A closed-loop artificial pancreas based on model predictive control: Human-friendly identification and automatic meal disturbance rejection

Type 1 diabetes is characterized by a lack of insulin production by the pancreas, causing high blood glucose concentrations and requiring external insulin infusion to regulate blood glucose. Continuous glucose sensors can be coupled with continuous insulin infusion pumps to create a closed-loop artificial pancreas. A novel procedure of “human-friendly” identification testing using multisine inputs is developed to estimate suitable models for use in an artificial pancreas. A constrained model predictive control (MPC) strategy is developed to reduce risks of hypo- and hyperglycemia (low and high blood glucose concentration). Meal detection and meal size estimation algorithms are developed to improve meal glucose disturbance rejection when incoming meals are not announced. Closed-loop performance is evaluated through simulation studies of a type 1 diabetic individual, illustrating the ability of the MPC-based artificial pancreas control strategy to handle announced and unannounced meal disturbances.     

Real-time estimation of plasma insulin concentration from continuous glucose monitor measurements

Continuous glucose monitors can measure interstitial glucose concentration in real time for closed-loop glucose control systems, known as artificial pancreas. These control systems use an insulin feedback to maintain plasma glucose concentration within a narrow and safe range, and thus to avoid health complications. As it is not possible to measure plasma insulin concentration in real time, insulin models have been used in literature to estimate them. Nevertheless, the significant inter- and intra-patient variability of insulin absorption jeopardizes the accuracy of these estimations. In order to reduce these limitations, our objective is to perform a real-time estimation of plasma insulin concentration from continuous glucose monitoring (CGM). Hovorka’s glucose–insulin model has been incorporated in an extended Kalman filter in which different selected time-variant model parameters have been considered as extended states. The observability of the original Hovorka’s model and of several extended models has been evaluated by their Lie derivatives. We have evaluated this methodology with an in-silico study with 100 patients with Type 1 diabetes during 25 h. Furthermore, it has been also validated using clinical data from 12 insulin pump patients with Type 1 diabetes who underwent four mixed meal studies. Real-time insulin estimations have been compared to plasma insulin measurements to assess performance showing the validity of the methodology here used in comparison with that formerly used for insulin models. Hence, real-time estimations for plasma insulin concentration based on subcutaneous glucose monitoring can be beneficial for increasing the efficiency of control algorithms for the artificial pancreas.     

New developments in the treatment and monitoring of type 1 diabetes mellitus

In recent years, insulin analogues are the benefits of the use in functional intensive insulin therapy for the treatment of diabetes. Shortacting insulin (lispro, aspart and glulisine) and long-acting insulin (glargine and detemir) have been developed for the management of diabetes. Short-acting insulin analogues are an alternative to regular human insulin before meals. These new short-acting insulin analogues show more rapid onset of activity and a shorter duration of action. As a result of these pharmacokinetic differences, an improved postprandial glycemic control is achieved, without increasing the risk of hypoglycemia. In addition, these insulin analogues can be administered immediately before a meal. The long-acting insulin analogues provide basal insulin levels for 24 h when administered once (glargine) or two (detemir) daily. Compared with previous intermediate- or long-acting conventional insulin, these insulins shows a flat profile of plasma insulin levels . The use of these long-acting insulin analogues appears to be associated with a reduced incidence of hypoglycemia, especially at night. The availability of these new insulin analogues has the potential to significantly improve long-term control over blood glucose in diabetic patients. In recent years more and more frequently the method of multiple daily injections (MDI) of insulin is being replaced by the method of continuous subcutaneous insulin infusion (CSII). It is the most physiological way to administer insulin. In recent years treatment with insulin pumps has been used more frequently in the pediatric patients and in the treatment of diabetes in pregnancy. Use of continuous glucose monitoring systems enables detection of glycemia fluctuations unrevealed by selfmonitoring of blood glucose, such as night hypoglycemias and early postprandial hyperglycemias. Real-time systems allow to reduce HbA1c levels and limit number of excursions. Non-invasive glucose measurement devices are introduced. Fully automated continuous glucose monitoring systems integrated with insulin pumps operating in closed-loop model, requiring no patient assistance, are still being researched. Commercially available systems operate in open-loop model, where the patient has to decide on administration and dose of insulin.     

