Search

This paper deals with subcutaneous blood glucose level control. Inverse optimal trajectory tracking for discrete time non-linear positive systems is applied. The scheme is developed for MIMO (multi-input, multi-output) affine systems. The control law calculates the subcutaneous insulin delivery rate in order to prevent hyperglycemia and hypoglycemia events. A neural model is obtained from an on-line neural identifier, which uses a recurrent neural network, trained with the extended Kalman filter (EKF); this neural model has an affine form, which permits the applicability of inverse optimal control scheme. The proposed algorithm is tuned to follow a desired trajectory; this trajectory reproduces the glucose absorption of a healthy person. Then this model is used to synthesize an inverse optimal controller in order to regulate the subcutaneous blood glucose level for a Type 1 Diabetes Mellitus patient the applicability of the proposed scheme is illustrated via simulation using a recurrent neural network in order to model the insulin-glucose dynamics. [ABSTRACT FROM AUTHOR]

SUMMARY Patients with type 1 diabetes mellitus require exogenous insulin infusion to avoid chronic complications related to elevated glucose levels. With diabetes reaching epidemic proportions, recent times have witnessed an increased attention in the field of optimal glucose management by closed-loop insulin delivery system. A proper glucose management scheme, in addition of maintaining the glucose level within the normal range of 80-120 mg/dL, should avoid excessive insulin delivery leading to hypoglycemia. By considering the glucose regulation as a linear quadratic problem, a constrained novel sub-optimal controller is proposed in the present work, for the maintenance of normal glucose level in type 1 diabetic subjects. The observer free state feedback controller is based on the feedback of only physiological variables (plasma glucose and plasma insulin). Constraining the feedback elements corresponding to the non-physiological variables avoids the use of an observer while maintaining the advantages of state feedback control. The implementation of the proposed scheme requires simple measurement protocol with no online computation. The closed-loop performance of the controller is evaluated on a physiologically relevant model for a meal disturbance and continuous glucose infusion. Copyright © 2013 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

OBJECTIVE--To compare two validated closed-loop (CL) algorithms versus patient self- control with CSII in terms of glycemic control. RESEARCH DESIGN AND METHODS--This study was a multicenter, randomized, d three-way crossover, open-label trial in 48 patients with type 1 diabetes mellitus for at least 6 months, treated with continuous subcutaneous insulin infusion. Blood glucose was controlled for 23 h by the algorithm of the Universities of Pavia and Padova with a Safety Supervision Module developed at the Universities of Virginia and California at Santa Barbara (international artificial pancreas [iAP]), by the algorithm of University of Cambridge (CAM), or by patients themselves in open loop (OL) during three hospital admissions including meals and exercise. The main analysis was on an intention-to-treat basis. Main outcome measures included time spent in target (glucose levels between 3.9 and 8.0 mmol/L or between 3.9 and 10.0 mmol/L after meals). RESULTS--Time spent in the target range was similar in CL and OL: 62.6% for OL, 59.2% for iAP, and 58.3% for CAM.While mean glucose level was significantly lower in OL (7.19, 8.15, and 8.26 mmol/L, respectively) (overall P = 0.001), percentage of time spent in hypoglycemia (<3.9mmol/L) was almost threefold reduced during CL (6.4%, 2.1%, and 2.0%) (overall P = 0.001) with less time ≤2.8 mmol/L (overall P = 0.038). There were no significant differences in outcomes between algorithms. CONCLUSIONS--Both CAM and iAP algorithms provide safe glycemic control. [ABSTRACT FROM AUTHOR]

Blood glucose control, for example, in diabetes mellitus or severe illness, requires strict adherence to a protocol of food, insulin administration and exercise personalized to each patient. An artificial pancreas for automated treatment could boost quality of glucose control and patients' independence. The components required for an artificial pancreas are: i) continuous glucose monitoring (CGM), ii) smart controllers and iii) insulin pumps delivering the optimal amount of insulin. In recent years, medical devices for CGM and insulin administration have undergone rapid progression and are now commercially available. Yet, clinically available devices still require regular patients' or caregivers' attention as they operate in open-loop control with frequent user intervention. Dosage-calculating algorithms are currently being studied in intensive care patients , for short overnight control to supplement conventional insulin delivery , and for short periods where patients rest and follow a prescribed food regime . Fully automated algorithms that can respond to the varying activity levels seen in outpatients, with unpredictable and unreported food intake, and which provide the necessary personalized control for individuals is currently beyond the state-of-the-art. Here, we review and discuss reinforcement learning algorithms, controlling insulin in a closed-loop to provide individual insulin dosing regimens that are reactive to the immediate needs of the patient. [ABSTRACT FROM AUTHOR]

Inverse optimal trajectory tracking via a control Lyapunov function (CLF) for discrete time non-linear systems is developed and applied to type 1 diabetes mellitus patients control. The control law calculates the insulin delivery rate in order to prevent hyperglycemia and hypoglycemia levels. To synthesize the inverse optimal control law a quadratic candidate CLF is used. The proposed algorithm is tuned to follow a desired trajectory; this trajectory reproduces the glucose absorption of a healthy person. Simulation results applied for two different patients illustrate the applicability of the control law and a comparison with inverse optimal neural control via passivity is included. [ABSTRACT FROM AUTHOR]

