose Concentration Dynamics |
|
|
66 | (4) |
|
7.2 Hypoglycemia Detection and Carbohydrate Suggestion |
|
|
70 | (1) |
|
7.3 Meal Detection and Hyperglycemia Prevention |
|
|
71 | (3) |
|
|
|
74 | (4) |
|
7.5 Acute Psychological Stress |
|
|
78 | (2) |
|
|
|
80 | (1) |
|
7.7 Multivariable Adaptive Control |
|
|
80 | (3) |
|
8 Dual-Hormone (Insulin and Glucagon) AP Systems |
|
|
83 | (6) |
|
9 Fault Detection and Data Reconciliation |
|
|
89 | (8) |
|
9.1 Sensor Error Detection and Data Reconciliation |
|
|
90 | (4) |
|
9.2 Controller Performance Assessment and Retuning |
|
|
94 | (3) |
|
|
|
97 | (4) |
|
|
|
101 | (4) |
| References |
|
105 | |
Ali Cinar is Professor of Chemical Engineering at the Illinois Institute of Technology. His research concentrates on three areas: modeling, simulation and control of biomedical systems, complex adaptive agent-based systems, and supervision of process operations. His research activities focus on the development of theory, methods, and tools to use in these application areas. He is a Fellow of the AIChE and the author of books on batch fermentation and chemical process performance evaluation.
Kamuran Turksoy is a postdoctoral researcher in biomedical engineering at IIT. His research focuses on the development of hypoglycemia early alarm systems, multivariable adaptive control systems and process monitoring, performance assessment, and fault diagnosis techniques for risk mitigation in artificial pancreas systems. He has developed software to implement the methods and algorithms developed, and tested them in simulation and clinical studies.