Internal model sliding mode control approach for glucose regulation in type 1 diabetes

Patients with type 1 diabetes require insulin therapy to maintain blood glucose levels within safe ranges since their pancreas is unable to complete its function. The development of a closed-loop artificial pancreas capable of maintaining normoglycemia during daily life will dramatically improve the quality of life for insulin-dependent diabetic patients. In this work, a closed-loop control strategy for blood glucose level regulation in type 1 diabetic patients is presented. A robust controller is designed using a combination of internal model and sliding mode control techniques. Also, the controller is provided with a feedforward loop to improve meal compensation. A simulation environment designed for testing the artificial pancreas control algorithms has been used to evaluate the controller. The simulation results show a good controller performance in fasting conditions and meal disturbance rejection, and robustness against model–patient mismatch and meal estimation errors.     

Rapid tightening of blood glucose control leads to transient deterioration of retinopathy in insulin dependent diabetes mellitus: the Oslo study.

In a study of retinopathy during one year of tight blood glucose control 45 type I (insulin dependent) diabetics without proliferative retinopathy were randomised to receive either continuous subcutaneous insulin infusion, multiple insulin injections, or conventional insulin treatment (controls). Near normoglycaemia was achieved with continuous infusion and multiple injections but not with conventional treatment. Blind evaluation of fluorescein angiograms performed three monthly showed progression of retinopathy in the control group, transient deterioration in the continuous infusion group, and no change in the multiple injection group. Half the patients receiving continuous infusion and multiple injections developed retinal cotton wool spots after three to six months. These changes regressed in all but four patients after 12 months. Control patients did not develop cotton wool spots. Patients who developed cotton wool spots are characterised by a larger decrement in glycosylated haemoglobin and blood glucose values, more frequent episodes of hypoglycaemia, a longer duration of diabetes, and more severe retinopathy at onset. A large and rapid fall in blood glucose concentration may promote transient deterioration of diabetic retinopathy.     

持续皮下输注赖脯胰岛素治疗老年非初诊2型糖尿病患者临床观察

目的：观察比较持续皮下输注赖脯胰岛素与常规注射预混赖脯胰岛素对老年非初诊2型糖尿病患者的疗效与安全性。方法：将58例老年2型糖尿病患者随机分为观察组（29例）与对照组（29例）,观察组用赖脯胰岛素经胰岛素泵持续皮下输注（CSI-I）,对照组用精蛋白锌重组赖脯胰岛素25注射液,2次／d,常规皮下注射。两组患者均给予糖尿病教育、饮食控制及适量运动,共治疗2周。比较治疗前后两组患者的血糖、胰岛素用量、血糖达标时间以及低血糖发生率。结果：治疗后两组患者空腹血糖、餐后血糖均较治疗前下降（P〈0．05）,观察组血糖达标时间、胰岛素用量均明显低于对照组（P〈0．05）。两组低血糖发生率无明显差异。结论：持续皮下输注赖脯胰岛素具有较好的疗效与安全性,是控制老年非初诊2型糖尿病患者较佳的方法。     

Flash Glucose Monitoring Compared to Capillary Glucose Levels in Patients With Diabetic Ketoacidosis: Potential Clinical Applications

ObjectiveFrequent, finger-prick capillary blood glucose measurement is standard care, used to drive insulin infusion rates for inpatients being resuscitated from diabetic ketoacidosis (DKA). Over recent years there has been a shift toward continuous interstitial glucose monitoring, allowing monitoring of glucose without repeated invasive testing. While continuous interstitial glucose monitoring has been safely and reliably utilized in the outpatient setting, it has yet to be studied in acutely unwell patients with DKA. The aim of this study, allowing for physiologically lower interstitial compared to capillary glucose, was to determine if interstitial flash glucose monitoring (FGM) would lead to insulin infusion rates that were similar to capillary blood glucose (CapBG) in DKA.MethodsIn this study, 10 patients with diabetes mellitus, assessed to be in DKA, were enrolled. At the same time as standard DKA management commencement, simultaneous FGM measurements were obtained. Duplicate paired glucose readings were then analyzed for agreement.ResultsActual (CapBG-driven) and predicted (FGM determined) insulin infusion rates were similar. Minor differences in predicted insulin infusion rates were noted in 2/10 patients at higher glucose concentrations, which may relate to the lag in change in glucose in the interstitial space.ConclusionBased on our results, a trial of clinical outcomes in patients with DKA treated with insulin infusion rates driven by CapBG versus subcutaneous FGM appears justified. The FGM method of testing may improve patient comfort, obviate fatigue, improve staff time and direct patient contact, and potentially facilitate rapid discharge.     

