Adaptive System Identification for Estimating Future Glucose Concentrations and Hypoglycemia Alarms

Generalization of a Deep Learning Model for Continuous Glucose Monitoring-Based Hypoglycemia Prediction: Algorithm Development and Validation Study

Background

Predicting hypoglycemia while maintaining a low false alarm rate is a challenge for the wide adoption of continuous glucose monitoring (CGM) devices in diabetes management. One small study suggested that a deep learning model based on the long short-term memory (LSTM) network had better performance in hypoglycemia prediction than traditional machine learning algorithms in European patients with type 1 diabetes. However, given that many well-recognized deep learning models perform poorly outside the training setting, it remains unclear whether the LSTM model could be generalized to different populations or patients with other diabetes subtypes.

Objective

The aim of this study was to validate LSTM hypoglycemia prediction models in more diverse populations and across a wide spectrum of patients with different subtypes of diabetes.

Methods

We assembled two large data sets of patients with type 1 and type 2 diabetes. The primary data set including CGM data from 192 Chinese patients with diabetes was used to develop the LSTM, support vector machine (SVM), and random forest (RF) models for hypoglycemia prediction with a prediction horizon of 30 minutes. Hypoglycemia was categorized into mild (glucose=54-70 mg/dL) and severe (glucose<54 mg/dL) levels. The validation data set of 427 patients of European-American ancestry in the United States was used to validate the models and examine their generalizations. The predictive performance of the models was evaluated according to the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC).

Results

For the difficult-to-predict mild hypoglycemia events, the LSTM model consistently achieved AUC values greater than 97% in the primary data set, with a less than 3% AUC reduction in the validation data set, indicating that the model was robust and generalizable across populations. AUC values above 93% were also achieved when the LSTM model was applied to both type 1 and type 2 diabetes in the validation data set, further strengthening the generalizability of the model. Under different satisfactory levels of sensitivity for mild and severe hypoglycemia prediction, the LSTM model achieved higher specificity than the SVM and RF models, thereby reducing false alarms.

Conclusions

Our results demonstrate that the LSTM model is robust for hypoglycemia prediction and is generalizable across populations or diabetes subtypes. Given its additional advantage of false-alarm reduction, the LSTM model is a strong candidate to be widely implemented in future CGM devices for hypoglycemia prediction.

Combined Use of Glucose-Specific Model Identification and Alarm Strategy Based on Prediction-Funnel to Improve Online Forecasting of Hypoglycemic Events

BACKGROUND

Advanced decision support systems for type 1 diabetes (T1D) management often embed prediction modules, which allow T1D people to take preventive actions to avoid critical episodes like hypoglycemia. Real-time prediction of blood glucose (BG) concentration relies on a subject-specific model of glucose-insulin dynamics. Model parameter identification is usually based on the mean square error (MSE) cost function, and the model is usually used to predict BG at a single prediction horizon (PH). Finally, a hypo-alarm is raised if the predicted BG crosses a threshold. This work aims to show that real-time hypoglycemia forecasting can be improved by leveraging: a glucose-specific mean square error (gMSE) cost function in model's parameters identification, and a "prediction-funnel," that is, confidence intervals (CIs) for multiple PHs, within the hypo-alarm-raising strategy.

METHODS

Autoregressive integrated moving average with exogenous input (ARIMAX) models are selected to illustrate the proposed solution (use of gMSE and prediction-funnel) and its assessment against the conventional approach (MSE and single PH). The gMSE penalizes the model misfit in unsafe BG ranges (e.g., hypoglycemia), and the prediction-funnel allows raising an alarm by monitoring if the CIs cross a suitable threshold. The algorithms were evaluated by measuring precision (P), recall (R), F1-score (F1), false positive per day (FP/day), and time gain (TG) on a real dataset collected in 11 T1D individuals.

RESULTS

The best performance is achieved exploiting both the gMSE and the prediction-funnel: P = 65%, R = 88%, F1 = 75%, FP/day = 0.29, and mean TG = 15 minutes.

CONCLUSIONS

The combined use of a glucose-specific metric and an alarm-raising strategy based on the prediction-funnel allows achieving a more effective and reliable hypoglycemia prediction algorithm.

Observer-based dynamical pattern recognition via deterministic learning

In this paper, based on the sampled-data observer and the deterministic learning theory, a rapid dynamical pattern recognition approach is proposed for univariate time series composed of the output signals of the dynamical systems. Specifically, locally-accurate identification of inherent dynamics of univariate time series is first achieved by using the sampled-data observer and the radial basis function (RBF) networks. The dynamical estimators embedded with the learned knowledge are then designed by resorting to the sampled-data observer. It is proved that generated estimator residuals can reflect the difference between the system dynamics of the training and test univariate time series. Finally, a recognition decision-making scheme is proposed based on the residual norms of the dynamical estimators. Through rigorous analysis, recognition conditions are given to guarantee the accurate recognition of the dynamical pattern of the test univariate time series. The significance of this paper lies in that the difficult problems of dynamical modeling and rapid recognition for univariate time series are solved by incorporating the sampled-data observer design and the deterministic learning theory. The effectiveness of the proposed approach is confirmed by a numerical example and compressor stall warning experiments.