Aims/hypothesis: Successful postprandial glycaemia management requires understanding of absorption patterns after meals containing variable complex carbohydrates. We studied eight young participants with type 1 diabetes to investigate a large low-glycaemic-load (LG) meal and another eight participants to investigate a high-glycaemic-load (HG) meal matched for carbohydrates (121 g). Methods: On Visit 1, participants consumed an evening meal. On follow-up Visit 2, a variable-target glucose clamp was performed to reproduce glucose and insulin levels from Visit 1. Adopting stable-label tracer dilution methodology, we measured endogenous glucose production on Visit 2 and subtracted it from total glucose appearance measured on Visit 1 to obtain meal-attributable glucose appearance. Results: After the LG meal, 25%, 50% and 75% of cumulative glucose appearance was at 88 ± 21, 175 ± 39 and 270 ± 54 min (mean ± SD), whereas glucose from the HG meal appeared significantly faster at 56 ± 12, 100 ± 25 and 153 ± 39 min ( p < 0.001 to 0.003), and resulted in a 50% higher peak appearance ( p < 0.001). Higher apparent bioavailability by 15% ( p = 0.037) was observed after the LG meal. We documented a 20 min deceleration of dietary mixed carbohydrates compared with dietary glucose for the HG meal and a twofold deceleration for the LG meal. Conclusions/interpretation: Absorption patterns may be influenced by glycaemic load and/or meal composition, affecting optimum prandial insulin dosing in type 1 diabetes. [ABSTRACT FROM AUTHOR]

A lifelong-implanted and completely automated artificial or bioartificial pancreas (BAP) is the holy grail for type 1 diabetes treatment, and could be a definitive solution even for other severe pathologies, such as pancreatitis and pancreas cancer. Technology has made several important steps forward in the last years, providing new hope for the realization of such devices, whose feasibility is strictly connected to advances in glucose sensor technology, subcutaneous and intraperitoneal insulin pump development, the design of closed-loop control algorithms for mechatronic pancreases, as well as cell and tissue engineering and cell encapsulation for biohybrid pancreases. Furthermore, smart integration of the mentioned components and biocompatibility issues must be addressed, bearing in mind that, for mechatronic pancreases, it is most important to consider how to recharge implanted batteries and refill implanted insulin reservoirs without requiring periodic surgical interventions. This review describes recent advancements in technologies and concepts related to artificial and bioartificial pancreases, and assesses how far we are from a lifelong-implanted and self-working pancreas substitute that can fully restore the quality of life of a diabetic (or other type of) patient. [ABSTRACT FROM AUTHOR]

The author comments on an article by Breton et al. on artificial pancreas in patients with type 1 diabetes. According to the author, ideal control by an artificial pancreas would therefore encompass autonomous control with normalization both of glycemia and of the glucose fluxes that contribute to it under all physiological circumstances. The author says that it would reduce the complications that arise form their dysregulation.

SUMMARY An optimal state feedback controller is designed with the objective of minimizing the elevated glucose levels caused by meal intake in Type 1 diabetic subjects, by the minimal infusion of insulin. The states for the controller based on linear quadratic regulator theory are estimated from noisy data using Kalman filter. The controller designed for a physiological relevant mathematical model is coupled with another model for simulating meal dynamics, which converts meal intake into glucose appearance rate in the plasma. The tuning parameters (weighting matrices) of the controller and the design parameters (noise covariance matrices) of the Kalman filter are optimized using genetic algorithm. The controller based on the combined framework of evolutionary computing and state estimated linear quadratic regulator is found to maintain normoglycemia for meal intakes of varying carbohydrate content. The proposed approach addresses noisy output measurement, modeling error and delay in sensor measurement. Copyright © 2012 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

SUMMARY In this paper, a discrete-time inverse optimal trajectory tracking for a class of nonlinear positive systems is proposed. The scheme is developed for MIMO affine discrete-time positive nonlinear systems. This optimal controller is based on discrete time passivity and positive systems theory. The advantage of this scheme is that it avoids solving the associated Hamilton-Jacobi-Bellman equation and minimizes a meaningful cost function. The affine discrete-time positive nonlinear system is obtained from an online neural identifier, which uses a recurrent neural network, trained with the extended Kalman filter. The applicability of the proposed approach is illustrated via simulation by trajectory tracking control of type 1 diabetes mellitus patients. Copyright © 2012 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

The article presents study that proposed a novel blend of model predictive control (MPC) and iterative learning control (ILC), also called learning-type MPC (L-MPC), for closed-loop control in an artificial pancreatic beta-cell. The proposed approach is initially tested on the Adult Average subject. The study showed that the blood glucose concentrations, after 20 days, can be maintained within 68-145 milligrams per decilitre (mg/dl) when the means are repetitive. L-MPC can generate superior control performance compared with that realized under MPC.

