Skip to main content

2014 | Buch

Data-driven Modeling for Diabetes

Diagnosis and Treatment

herausgegeben von: Vasilis Marmarelis, Georgios Mitsis

Verlag: Springer Berlin Heidelberg

Buchreihe : Lecture Notes in Bioengineering

insite
SUCHEN

Über dieses Buch

This contributed volume presents computational models of diabetes that quantify the dynamic interrelationships among key physiological variables implicated in the underlying physiology under a variety of metabolic and behavioral conditions. These variables comprise for example blood glucose concentration and various hormones such as insulin, glucagon, epinephrine, norepinephrine as well as cortisol. The presented models provide a powerful diagnostic tool but may also enable treatment via long-term glucose regulation in diabetics through closed-look model-reference control using frequent insulin infusions, which are administered by implanted programmable micro-pumps. This research volume aims at presenting state-of-the-art research on this subject and demonstrating the potential applications of modeling to the diagnosis and treatment of diabetes. The target audience primarily comprises research and experts in the field but the book may also be beneficial for graduate students.

Inhaltsverzeichnis

Frontmatter
Data-Driven and Minimal-Type Compartmental Insulin-Glucose Models: Theory and Applications
Abstract
This chapter initially presents the results of a computational study that compares simulated compartmental and Volterra models of the dynamic effects of insulin on blood glucose concentration in humans. In this context, we employ the general class of Volterra-type models that are estimated from input-output data, and the widely used “minimal model” as well as an augmented form of it, which incorporates the effect of insulin secretion by the pancreas. We demonstrate both the equivalence between the two approaches analytically and the feasibility of obtaining accurate Volterra models from insulin-glucose data generated from the compartmental models. We also present results from applying the proposed approach to quantifying the dynamic interactions between spontaneous insulin and glucose fluctuations in a fasting dog. The results corroborate the proposition that it may be feasible to obtain data-driven models in a more general and realistic operating context, without resorting to the restrictive prior assumptions and simplifications regarding model structure and/or experimental protocols (e.g. glucose tolerance tests) that are necessary for the compartmental models proposed previously. These prior assumptions may lead to results that are improperly constrained or biased by preconceived (and possibly erroneous) notions—a risk that is avoided when we let the data guide the inductive selection of the appropriate model.
Georgios D. Mitsis, Vasilis Z. Marmarelis
Ensemble Glucose Prediction in Insulin-Dependent Diabetes
Abstract
Real-time prediction of glucose in type 1 Diabetes Mellitus has received a considerable amount of scientific and commercial interest over the last decade. Numerous different models have been suggested using both physiological and data-driven approaches. Insulin-dependent diabetic glucose dynamics are known to be subject to time-shifting dynamics. Considering this, as well as the vast number of models developed in the literature, it is unclear if a single model can be determined to be optimal under every possible situation. This raises the question whether it is more useful to use one of the models solely, or if it is possible to gain additional prediction accuracy by combining their outcomes. Here, a novel merging approach—combining elements from both switching and averaging techniques, forming a ‘soft’ switcher in a Bayesian framework—is presented for the glucose prediction application. The method is demonstrated on both simulated and empirical data sets.
Fredrik Ståhl, Rolf Johansson, Eric Renard
Hypoglycemia Prevention Using Low Glucose Suspend Systems
Abstract
The fear of nocturnal hypoglycemia was one of the main motivations behind the development of continuous glucose monitors (CGM) and hypoglycemic alarms. Since many individuals do not awake to these alarms, the next step is to implement a low glucose suspend (LGS) algorithm that shuts off the insulin infusion pump to avoid hypoglycemia. A threshold-based LGS simply shuts off the pump when the CGM signal has violated a lower limit on glucose. A predictive LGS shuts off the pump when a glucose threshold is predicted to occur within a specified time limit, or horizon. A review of algorithms hypoglycemic predictors/alarms is followed by a presentation of threshold and predictive LGS algorithms. Finally, an overview of the implementation challenges that remain is provided.
B. Wayne Bequette
Adaptive Algorithms for Personalized Diabetes Treatment
Abstract