Mathematical model of glucose–insulin homeostasis in healthy rats

According to the World Health Organization there are over 220 million people in the world with diabetes and 3.4 million people died in 2004 as a consequence of this pathology. Development of an artificial pancreas would allow to restore control of blood glucose by coupling an infusion pump to a continuous glucose sensor in the blood. The design of such a device requires the development and application of mathematical models which represent the gluco-regulatory system. Models developed by other research groups describe very well the gluco-regulatory system but have a large number of mathematical equations and require complex methodologies for the estimation of its parameters. In this work we propose a mathematical model to study the homeostasis of glucose and insulin in healthy rats. The proposed model consists of three differential equations and 8 parameters that describe the variation of: blood glucose concentration, blood insulin concentration and amount of glucose in the intestine. All parameters were obtained by setting functions to the values of glucose and insulin in blood obtained after oral glucose administration. In vivo and in silico validations were performed. Additionally, a qualitative analysis has been done to verify the aforementioned model. We have shown that this model has a single, biologically consistent equilibrium point. This model is a first step in the development of a mathematical model for the type I diabetic rat.     

An iterative learning strategy for the auto-tuning of the feedforward and feedback controller in type-1 diabetes

This paper proposes a scheme for the control of the blood glucose in subjects with type-1 diabetes mellitus based on the subcutaneous (s.c.) glucose measurement and s.c. insulin administration. The tuning of the controller is based on an iterative learning strategy that exploits the repetitiveness of the daily feeding habit of a patient. The control consists of a mixed feedback and feedforward contribution whose parameters are tuned through an iterative learning process that is based on the day-by-day automated analysis of the glucose response to the infusion of exogenous insulin. The scheme does not require any a priori information on the patient insulin/glucose response, on the meal times and on the amount of ingested carbohydrates (CHOs). Thanks to the learning mechanism the scheme is able to improve its performance over time. A specific logic is also introduced for the detection and prevention of possible hypoglycaemia events. The effectiveness of the methodology has been validated using long-term simulation studies applied to a set of nine in silico patients considering realistic uncertainties on the meal times and on the quantities of ingested CHOs.     

Theoretical evaluation of the parameters of glucose metabolism on the basis of continuous glycemia monitoring data using mathematical modeling

The aim of this paper is to construct a mathematical model that takes the main physiological parameters of blood-glucose regulation into account, in order to identify these parameters for an individual patient according to continuous glucose-monitoring data. The constructed mathematical model consists of six ordinary differential equations that describe the dynamics of changes in glucose concentrations, as well as insulin and anti-insulin factors in the blood. Estimation of the parameters of the equations was performed using an evolutionary programming method. The model predictions were fitted to the continuous glucosemonitoring data. As a result of the identification of the model parameters for two patients with type 1 diabetes mellitus, the estimated insulin secretion was close to zero and the estimated glucose utilization and insulin clearance were increased in comparison with the data for healthy donors. Here, we present a personalized model of the regulation of blood glucose, which can be used to predict the results of continuous glucose monitoring depending on modification of the prescribed glucose-lowering therapy. This approach can significantly reduce the number of iterations of the selection of medical hypoglycemic therapy and therefore increase the effectiveness of treatment according to glucose-monitoring data.

     

Development of a closed-loop artificial pancreas.

A closed-loop glucose controlled insulin infusion system was developed, consisting of elements for continuous blood glucose analysis, a computer control system, and infusion systems. Improvements include decreased size, cost reduction and better performance. The algorithm used was a piecewise linear representation of the sigmoidal curve commonly employed. The apparatus has been applied to simulation of the healthy beta cell and glucose clamp studies.     