OBJECTIVE -- Current state-of-the-art artificial pancreas systems are either based on traditional linear control theory or rely on mathematical models of glucose-insulin dynamics. Blood glucose control using these methods is limited due to the complexity of the biological system. The aim of this study was to describe the principles and clinical performance of the novel MD-Logic Artificial Pancreas (MDLAP) System. RESEARCH DESIGN AND METHODS -- The MDLAP applies fuzzy logic theory to imitate lines of reasoning of diabetes caregivers. It uses a combination of control-to-range and control-to-target strategies to automatically regulate individual glucose levels. Feasibility clinical studies were conducted in seven adults with type 1 diabetes (aged 19-30 years, mean diabetes duration 10 ± 4 years, mean A1C 6.6 ± 0.7%). All underwent 14 full, closed-loop control sessions of 8 h (fasting and meal challenge conditions) and 24 h. RESULTS -- The mean peak postprandial (overall sessions) glucose level was 224 ± 22 mg/dl. Postprandial glucose levels returned to <180 mg/dl within 2.6 ± 0.6 h and remained stable in the normal range for at least 1 h. During 24-h closed-loop control, 73% of the sensor values ranged between 70 and 180 mg/dl, 27% were >180 mg/dl, and none were <70 mg/dl. There were no events of symptomatic hypoglycemia during any of the trials. CONCLUSIONS -- The MDLAP system is a promising tool for individualized glucose control in patients with type 1 diabetes. It is designed to minimize high glucose peaks while preventing hypoglycemia. Further studies are planned in the broad population under daily-life conditions. [ABSTRACT FROM AUTHOR]

Diabet. Med. 27, 480–484 (2010) Aims We assessed an extended interruption of subcutaneous insulin delivery during overnight closed-loop glucose control in children and adolescents with Type 1 diabetes (T1D). Methods In seven young subjects with T1D [age 14.2 ± 2.1 years, diabetes duration 6.9 ± 4.0 years, glycated haemoglobin (HbA1c) 8.0 ± 1.5%, body mass index (BMI) 21.4 ± 4.0 kg/m2, total daily insulin dose 0.9 ± 0.2 units/kg/day; mean ± sd) participating in overnight closed-loop glucose control studies, insulin delivery was interrupted for at least 90 min on the basis of predicted hypoglycaemia, low prevailing glucose levels or a too-steep decline in glucose levels. Results Insulin delivery was interrupted for 165 (105, 210) min [median, interquartile range (IQR)]. Plasma glucose was 6.2 ± 3.2 mmol/l at the time of interruption and 5.5 ± 2.0 mmol/l 105 min later ( P = 0.15, paired t-test). Plasma glucose declined during the first hour of the interruption at a rate of 0.02 ± 0.03 mmol/l per min and reached a nadir of 5.2 ± 2.7 mmol/l; 105 min after the interruption, plasma glucose was increasing at a rate of 0.01 ± 0.03 mmol/l per min. When insulin delivery restarted, plasma glucose was 6.4 ± 2.2 mmol/l and peaked at 7.9 ± 2.1 mmol/l in 60 min ( P = 0.01). Physiological levels of plasma insulin were measured throughout with a nadir of 119 ± 78 pmol/l. Conclusions A prolonged interruption of insulin delivery during overnight closed-loop glucose control to prevent hypoglycaemia was not associated with an increased risk of hyperglycaemia in young people with T1D. [ABSTRACT FROM AUTHOR]

Intensive insulin therapy aimed at achieving normal glucose levels significantly reduces the complications that are associated with diabetes but is also associated with an increased risk of low glucose levels (hypoglycemia). The growing use of continuous glucose monitors has stimulated the development of the artificial pancreas, a closed-loop insulin-delivery system aimed at restoring near-normal glucose levels while reducing the risk of hypoglycemia. The artificial pancreas comprises three components: a continuous glucose monitor, an insulin infusion pump and a control algorithm delivering insulin according to real-time glucose readings. In this article, we review closed-loop glucose control, including its components, development, testing and clinical application. INSET: Key issues. [ABSTRACT FROM AUTHOR]

The article presents a study examining the influence of input excitation upon the quality of empirical dynamic models used for the prediction of blood glucose trends in type 1 diabetes patients. The ability for high-quality models to improve glycemic regulation for individuals with type 1 diabetes is noted. The fact that bolused insulin and meal carbohydrates are the most influential inputs influencing the concentration of glucose in the blood is mentioned. Also noted is the possibility that inputs with high levels of excitation may cause extreme hypoglycemia or hyperglycemia for the patient.