Dynamic systems, especially in real-life applications, are often determined by inter-/intra-variability, uncertainties and time-varying components. Physiological systems are probably the most representative example in which population variability, vital signal measurement noise and uncertain dynamics render their explicit representation and optimization a rather difficult task. Systems characterized by such challenges often require the use of adaptive algorithmic solutions able to perform an iterative structural and/or parametrical update process towards optimized behavior. Adaptive optimization presents the advantages of (i) individualization through learning of basic system characteristics, (ii) ability to follow time-varying dynamics and (iii) low computational cost. In this chapter, the use of online adaptive algorithms is investigated in two basic research areas related to diabetes management: (i) real-time glucose regulation and (ii) real-time prediction of hypo-/hyperglycemia. The applicability of these methods is illustrated through the design and development of an adaptive glucose control algorithm based on reinforcement learning and optimal control and an adaptive, personalized early-warning system for the recognition and alarm generation against hypo- and hyperglycemic events.
Elena Daskalaki, Peter Diem, Stavroula Mougiakakou
Pitfalls in Model Identification: Examples from Glucose-Insulin Modelling
Abstract
Two important statistical parameter estimation pitfalls, examples of which can be found in the literature, are here reviewed and discussed. The first concerns the lack of model qualitative behaviour analysis before proceeding to the actual parameter estimation phase: this may give rise in the worst cases to aberrant model behaviour and to meaningless parameter estimates. The second concerns the use of interpolated noisy observations taken to represent the real input or driving variable into a model: this gives rise to the artifactual reproduction of meaningful features of the output variables, based on data errors and hence inherently non-reproducible. This is particularly dangerous when using noisy observations instead of model predictions in coupled systems. Examples of these pitfalls are drawn from existing glucose-insulin modelling literature and recommendations are made.
Simona Panunzi, Andrea DeGaetano
Simulation Models for In-Silico Evaluation of Closed-Loop Insulin Delivery Systems in Type 1 Diabetes
Abstract
This chapter presents simulation models created to support the development of closed-loop insulin delivery systems in type 1 diabetes. Such models, also known as ‘virtual patient’ models, represent an input-output relationship between insulin delivery and other external inputs such as meals or exercise, and the resulting glucose response. It is argued that these simulation models are an essential prerequisite for an accelerated development of the artificial pancreas systems in various populations of type 1 diabetes ranging from children to adults and pregnancies. The present review provides a general introduction to the models of glucose regulation in type 1 diabetes and then proceeds to discussing the individual submodels of glucose kinetics and insulin action, subcutaneous insulin kinetics, subcutaneous glucose kinetics, glucose absorption from the gut, and the exercise effect on the glucose kinetics. Finally, several important virtual-patient models used for in silico testing of glucose controllers are reviewed.
Malgorzata E. Wilinska, Roman Hovorka
Simple Parameters Describing Gut Absorption and Lipid Dynamics in Relation to Glucose Metabolism During a Routine Oral Glucose Test
Abstract
The oral glucose tolerance test (OGTT) is a simple and physiological test used for the diagnosis of diabetes and, more generally, for the assessment of the metabolic condition of an individual. For a deep analysis of the OGTT data, the exploitation of mathematical models for the interpretation of the test results is necessary. The focus of this chapter is on some recent models based on OGTT data, describing the glucose absorption from the gut, and the effects of insulin on the dynamics of non-esterified free fatty acids (NEFA). As regards the glucose absorption model, it requires measure of plasma glucose and insulin during the OGTT. The model allows the estimation of absorption rates, the total glucose absorbed, and half-life in the gastrointestinal tract. In fact, the increase of postprandial circulating plasma glucose over time is the result of gain from gut glucose absorption and liver endogenous glucose production, and loss because of glucose uptake, predominantly by skeletal muscle. Endogenous glucose production can be assessed by some mathematical expressions depending on plasma insulin values, whereas the glucose loss is calculated as a function of the endogenous glucose production and of the insulin sensitivity in each individual. Once these variables have been determined, glucose absorption can be determined by solving a nonlinear least squares problem fitting the plasma glucose values during the OGTT. The model was used to analyze sex related differences in OGTT glucose metabolism, including gut absorption, in healthy humans. We found that, in the early phase of the OGTT, males had markedly increased glucose absorption rates by approximately 200 mg/min from the gastrointestinal tract, whereas in the final phase of the OGTT, females absorbed approximately 60 mg/min more glucose. In another study, the model was used to assess glucose absorption in 15 pregnant women with gestational diabetes. As regards the model of NEFA dynamics, it was postulated that NEFA kinetics could be described by first order (single compartment) kinetics, with NEFA production controlled by insulin in remote compartment. It is assumed that plasma insulin enters a remote compartment, before having an inhibitory effect on NEFA production. Starting from basal production at basal insulin, NEFA production decreases, for suprabasal increase in remote insulin up to a prescribed value. The identification of the unknown parameters of the model was performed by applying Genetic Algorithms. The model was used to study NEFA kinetics in a group of women with a history of gestational diabetes (fGDM). The fGDM women were divided into normal glucose tolerance group (NGT) and a impaired glucose metabolism (IGM). We also studied 15 control (CNT) women. We found that, while fasting NEFA were not different between groups, IGM exhibited slower decline in plasma NEFA during the OGTT. We conclude that appropriate mathematical modeling allows sophisticated analyses of the OGTT data, thus possibly providing a comprehensive picture of the metabolic condition.
