Background: To improve type 1 diabetes mellitus (T1DM) management, we developed a model predictive control (MPC) algorithm for closed-loop (CL) glucose based on linear second-order deterministic-stochastic model. The deterministic part of the is specified by three patient-specific parameters: Insulin sensitivity factor, insulin action time, and basal infusion rate. stochastic identical all patients but identified from data single patient. Results first clinical feasibility test are...

Bolus calculators help patients with type 1 diabetes to mitigate the effect of meals on their blood glucose by administering a large amount insulin at mealtime. Intraindividual changes in physiology and nonlinearity insulin-glucose dynamics pose challenge accuracy such calculators.We propose method based continuous-discrete unscented Kalman filter continuously track postprandial sensitivity. We augment Medtronic Virtual Patient (MVP) model simulate noise-corrupted data from continuous...

This paper addresses model identification of continuous-discrete nonlinear models for people with type 1 diabetes using sampled data from a continuous glucose monitor (CGM). We compare five techniques: least squares, weighted Huber regression, maximum likelihood extended Kalman filter and unscented filter. perform the on 24-hour simulation stochastic differential equation (SDE) version Medtronic Virtual Patient (MVP) including process output noise. fits actual CGM signal, as well short-...

The risk of hypoglycemia is one the main concerns in treatment type 1 diabetes (T1D). In this paper we present a head-to-head comparison currently used insulin-only controller and prospective bihormonal for blood glucose people with T1D. strategy uses insulin to treat hyperglycemia as well glucagon ensure fast recovery from hypoglycemic episodes. Two separate model predictive controllers (MPC) based on patient-specific models handle infusion. addition, control algorithm consists Kalman...

The primary goal of this paper is to predict fasting glucose levels in type 2 diabetes (T2D) long-acting insulin treatment. presents a model for simulating insulin-glucose dynamics T2D patients. combines physiological 1 (T1D) and an endogenous production T2D. We include review sources variance values treatment, with respect dose guidance algorithms. use the simulate treatment compare results clinical trial where algorithm was used. investigate through simulations evaluate contribution...

Background: Treatment inertia and prescription complexity are among reasons that people with type 2 diabetes (T2D) do not reach glycemic targets. This study investigated feasibility of a new approach to basal insulin initiation, where the dose needed target is estimated from two weeks continuous glucose monitoring (CGM) data. Methods: was an exploratory single arm maximum length 84 days. Eight naïve T2D, planning initiate insulin, wore CGM throughout period. A predetermined regime followed...

This paper presents a novel ensemble nonlinear model predictive control (NMPC) algorithm for glucose regulation in type 1 diabetes. In this approach, we consider number of scenarios describing different uncertainties, instance meals or metabolic variations. We simulate population 9 patients with physiological parameters and time-varying insulin sensitivity using the Medtronic Virtual Patient (MVP) model. augment MVP stochastic diffusion terms, noise-corrupted CGM measurements. meal...

Around a third of type 2 diabetes patients (T2D) are escalated to basal insulin injections. Basal dose is titrated achieve tight glycemic target without undue hypoglycemic risk. In the standard care (SoC), titration based on intermittent fasting blood glucose (FBG) measurements. Lack adherence and day-to-day variabilities in FBG measurements limiting factors existing procedure. We propose an adaptive receding horizon control strategy where glucose-insulin model identified used predict...

In this paper we compare the performance of five different continuous time transfer function models used in closed-loop model predictive control (MPC). These describe glucose-insulin and glucose-glucagon dynamics. They are discretized into a state-space description as prediction MPC algorithm. We simulate scenario including meals daily variations parameters. The numerical results do not show significant changes glucose traces for any models, excepted first order model. From present study,...

This letter presents a model-free insulin titration algorithm for patients with type 2 diabetes that automatically finds and maintains the optimal dosage in order to maintain blood glucose concentration at desired levels. The proposed method is based on recursive least square-based extremum seeking control. Since does not require detailed model, it can be applied wide population of without need identify adapt models patient data. We demonstrate effectiveness using silico simulations, which...

With the fast growth of diabetes prevalence, disease is now considered an epidemic. Diabetes characterized by elevated glucose levels, that may be treated with insulin. Tight control essential for prevention complications and patients' well-being. In this paper we model fasting glucose-insulin dynamics in type 2 diabetes, aiming at controlling level. Relevant clinical data are typically sparse have a sampling period much greater than humans. We adapt physiological such important slow...

In this paper, we discuss the identification of a physiological model describing glucose-insulin dynamics in people with type 1 diabetes (TID). The identified has to be applied nonlinear predictive control (NMPC). We propose stochastic TID. Discrete-time glucose data are provided by continuous monitor (CGM). use maximum likelihood for parameter estimation, combined procedure compute gradient function. To test our procedure, generate virtual population 10 patients using Hovorka and its...

In patients with type 1 diabetes, the effects of meals intake on blood glucose level are usually mitigated by administering a large amount insulin (bolus) at mealtime or even slightly before. This strategy assumes, among other things, prior knowledge meal size and postprandial dynamics. On hand, bolus during after could benefit from information provided dynamics expense delayed bolus. The present paper investigates different administration strategies (at mealtime, 15 minutes 30 beginning...

The purpose of this study is the online detection faults and anomalies a continuous glucose monitor (CGM). We simulated type 1 diabetes patient using Medtronic virtual model. model system stochastic differential equations includes insulin pharmacokinetics, insulin-glucose interaction, carbohydrate absorption. detected two types CGM faults, i.e., spike drift. A fault was defined as value in any zones C, D, E Clarke error grid analysis classification. Spike modelled by binomial distribution,...