Subcutaneous continuous insulin infusion and control of blood glucose concentration in diabetics in third trimester of pregnancy.

The effect of subcutaneous continuous insulin infusion on the control of blood glucose concentrations was assessed in eight pregnant diabetics in the third trimester. Twenty-four-hour glucose profiles were obtained after strict inpatient control on conventional insulin regimens and after the start of the continuous infusion, which was maintained for 5-55 days. Mean 24-hour glucose concentrations (6.2 mmol/l on conventional regimen, 5.9 mmol/l on continuous infusion; 111.6 and 106.2 mg/100 ml respectively) and mean fasting concentrations (5.3 v 6.2 mmol/l; 95.4 v 111.6 mg/100 ml) were not significantly changed by continuous infusion. Diurnal variations in glucose concentration tended to be smaller on continuous infusion: standard deviation from mean 24-hour glucose concentration was reduced from 2.5 to 2.0 mmol/l (from 45 to 36 mg/100 ml), maximum excursion from 8.4 to 7.4 mmol/l (151.2 to 133.2 mg/100 ml), and M value from 16 to 14. Subcutaneous continuous insulin infusion may be useful in limiting diurnal variations of blood glucose concentrations and warrants further investigation since such an action may be beneficial in the management of pregnant diabetics, in whom the best possible control of blood glucose concentrations is sought for the good of the fetus.     

Beta cell function, insulin resistance and plasma adiponectin concentrations are predictors for the change of postprandial glucose in non-diabetic subjects at risk for type 2 diabetes.

Insulin resistance, impaired insulin secretion, and low adiponectin levels have been shown to be predictors for type 2 diabetes. However, it is not yet clear whether these associations (1) are independent of changes in body weight, or (2) are valid for changes in glucose tolerance in the prediabetic state. Sixty-two non-diabetics (50 with normal glucose tolerance) aged 41 +/- 11 years, BMI 30.5 +/- 5.3 kg/m2 (mean +/- SD) were studied twice with a standard oral glucose tolerance test (oGTT, mean follow-up time 3.0 +/- 1.8 years (mean +/- SD) [range 0.5-6.5 years]). Insulin sensitivity and insulin secretion were estimated from oGTT using validated indices. Two-hour blood glucose during oGTT deteriorated over time (baseline 2 h glucose 6.32 +/- 0.21 VS. follow-up 2 h glucose 7.14 +/- 0.22 mM, p < 0.001) while the percentage body fat did not change (32.7 +/- 1.2 VS. 32.6 +/- 1.2%, p = 0.46). Follow-up 2 h blood glucose was predicted by adiponectin (p = 0.01), baseline insulin sensitivity (p = 0.02) and baseline insulin secretion relative to insulin sensitivity (p = 0.03) independent of sex, age, baseline 2 h blood glucose or change in percentage body fat. Our results suggest that low adiponectin levels, insulin resistance and low beta cell function predict the continuous deterioration of glucose tolerance in early prediabetic states, independent of changes in adiposity. Therefore, the early influence of these parameters should be the subject of future prevention programs to prevent deterioration of glucose tolerance.     

The use of the artificial pancreas in uremic diabetic patients.

Maintenance hemodialysis and renal transplantation are increasingly used for treating diabetic patients with end-stage renal failure. The use of the artificial pancreas is able to prevent large blood glucose fluctuations in these patients with atherosclerosis, advanced retinopathy or neuropathy in which hyper- and hypoglycemia are potentially deleterious. For this purpose, we have developed and are utilizing an artificial pancreas easily utilizable without special training by the staff of a dialysis unit. This artificial pancreas uses a polarographic glucose electrode with a fast response time (45 to 90 seconds), a terminal display for operator communication, and a continuous digital and analogyl display for control of the running operation. There is also a printer to display in tabular and graphical form the values at any time during the operation. In this preliminary study, 7 patients have been studied: five under repetitive hemodialysis for four hours, 3 times a week; one treated by peritoneal dialysis for 12 hours, twice a week and one controlled during, and 48 hours after, renal transplantation. The macroscopic pancreas normalizes blood glucose under these circumstances, helps in a better understanding of blood glucose homeostasis in uremic patients under dialysis, leads to a more precise evaluation of insulin needs, may help to improve the nutritional status of the patients, and has an educational value for the patient and the medical staff.     