GLYFE: review and benchmark of personalized glucose predictive models in type 1 diabetes

Due to the sensitive nature of diabetes-related data, preventing them from being easily shared between studies, and the wide discrepancies in their data processing pipeline, progress in the field of glucose prediction is hard to assess. To address this issue, we introduce GLYFE (GLYcemia Forecasting Evaluation), a benchmark of machine learning-based glucose predictive models. We present the accuracy and clinical acceptability of nine different models coming from the literature, from standard autoregressive to more complex neural network-based models. These results are obtained on two different datasets, namely UVA/Padova Type 1 Diabetes Metabolic Simulator (T1DMS) and Ohio Type-1 Diabetes Mellitus (OhioT1DM), featuring artificial and real type 1 diabetic patients respectively. By providing extensive details about the data flow as well as by providing the whole source code of the benchmarking process, we ensure the reproducibility of the results and the usability of the benchmark by the community. Those results serve as a basis of comparison for future studies. In a field where data are hard to obtain, and where the comparison of results from different studies is often irrelevant, GLYFE gives the opportunity of gathering researchers around a standardized common environment.

A predictive model incorporating the change detection and Winsorization methods for alerting hypoglycemia and hyperglycemia

This paper focuses on establishing an effective predictive model to quickly and accurately alert hypoglycemia and hyperglycemia for helping control blood glucose levels of people with diabetes. In general, a good predictive model is established on the features of data. Inspired by this, we first analyze the characteristics of continuous glucose monitoring (CGM) data by the equality of variances test and outlier detection, which show time-varying fluctuations and jump points in CGM data. Therefore, we incorporate the change detection method and the Winsorization method into the predictive model based on the autoregressive moving average (ARMA) model and the recursive least squares (RLS) method to fit the above characteristics. To the best of our knowledge, the proposed method is the first attempt to give a solution for matching the time-varying fluctuations and jump points of CGM data simultaneously. A case study using CGM data is given to validate the effectiveness of the proposed method under 30-min-ahead prediction. The results show that the proposed method can improve the true alarm ratio of hypoglycemia and hyperglycemia from 0.7983 to 0.8783, and lengthen the average advance detection time of hypoglycemia and hyperglycemia from 19.77 to 22.64 min.

A review of biosensor technology and algorithms for glucose monitoring

Diabetes mellitus (DM) has become a serious illness in the whole world. Until now, there is no effective cure for patients with DM. It is well known that the glucose level is one key factor to determine the progress of DM. It is also an important reference to carry out the accurate and timely treatment for patients with DM. In this article, the related biosensors technology that can be utilized to identify and predict glucose level are reviewed in detail, including the algorithms that can help to achieve numerical value of glucose level. Firstly, the biosensor technology based on the physiological fluids are illustrated, including blood, sweat, interstitial fluid, ocular fluid, and other available fluids. Secondly, the algorithms for achieving numerical value of glucose level are investigated, including the physiological model-based method and the machine learning-based method. Finally, the future development trend and challenges of glucose level monitoring are given and the conclusions are drawn.

Adversarial multi-source transfer learning in healthcare: Application to glucose prediction for diabetic people

BACKGROUND AND OBJECTIVES

Deep learning has yet to revolutionize general practices in healthcare, despite promising results for some specific tasks. This is partly due to data being in insufficient quantities hurting the training of the models. To address this issue, data from multiple health actors or patients could be combined by capitalizing on their heterogeneity through the use of transfer learning.

METHODS

To improve the quality of the transfer between multiple sources of data, we propose a multi-source adversarial transfer learning framework that enables the learning of a feature representation that is similar across the sources, and thus more general and more easily transferable. We apply this idea to glucose forecasting for diabetic people using a fully convolutional neural network. The evaluation is done by exploring various transfer scenarios with three datasets characterized by their high inter and intra variability.

RESULTS

While transferring knowledge is beneficial in general, we show that the statistical and clinical accuracies can be further improved by using of the adversarial training methodology, surpassing the current state-of-the-art results. In particular, it shines when using data from different datasets, or when there is too little data in an intra-dataset situation. To understand the behavior of the models, we analyze the learnt feature representations and propose a new metric in this regard. Contrary to a standard transfer, the adversarial transfer does not discriminate the patients and datasets, helping the learning of a more general feature representation.