For people with type 1 diabetes, automatic controllers aim to maintain the blood glucose concentration within the desired range of 60–120 mg/dL by infusing the appropriate amount of insulin in the presence of meal and exercise disturbances. Blood glucose concentration outside the desired range can be harmful to an individual’s health but concentration below 60 mg/dL, a state known as hypoglycemia, is considered to be more harmful than the concentration above 120 mg/dL, a state known as hyperglycemia. In this paper, two techniques to address this issue within a multi-parametric model based control framework are presented. The first technique introduces asymmetry into the objective function to penalize the deviation towards hypoglycemia more than the deviation towards hyperglycemia. The second technique is based upon placing higher priority on satisfaction of constraints on hypoglycemia than on satisfaction of constraints on hyperglycemia. The performance of both the control techniques is analyzed and compared in the presence of disturbances. [ABSTRACT FROM AUTHOR]

Nocturnal hypoglycemia (NH) is one of the most challenging events for multiple dose insulin therapy (MDI) in people with type 1 diabetes (T1D). The goal of this study is to design a method to reduce the incidence of NH in people with T1D under MDI therapy, providing a decision-support system and improving confidence toward self-management of the disease considering the dataset used by Bertachi et al. Different machine learning (ML) algorithms, data sources, optimization metrics and mitigation measures to predict and avoid NH events have been studied. In addition, we have designed population and personalized models and studied the generalizability of the models and the influence of physical activity (PA) on them. Obtaining 30 g of rescue carbohydrates (CHO) is the optimal value for preventing NH, so it can be asserted that this is the value with which the time under 70 mg/dL decreases the most, with almost a 35% reduction, while increasing the time in the target range by 1.3%. This study supports the feasibility of using ML techniques to address the prediction of NH in patients with T1D under MDI therapy, using continuous glucose monitoring (CGM) and a PA tracker. The results obtained prove that BG predictions can not only be critical in achieving safer diabetes management, but also assist physicians and patients to make better and safer decisions regarding insulin therapy and their day-to-day lives. [ABSTRACT FROM AUTHOR]

OBJECTIVE -- To assess whether prandial insulin, in addition to basal insulin, has an effect on the rate of glucose appearance from a meal in people with type 1 diabetes. RESEARCH DESIGN AND METHODS -- The rate of glucose appearance from a mixed meal (Ra[sub meal]) was investigated in six adult (aged 24 ± 2 years), lean (BMI 23.6 ± 1.5 kg/m²) subjects with well-controlled type 1 diabetes (duration 7.9 ± 6.9 years, A1C 7.6 ± 0.9%) with/without prandial insulin. Actrapid was infused to maintain euglycemia before meals were consumed. Subjects consumed two identical meals on separate occasions, and Ra[sub meal] was measured using a dual isotope method. [6,6-²H[sub 2]] glucose was incorporated into the meal (0.081 g/kg body wt), and a primed constant/variable rate infusion of [1,2,3,4,5,6,6-²H[sub 2]]glucose was administered. In the tests with prandial insulin, an additional bolus dose of Actrapid was given 20 min before the meal at 0.1 units/kg body wt. RESULTS -- Insulin concentration with prandial insulin was significantly higher than during basal insulin studies (119 ± 16 vs. 66 ± 15 pmol/l, P = 0.03 by paired t test). Despite differences in insulin concentration, there were no differences in total glucose appearance (3,398 ± 197 vs. 3,307 ± 343 µmol/kg) or time taken for 25% (33.1 ± 3.3 vs. 31.7 ± 3.5 min), 50% (54.6 ± 3.5 vs. 54.1 ± 4.7 min), and 75% (82.9 ± 7.1 vs. 82.8 ± 5.8 min) of total glucose appearance. The fraction of the glucose dose appearing in the circulation was the same for basal (73 ± 8%) and prandial (75 ± 4%) study days. CONCLUSIONS -- These results suggest that meal glucose appearance is independent of prandial insulin concentration in people with type 1 diabetes. [ABSTRACT FROM AUTHOR]

Tight glycaemic control (TGC) in critically ill patients improves clinical outcome, but is difficult to establish The primary objective of the present study was to compare glucose control in medical ICU patients applying a computer-based enhanced model predictive control algorithm (eMPC) extended to include time-variant sampling against an implemented glucose management protocol. Open randomised controlled trial. Nine-bed medical intensive care unit (ICU) in a tertiary teaching hospital. Fifty mechanically ventilated medical ICU patients. Patients were included for a study period of up to 72 h. Patients were randomised to the control group ( n = 25), treated by an implemented insulin algorithm, or to the eMPC group ( n = 25), using the laptop-based algorithm. Target range for blood glucose (BG) was 4.4–6.1 mM. Efficacy was assessed by mean BG, hyperglycaemic index (HGI) and BG sampling interval. Safety was assessed by the number of hypoglycaemic-episodes < 2.2 mM. Each participating nurse filled-in a questionnaire regarding the usability of the algorithm. BG and HGI were significantly lower in the eMPC group [BG 5.9 mM (5.5–6.3), median (IQR); HGI 0.4 mM (0.2–0.9)] than in control patients [BG 7.4 mM (6.9–8.6), p < 0.001; HGI 1.6 mM (1.1–2.4), p < 0.001]. One hypoglycaemic episode was detected in the eMPC group; no such episodes in the control group. Sampling interval was significantly shorter in the eMPC group [eMPC 117 min (± 34), mean (± SD), vs 174 min (± 27); p < 0.001]. Thirty-four nurses filled-in the questionnaire. Thirty answered the question of whether the algorithm could be applied in daily routine in the affirmative. The eMPC algorithm was effective in maintaining tight glycaemic control in severely ill medical ICU patients. [ABSTRACT FROM AUTHOR]