Model predictive control of glucose concentration in type I diabetic patients: An in silico trial

In this paper, the feedback control of glucose concentration in type I diabetic patients using subcutaneous insulin delivery and subcutaneous continuous glucose monitoring is considered. A recently developed in silico model of glucose metabolism is employed to generate virtual patients on which control algorithms can be validated against interindividual variability. An in silico trial consisting of 100 patients is used to assess the performances of a linear output feedback and a nonlinear state-feedback model predictive controller, designed on the basis of the in silico model. More than satisfactory results are obtained in the great majority of virtual patients. The experiments highlight the crucial role of the anticipative feedforward action driven by the meal announcement information. Preliminary results indicate that further improvements may be achieved by means of a nonlinear model predictive control scheme.     

接受肠外营养液的非重症住院患者的血糖调控

接受肠外营养液的非重症住院患者常因手术或其他原因无法进食而采用胰岛素控制血糖,加之一些影响血糖的治疗因素存在,致使该类患者的血糖控制更加复杂。从影响血糖的因素、血糖的监测和调控方式及血糖波动的评价指标等方面论述非重症住院患者血糖的调控,着重介绍胰岛素静脉控制血糖的使用方法和基本算法,以期为临床治疗提供参考     

Glucose Homeostatic Law: Insulin Clearance Predicts the Progression of Glucose Intolerance in Humans

Homeostatic control of blood glucose is regulated by a complex feedback loop between glucose and insulin, of which failure leads to diabetes mellitus. However, physiological and pathological nature of the feedback loop is not fully understood. We made a mathematical model of the feedback loop between glucose and insulin using time course of blood glucose and insulin during consecutive hyperglycemic and hyperinsulinemic-euglycemic clamps in 113 subjects with variety of glucose tolerance including normal glucose tolerance (NGT), impaired glucose tolerance (IGT) and type 2 diabetes mellitus (T2DM). We analyzed the correlation of the parameters in the model with the progression of glucose intolerance and the conserved relationship between parameters. The model parameters of insulin sensitivity and insulin secretion significantly declined from NGT to IGT, and from IGT to T2DM, respectively, consistent with previous clinical observations. Importantly, insulin clearance, an insulin degradation rate, significantly declined from NGT, IGT to T2DM along the progression of glucose intolerance in the mathematical model. Insulin clearance was positively correlated with a product of insulin sensitivity and secretion assessed by the clamp analysis or determined with the mathematical model. Insulin clearance was correlated negatively with postprandial glucose at 2h after oral glucose tolerance test. We also inferred a square-law between the rate constant of insulin clearance and a product of rate constants of insulin sensitivity and secretion in the model, which is also conserved among NGT, IGT and T2DM subjects. Insulin clearance shows a conserved relationship with the capacity of glucose disposal among the NGT, IGT and T2DM subjects. The decrease of insulin clearance predicts the progression of glucose intolerance.     

Hypoglycaemia and counterregulation during childhood

Hypoglycaemia is particularly common in young children with type 1 diabetes mellitus yet the normal protective counterregulatory responses have been little studied in this age group. The studies reported have shown conflicting results, in part related to prior glycaemic control and also to the method of investigation used. Counterregulatory hormone responses during both spontaneous and experimentally induced episodes of nocturnal hypoglycaemia do appear to be blunted, which may be a function of sleep itself. Although studies of cognitive function have consistently shown defects in certain areas of neurocognitive performance, particularly in those children with early-onset diabetes or a prior history of severe hypoglycaemia, the contribution of nocturnal hypoglycaemia to the development of these impairments has not been evaluated. In young adults and adolescents, nocturnal hypoglycaemia has been linked to cardiac arrhythmia and the risk of sudden death. The development of new techniques for continuous subcutaneous glucose monitoring may allow detailed study of counterregulatory responses and symptom recognition in young children. Effective intensification of insulin therapy without an increased risk of hypoglycaemia may be possible using new insulin analogues or continued subcutaneous intravenous infusion (CSII), thus improving patient compliance and overall quality of clinical care.