CONCLUSION

The adversarial training framework improves the learning of a general feature representation in a multi-source environment, enhancing the knowledge transfer to an unseen target. The proposed method can help improve the efficiency of data shared by different health actors in the training of deep models.

Comparing Machine Learning Techniques for Blood Glucose Forecasting Using Free-living and Patient Generated Data

Managing a chronic disease like Type 1 diabetes (T1D) is both challenging and time consuming, but new technologies that allow continuous measurement of glucose and delivery of insulin have led to significant improvements. The development of an artificial pancreas (AP), which algorithmically determines insulin dosing and delivers insulin in a fully automated way, may transform T1D care but it is not yet widely available. Patient-led alternatives, like the Open Artificial Pancreas (OpenAPS), are being used by hundreds of individuals and have also led to a dramatic increase in the availability of patient generated health data (PGHD). All APs require an accurate forecast of blood glucose (BG). While there have been efforts to develop better forecasts and apply new ML techniques like deep learning to this problem, methods are often tested on small controlled datasets that do not indicate how they may perform in reality - and the most advanced methods have not always outperformed the simplest. We introduce a rigorous comparison of BG forecasting using both a small controlled research dataset and large heterogeneous PGHD. Our comparison advances the state of the art in BG forecasting by providing insight into how methods may fare when moving beyond small controlled studies to real-world use.

A Data-Driven Personalized Model of Glucose Dynamics Taking Account of the Effects of Physical Activity for Type 1 Diabetes: An In Silico Study

This paper aims to develop a data-driven model for glucose dynamics taking into account the effects of physical activity (PA) through a numerical study. It intends to investigate PA's immediate effect on insulin-independent glucose variation and PA's prolonged effect on insulin sensitivity. We proposed a nonlinear model with PA (NLPA), consisting of a linear regression of PA and a bilinear regression of insulin and PA. The model was identified and evaluated using data generated from a physiological PA-glucose model by Dalla Man et al. integrated with the uva/padova Simulator. Three metrics were computed to compare blood glucose (BG) predictions by NLPA, a linear model with PA (LPA), and a linear model with no PA (LOPA). For PA's immediate effect on glucose, NLPA and LPA showed 45-160% higher mean goodness of fit (FIT) than LOPA under 30 min-ahead glucose prediction (P < 0.05). For the prolonged PA effect on glucose, NLPA showed 87% higher FIT than LPA (P < 0.05) for simulations using no previous measurements. NLPA had 25-37% and 31-54% higher sensitivity in predicting postexercise hypoglycemia than LPA and LOPA, respectively. This study demonstrated the following qualitative trends: (1) for moderate-intensity exercise, accuracy of BG prediction was improved by explicitly accounting for PA's effect; and (2) accounting for PA's prolonged effect on insulin sensitivity can increase the chance of early prediction of postexercise hypoglycemia. Such observations will need to be further evaluated through human subjects in the future.

Predicting Nocturnal Hypoglycemia from Continuous Glucose Monitoring Data with Extended Prediction Horizon

Nocturnal hypoglycemia is a serious complication of insulin-treated diabetes, which commonly goes undetected. Continuous glucose monitoring (CGM) devices have enabled prediction of impending nocturnal hypoglycemia, however, prior efforts have been limited to a short prediction horizon (~ 30 minutes). To this end, a nocturnal hypoglycemia prediction model with a 6-hour horizon (midnight-6 am) was developed using a random forest machine- learning model based on data from 10,000 users with more than 1 million nights of CGM data. The model demonstrated an overall nighttime hypoglycemia prediction performance of ROC AUC = 0.84, with AUC = 0.90 for early night (midnight-3 am) and AUC = 0.75 for late night (prediction at midnight, looking at 3-6 am window). While instabilities and the absence of late-night blood glucose patterns introduce predictability challenges, this 6-hour horizon model demonstrates good performance in predicting nocturnal hypoglycemia. Additional study and specific patient-specific features will provide refinements that further ensure safe overnight management of glycemia.

Seasonal Local Models for Glucose Prediction in Type 1 Diabetes

Linear empirical dynamic models have been widely used for blood glucose prediction and risks prevention in people with type 1 diabetes. More accurate blood glucose prediction models with longer prediction horizon (PH) are desirable to enable warnings to patients about imminent blood glucose changes with enough time to take corrective actions. In this study, a blood glucose prediction method is developed by integrating the predictions of a set of seasonal local models (each of them corresponding to different glucose profiles observed along historical data). In the modeling step, the number of sets and their corresponding glucose profiles characteristics are obtained by clustering techniques (Fuzzy C-Means). Then, Box-Jenkins methodology is used to identify a seasonal model for each set. Finally, blood glucose predictions of local models are integrated using different techniques. The proposed method is tested by using 18 60-h closed-loop experiments (including different exercise types and artificial pancreas strategies) and achieving mean absolute percentage error (MAPE) of 2.94%, 3.89%, 5.41%, 6.29% and 8.66% for 15-, 30-, 45-, 60-, and 90-min PHs, respectively.