Background: Analyzing the dynamics of insulin concentration in the blood is necessary for a comprehensive understanding of the effects of insulin in vivo. Insulin removal from the blood has been addressed in many studies. The results are highly variable with respect to insulin clearance and the relative contributions of hepatic and renal insulin degradation. Results: We present a dynamic mathematical model of insulin concentration in the blood and of insulin receptor activation in hepatocytes. The model describes renal and hepatic insulin degradation, pancreatic insulin secretion and nonspecific insulin binding in the liver. Hepatic insulin receptor activation by insulin binding, receptor internalization and autophosphorylation is explicitly included in the model. We present a detailed mathematical analysis of insulin degradation and insulin clearance. Stationary model analysis shows that degradation rates, relative contributions of the different tissues to total insulin degradation and insulin clearance highly depend on the insulin concentration. Conclusion: This study provides a detailed dynamic model of insulin concentration in the blood and of insulin receptor activation in hepatocytes. Experimental data sets from literature are used for the model validation. We show that essential dynamic and stationary characteristics of insulin degradation are nonlinear and depend on the actual insulin concentration. [ABSTRACT FROM AUTHOR]

A new calculation method is proposed to quantify the endogenous glucose production (EGP), the glucose appearance rate due to meal ingestion (Ra meal), and the glucose disposal (Rd) during a three-tracer study design. The method utilizes the maximum likelihood theory combined with a regularization method to achieve a theoretically coherent computational framework. The method uses the two-compartment formulation of the glucose kinetics. Instead of assuming smoothness of unlabeled and labeled glucose concentrations, the method assumes that the EGP, the Ra meal, and the fractional glucose clearance are smooth, increasing plausibility of their individual estimates. The method avoids transformation of the measurement errors, which may skew the estimates of the EGP, Ra meal, and Rd with the traditional approach. Finally, the sequential nature of the calculations is replaced by calculating the EGP, Ra meal, and Rd in "one go" to avoid the propagation of the errors from the EGP and Ra meal into Rd. An example study is shown demonstrating the utility of the approach. A better performance of the new method is demonstrated in a simulation study. [ABSTRACT FROM AUTHOR]

OBJECTIVE: We propose a novel algorithm to adjust prandial insulin dose using sparse blood glucose measurements. The dose is adjusted on the basis of a performance measure for the same meal on the previous day. We determine the best performance measure and tune the algorithm to match the recommendations of experienced physicians. RESEARCH DESIGN AND METHODS: Eleven subjects with type 1 diabetes, using continuous subcutaneous insulin infusion, were recruited (seven women and four men, aged 21-65 years with A1C of 7.1 +/- 1.3%). Basal insulin infusion rates were optimized. Target carbohydrate content for the lunch meal was calculated on the basis of a weight-maintenance diet. Over a period of 2-4 days, subjects were asked to measure their blood glucose according to the algorithm's protocol. Starting with their usual insulin-to-carbohydrate ratio, the insulin bolus dose was titrated downward until postprandial glucose levels were high (180-250 mg/dl [10-14 mmol/l]). Subsequently, physicians made insulin bolus recommendations to normalize postprandial glucose concentrations. Graphical methods were then used to determine the most appropriate performance measure for the algorithm to match the physician's decisions. For the best performance measure, the gain of the controller was determined to be the best match to the dose recommendations of the physicians. RESULTS: The correlation between the clinically determined dose adjustments and those of the algorithm is R2 = 0.95, P < 1e - 18. CONCLUSIONS: We have shown how engineering methods can be melded with medical expertise to develop and refine a dosing algorithm. This algorithm has the potential of drastically simplifying the determination of correct insulin-to-carbohydrate ratios. [ABSTRACT FROM AUTHOR]

The article describes a research activity that proposed a model predictive control (MPC) algorithm for the optimal oral anticoagulant drug administration. It provides information on MPC. It describes the main modules to which the proposed algorithm is based. It discusses the clinical tests conducted on patients who were never treated with warfarin.

A continuous closed-loop insulin delivery system using subcutaneous insulin delivery was evaluated in eight diabetic canines. Continuous glucose profiles were obtained by extrapolation of blood glucose measurements. Insulin delivery rate was calculated, using a model of β-cell insulin secretion, and delivered with a Medtronic MiniMed subcutaneous infusion pump. The model acts like a classic proportional-integral-derivative controller, delivering insulin in proportion to glucose above target, history of past glucose values, and glucose rate of change. For each dog, a proportional gain was set relative to the open-loop total daily dose (TDD) of insulin. Additional gains based on 0.5 x TDD and 1.5 x TDD were also evaluated (gain dose response). Control was initiated 4 h before the meal with a target of 6.7 mmol/l. At the time of the meal, glucose was similar for all three gains (6.0 ± 0.3, 5.2 ± 0.3, and 4.9 ± 0.5 mmol/l for 0.5 x TDD, TDD, and 1.5 x TDD, respectively; P > 0.05) with near-target values restored at the end of experiments (8.2 ± 0.9, 6.0 ± 0.6, and 6.0 ± 0.5, respectively). The peak postprandial glucose level decreased significantly with increasing gain (12.1 ± 0.6, 9.6 ± 1.0, and 8.5 ± 0.6 mmol/l, respectively; P < 0.05). The data demonstrate that closed-loop insulin delivery using the subcutaneous site can provide stable glycemic control within a range of gain. Diabetes 55:1995-2000, 2006 [ABSTRACT FROM AUTHOR]