Andrea Tura, Giovanni Pacini
Data-Driven Modeling of Diabetes Progression
Abstract
A realistic representation of the long-term physiologic adaptation to developing insulin resistance would facilitate the effective design of clinical trials evaluating diabetes prevention or disease modification therapies. In the present work, a realistic, robust description of the evolution of the compensation of the glucose-insulin system in healthy and diabetic individuals, with particular attention to the physiological compensation to worsening insulin resistance is formulated, its physiological assumptions are presented, and its performance over the span of a lifetime is simulated. Model-based simulations of the long-term evolution of the disease and of its response to therapeutic interventions are consistent with the transient benefits observed with conventional therapies, and with promising effects of radical improvement of insulin sensitivity (as by metabolic surgery) or of β-cell protection. The mechanistic Diabetes Progression Model provides a credible tool by which long-term implications of anti-diabetic interventions can be evaluated.
Andrea DeGaetano, Simona Panunzi, Pasquale Palumbo, Claudio Gaz, Thomas Hardy
Linear Modeling and Prediction in Diabetes Physiology
Abstract
Diabetes Mellitus is a chronic disease characterized by the inability of the organism to autonomously regulate the blood glucose level due to insulin deficiency or resistance, leading to serious health damages. The therapy is essentially based on insulin injections and depends strongly on patient daily decisions, being mainly based upon empirical experience and rules of thumb. This chapter presents work on data-driven glucose metabolism modeling and short-term, that is, up to 120 min, blood-glucose prediction in Type 1 Diabetes Mellitus (T1DM) subjects. Low-order, individualized, data-driven, stable, physiologically relevant models were identified from a population of 9 datasets from T1DM patients. Model structures include: autoregressive moving average with exogenous inputs (ARMAX) models and state-space models. The performances of the model-based predictors were compared to those achieved by the zero-order hold (ZOH).
Marzia Cescon, Rolf Johansson
Nonlinear Modeling of the Dynamic Effects of Free Fatty Acids on Insulin Sensitivity
Abstract
This chapter presents a nonlinear model of the combined dynamic effects of spontaneous variations of plasma insulin and free fatty acids on glucose concentration in a fasting dog. The model is based on the general nonparametric modeling methodology that employs the concept of Principal Dynamic Modes (PDMs) to obtain a Volterra-equivalent nonlinear dynamic model with two inputs (insulin and free fatty acids) and one output (glucose) that are measured experimentally every 3 min in a fasting dog as time-series data over 10 hr. This model is deemed valid and predictive for all input waveforms within the experimental dynamic range. The obtained model is composed of two PDMs for each input and cubic Associated Nonlinear Functions (ANFs), in addition to two cross-terms. The waveform of the obtained PDMs offers potentially valuable interpretation of the implicated physiological mechanisms. The system nonlinearities are described, in turn, by the obtained ANFs. The evaluation of the overall model performance is facilitated by the use of specialized inputs, such as pulses or impulses. For instance, the use of insulin input pulses can yield estimates of “dynamic insulin sensitivity” (as the ratio of the steady-state glucose response to the input pulse amplitude) for various levels of free fatty acids. The obtained result indicates (in a quantitative manner) the widely held view that insulin sensitivity decreases with rising levels of free fatty acids. Furthermore, it indicates that this effect depends on the input insulin strength (dose-dependent insulin sensitivity). Drastic reduction of insulin sensitivity is predicted by the model above a critical level of free fatty acids for low-to-moderate values of plasma insulin. This result demonstrates the potential utility of the proposed modeling approach for advancing our quantitative understanding of the processes underpinning obesity and Type II diabetes.
Vasilis Z. Marmarelis, Dae C. Shin, Georgios D. Mitsis
Metadaten
Titel
Data-driven Modeling for Diabetes
herausgegeben von
Vasilis Marmarelis
Georgios Mitsis
Copyright-Jahr
2014
Verlag
Springer Berlin Heidelberg
Electronic ISBN
978-3-642-54464-4
Print ISBN
978-3-642-54463-7
DOI
https://doi.org/10.1007/978-3-642-54464-4

Neuer Inhalt