OBJECTIVE -- To evaluate a fully automated algorithm for the establishment of tight glycemic control in critically ill patients and to compare the results with different routine glucose management protocols of three intensive care units (ICUs) across Europe (Graz, Prague, and London). RESEARCH DESIGN AND METHODS -- Sixty patients undergoing cardiac surgery (age 67 ± 9 years, BMI 27.7 ± 4.9 kg/m², 17 women) with postsurgery blood glucose levels > 120 mg/dl (6.7 mmol/l) were investigated in three different ICUs (20 per center). Patients were randomized to either blood glucose management (target range 80-110 mg/dl [4.4-6.1 mmol/l]) by the fully automated model predictive control (MPC) algorithm (n = 30, 10 per center) or implemented routine glucose management protocols (n = 30, 10 per center). In all patients, arterial glucose was measured hourly to describe the glucose profile until the end of the 1CU stay but for a maximum period of 48 h. RESULTS -- Compared with routine protocols, MPC treatment resulted in a significantly higher percentage of time within the target glycemic range (% median [min-max]: 52 [17-92] vs. 19 [0-71]) over 0-24 h (P < 0.01). Improved glycemic control with MPC treatment was confirmed in patients remaining in the 1CU for 48 h (0-24 h: 50 [17-71] vs. 21 [4-67], P < 0.05, and 24-48 h: 65 [38-96] vs. 25 [8-79], P < 0.05, for MPC [n = 16] vs. routine protocol In = 13], respectively). Two hypoglycemic events (<54 mg/dl [3.0 mmol/l]) were observed with routine protocol treatment. No hypoglycemic event occurred with MPC. CONCLUSIONS-- The data suggest that the MPC algorithm is safe and effective in controlling glycemia in critically ill postsurgery patients. [ABSTRACT FROM AUTHOR]

Background The last two decades have witnessed unprecedented technological progress in the development of continuous glucose sensors, resulting in the first generation of commercial glucose monitors. This has fuelled the development of prototypes of a closed-loop system based on the combination of a continuous monitor, a control algorithm, and an insulin pump. Method A review of electromechanical closed-loop approaches is presented. This is followed by a review of existing prototypes and associated glucose sensors. A literature review was undertaken from 1960 to 2004. Results Two main approaches exist. The extracorporeal s.c.–s.c. approach employs subcutaneous glucose monitoring and subcutaneous insulin delivery. The implantable i.v.–i.p. approach adopts intravenous sampling and intraperitoneal insulin delivery. Feasibility of both solutions has been demonstrated in small-scale laboratory studies using either the classical proportional–integral–derivative controller or a model predictive controller. Performance in the home setting has yet to be demonstrated. Conclusions The glucose monitor remains the main limiting factor in the development of a commercially viable closed-loop system, as presently available monitors fail to demonstrate satisfactory characteristics in terms of reliability and/or accuracy. Regulatory issues are the second limiting factor. Closed-loop systems are likely to be used first by health-care professionals in controlled environments such as intensive care units. [ABSTRACT FROM AUTHOR]

Aims To evaluate an algorithm with glucose prediction capacity and continuous adaptation of patient parameters—a model predictive control (MPC) algorithm—to control blood glucose concentration during fasting conditions in patients with Type 1 diabetes. In the subcutaneous (sc) route within a closed loop system. Methods Paired experiments were performed in six patients. Over 8 h the MPC algorithm was used to control glucose with s.c. insulin administration and two different glucose monitoring protocols: first, the algorithm was provided with intravenous (i.v.) glucose values for insulin dosage calculation directly (i.v.–s.c. route). Then, in the second experiment, i.v. glucose values were fed to the MPC with a delay of 30 min to simulate s.c. glucose measurements (‘s.c.’–s.c. route). In both experiments plasma glucose, insulin dosage, and serum insulin levels were analysed. Results Glucose concentration was brought from hyper- to normoglycaemia and kept in the physiological range (6–7 mmol/l) with both routes in all subjects. Mean glucose concentration reached the threshold of 7 mmol/l approximately 2 (i.v.–s.c. route) and 3 (‘s.c.’–s.c. route) hours after the start of glucose control with the MPC. During the last 2 h of automated glucose control, mean glucose concentration was 6.3 ± 0.2 mmol/l and 6.6 ± 0.3 mmol/l for i.v.–s.c. and ‘s.c.’–s.c. route, respectively. Glucose concentration, insulin doses, and serum insulin levels did not differ significantly between routes ( P > 0.05). Conclusions The MPC algorithm is suitable for glucose control during fasting within an extracorporeal artificial β-cell in the subcutaneous route Type 1 diabetic patients. [ABSTRACT FROM AUTHOR]

An algorithm designed to automatically control insulin delivery was tested in rats with Type 1 diabetes. This nonlinear algorithm included a fading memory component of proportional and derivative errors in order to simulate normal insulin secretion. Error-weighting functions for the proportional and derivative terms were used with a performance index designed for error adaptation. In the first version of the algorithm, the proportional gain was adaptively varied. In the second version, a low rate of basal insulin delivery was adaptively varied. Six 6-h studies with each version were conducted using frequent blood sampling and intravenous insulin delivery. In Version 2 studies, blood glucose levels during the last two hours were well-controlled and significantly lower than in Version 1 (118 ± 2.0 vs. 130 ± 2.9 mg/dL). Neither version produced hypoglycemia. Future research using this algorithm needs to focus on automated glucose sensing in combination with insulin delivery. [ABSTRACT FROM AUTHOR]

Background: Many studies of clients with 'pure' anomia (i.e. word finding difficulties in the absence of semantic impairment) have found a lack of generalization to the naming of items that are not treated in therapy. One explanation for this result is that the phonological output lexicon consists of discrete entries which are not as interconnective as those in the semantic system (Miceli, Amatrano, Capasso, and Caramazza 1996). However, more recent studies suggest that therapy which requires more explicit and unaided access on the part of the client might indeed effect generalization (McNeil, Small, Masterton, and Fossett, 1995; Spencer, Doyle, McNeil, Wambaugh, Park, and Carroll, 2000). Aims: The main aim of our study was to investigate whether requiring a client to name pictures with minimal help (i.e. cueing, repetition) from therapists would result in generalization. Methods & Procedures: We used a single case design, employing multiple baselines across behaviours to control for possible spontaneous recovery. The participant was MB, a 79 year old client who was anomic subsequent to a stroke. Therapy sessions required MB to name a set of pictures. Instead of cues, she was required to describe and 'talk around' each picture for as long as it took until the name came to her. We termed this therapy 'Circumlocutioninduced naming' (CIN). Nonparametric statistical analysis was performed on pre and post test results. Outcomes & Results: Therapy resulted in improvement to untreated as well as to treated words, and there was also a qualitative change in error patterns on untreated words (namely fewer unrelated errors and more semantic errors). Performance on unrelated control tasks did not improve. Conclusions: Our results suggest that CIN may effect a generalized improvement on the phonological output lexicon. We hypothesize that the therapy 'exercises' the impaired link between semantics and phonology to a greater degree than do other methods (e.g. phonemic cueing, repetition). This suggests that clients with moderate to mild anomia might benefit from therapy that requires them to take a more active and independent part in accessing words. We discuss some difficulties in the interpretation of our results, including the possible existence of spontaneous recovery, and the issue of exactly what our therapy targeted. [ABSTRACT FROM AUTHOR]

The effects of unsupervised, computerised rehabilitation of anomia in aphasia were investigated with three single-case studies. The training was carried out in the patients' own homes without any supervision. It included semantic and phonemic tasks as well as written naming tasks. The computer was programmed to progress through a preselected set of tasks according to set criteria of performance. The patient could not alter the preselected sequence, but could start and stop a session at any time. The length of the training was determined by the time it took the patient to work through the full set of tasks. Comparison was made of their progress in naming of trained and untrained words. A cross-over design was used, and baselines were established with two patients. All patients improved in their general naming performance, but to different degrees. A relatively specific effect for the trained words was found in two patients, whereas the third showed a somewhat more general effect. The chosen therapy was probably not equally appropriate for all three patients. It is concluded that oral naming can be improved with unsupervised computer rehabilitation in the home even though the patient is not required to speak during the training. [ABSTRACT FROM AUTHOR]

Diabetes is a chronic disease caused by the inability of the pancreas to produce insulin or problems in the body to use it efficiently. It is one of the fastest growing health challenges affecting more than 400 million people worldwide, according to the World Health Organization. Intensive research is being carried out on artificial intelligence methods to help people with diabetes to optimize the way in which they use insulin, carbohydrate intakes, or physical activity. By predicting upcoming levels of blood glucose concentrations, preventive actions can be taken. Previous research studies using machine learning methods for blood glucose level predictions have mainly focused on the machine learning model used. Little attention has been given to the pre-processing of insulin and carbohydrate signals in order to mimic the human absorption processes. In this manuscript, a recurrent neural network (RNN) based model for predicting upcoming blood glucose levels in people with type 1 diabetes is combined with several carbohydrate and insulin absorption curves in order to optimize the prediction results. The proposed method is applied to data from real patients suffering type 1 diabetes mellitus (T1DM). The achieved results are encouraging, obtaining accuracy levels around 0.510 mmol/L (9.2 mg/dl) in the best scenario. [ABSTRACT FROM AUTHOR]

Type 1 diabetes is a chronic disease caused by the inability of the pancreas to produce insulin. Patients suffering type 1 diabetes depend on the appropriate estimation of the units of insulin they have to use in order to keep blood glucose levels in range (considering the calories taken and the physical exercise carried out). In recent years, machine learning models have been developed in order to help type 1 diabetes patients with their blood glucose control. These models tend to receive the insulin units used and the carbohydrate taken as inputs and generate optimal estimations for future blood glucose levels over a prediction horizon. The body glucose kinetics is a complex user-dependent process, and learning patient-specific blood glucose patterns from insulin units and carbohydrate content is a difficult task even for deep learning-based models. This paper proposes a novel mechanism to increase the accuracy of blood glucose predictions from deep learning models based on the estimation of carbohydrate digestion and insulin absorption curves for a particular patient. This manuscript proposes a method to estimate absorption curves by using a simplified model with two parameters which are fitted to each patient by using a genetic algorithm. Using simulated data, the results show the ability of the proposed model to estimate absorption curves with mean absolute errors below 0.1 for normalized fast insulin curves having a maximum value of 1 unit. [ABSTRACT FROM AUTHOR]

In this study we examined the value of errorless learning as a means of restoring memory for proper names in an individual who had becomedysnomic following herpes simplex encephalitis. In Experiment 1 the errorless learning technique was used to teach the subject (PM) the names of eightpoliticians thathe had forgotten. In evaluating thetechnique, we also examined whether there was any generalisation to the naming of untrained politicians' names that he had also forgotten and whether the technique had any effect on politicians'names he was able to produce at the outset of training. Training was successful in that PM reached criterion on naming the trained faces but there was no recovery of untrained names. Moreover, we also observed inhibition of politicians' name he was able to produce at the outset of the study. Experiment 2 used a similar procedure to teach PM the names of personal friends that he could not remember. However, the technique was modified so as to include a ''refresh" trial at the end of each training session for those friends' names he was already able to produce. Errorless learning again facilitated the reacquisition of names but there was no inhibition of names available at the outset. These data illustrate that errorless learning can be used as a means of restoring lost vocabulary but demonstrate how retrieval inhibition is an important element in evaluating overall effectiveness. In addition, the data are discussed in relation to recent theories of memory restoration. [ABSTRACT FROM AUTHOR]

The most obvious neuropsychological application of cognitive psychology would concern rehabilitation for neurological patients with deficits in cognitive processes. We debate here the extent to which theoretical cognitive psychology is currently contributing to the design of rehabilitation. On one side of the debate our assessment suggests that, at present, theories from cognitive psychology have link impact on the majority of treatment programmes; on the other side we note some current and future theoretical developments which could enhance rapport between theory and therapy. [ABSTRACT FROM AUTHOR]

When, as projected, nearly one out of every four Americans is age sixty‐five or more, one out of every four Americans over sixty‐five will spend time in a nursing home, and caregiver families scarcely exist, we will no longer be able to avert our eyes from the grim options faced by our senior citizens. Nor will we be able to continue to operate under the comforting but often false illusion that our convenient dispositions are actually serving the best interests of the elderly. [ABSTRACT FROM PUBLISHER]

Accurate glucose prediction along a long-enough time horizon is a key component for technology to improve type 1 diabetes treatment. Subjects with diabetes might benefit from supervision and control systems that accurately predict risks and trigger corrective actions early enough with improved mitigation. However, large intra-patient variability poses big challenges to glucose prediction. In previous works by the authors, clustering and local modeling techniques with seasonal stochastic models proved to be efficient, allowing for good glucose prediction accuracy for long prediction horizons. Continuous glucose monitoring (CGM) data were partitioned into fixed-length postprandial time subseries and clustered with Fuzzy C-Means to collect similar behaviors, enforcing seasonality at each cluster after subseries concatenation. Then, seasonal stochastic models were identified for each cluster and local predictions were integrated into a global prediction. However, free-living conditions do not support the fixed-length partition of CGM data since daily events duration is variable. In this work, a new algorithm is provided to overcome this constraint, allowing better coping with patient's variability under variable-length time-stamped daily events in supervision and control applications. Besides predicted glucose, two real-time indices are additionally provided—a crispness index, indicating good representation of current glucose behavior by a single model, and a normality index, allowing for the detection of an abnormal glucose behavior (unusual according to registered historical data). The framework is tested in a proof-of-concept in silico study with ten patients over four month training data and two independent two month validation datasets, with and without abnormal behaviors, from the distribution version of the UVA/Padova simulator extended with diverse sources of intra-patient variability. [ABSTRACT FROM AUTHOR]

Accurate estimations for the near future levels of blood glucose are crucial for Type 1 Diabetes Mellitus (T1DM) patients in order to be able to react on time and avoid hypo and hyper-glycemic episodes. Accurate predictions for blood glucose are the base for control algorithms in glucose regulating systems such as the artificial pancreas. Numerous research studies have already been conducted in order to provide predictions for blood glucose levels with particularities in the input signals and underlying models used. These models can be categorized into two major families: those based on tuning glucose physiological-metabolic models and those based on learning glucose evolution patterns based on machine learning techniques. This paper reviews the state of the art in blood glucose predictions for T1DM patients and proposes, implements, validates and compares a new hybrid model that decomposes a deep machine learning model in order to mimic the metabolic behavior of physiological blood glucose methods. The differential equations for carbohydrate and insulin absorption in physiological models are modeled using a Recurrent Neural Network (RNN) implemented using Long Short-Term Memory (LSTM) cells. The results show Root Mean Square Error (RMSE) values under 5 mg/dL for simulated patients and under 10 mg/dL for real patients. [ABSTRACT FROM AUTHOR]