Simulation Software for Assessment of Nonlinear and Adaptive Multivariable Control Algorithms: Glucose - Insulin Dynamics in Type 1 Diabetes

A simulator for testing automatic control algorithms for nonlinear systems with time-varying parameters, variable time delays, and uncertainties is developed. It is based on simulation of virtual patients with Type 1 diabetes (T1D). Nonlinear models are developed to describe glucose concentration (GC) variations based on user-defined scenarios for meal consumption, insulin administration, and physical activity. They compute GC values and physiological variables, such as heart rate, skin temperature, accelerometer, and energy expenditure, that are indicative of physical activities affecting GC dynamics. This is the first simulator designed for assessment of multivariable controllers that consider supplemental physiological variables in addition to GC measurements to improve glycemic control. Virtual patients are generated from distributions of identified model parameters using clinical data. The simulator will enable testing and evaluation of new control algorithms proposed for automated insulin delivery as well as various control algorithms for nonlinear systems with uncertainties, time-varying parameters and delays.

Long-Term Glucose Forecasting Using a Physiological Model and Deconvolution of the Continuous Glucose Monitoring Signal

(1) Objective: Blood glucose forecasting in type 1 diabetes (T1D) management is a maturing field with numerous algorithms being published and a few of them having reached the commercialisation stage. However, accurate long-term glucose predictions (e.g., >60 min), which are usually needed in applications such as precision insulin dosing (e.g., an artificial pancreas), still remain a challenge. In this paper, we present a novel glucose forecasting algorithm that is well-suited for long-term prediction horizons. The proposed algorithm is currently being used as the core component of a modular safety system for an insulin dose recommender developed within the EU-funded PEPPER (Patient Empowerment through Predictive PERsonalised decision support) project. (2) Methods: The proposed blood glucose forecasting algorithm is based on a compartmental composite model of glucose-insulin dynamics, which uses a deconvolution technique applied to the continuous glucose monitoring (CGM) signal for state estimation. In addition to commonly employed inputs by glucose forecasting methods (i.e., CGM data, insulin, carbohydrates), the proposed algorithm allows the optional input of meal absorption information to enhance prediction accuracy. Clinical data corresponding to 10 adult subjects with T1D were used for evaluation purposes. In addition, in silico data obtained with a modified version of the UVa-Padova simulator was used to further evaluate the impact of accounting for meal absorption information on prediction accuracy. Finally, a comparison with two well-established glucose forecasting algorithms, the autoregressive exogenous (ARX) model and the latent variable-based statistical (LVX) model, was carried out. (3) Results: For prediction horizons beyond 60 min, the performance of the proposed physiological model-based (PM) algorithm is superior to that of the LVX and ARX algorithms. When comparing the performance of PM against the secondly ranked method (ARX) on a 120 min prediction horizon, the percentage improvement on prediction accuracy measured with the root mean square error, A-region of error grid analysis (EGA), and hypoglycaemia prediction calculated by the Matthews correlation coefficient, was 18.8 % , 17.9 % , and 80.9 % , respectively. Although showing a trend towards improvement, the addition of meal absorption information did not provide clinically significant improvements. (4) Conclusion: The proposed glucose forecasting algorithm is potentially well-suited for T1D management applications which require long-term glucose predictions.

Feature Selection for Blood Glucose Level Prediction in Type 1 Diabetes Mellitus by Using the Sequential Input Selection Algorithm (SISAL)

Feature selection is a primary exercise to tackle any forecasting task. Machine learning algorithms used to predict any variable can improve their performance by lessening their computational effort with a proper dataset. Anticipating future glycemia in type 1 diabetes mellitus (DM1) patients provides a baseline in its management, and in this task, we need to carefully select data, especially now, when novel wearable devices offer more and more information. In this paper, a complete characterization of 25 diabetic people has been carried out, registering innovative variables like sleep, schedule, or heart rate in addition to other well-known ones like insulin, meal, and exercise. With this ground-breaking data compilation, we present a study of these features using the Sequential Input Selection Algorithm (SISAL), which is specially prepared for time series data. The results rank features according to their importance, regarding their relevance in blood glucose level prediction as well as indicating the most influential past values to be taken into account and distinguishing features with person-dependent behavior from others with a common performance in any patient. These ideas can be used as strategies to select data for predicting glycemia depending on the availability of computational power, required speed, or required accuracy. In conclusion, this paper tries to analyze if there exists symmetry among the different features that can affect blood glucose levels, that is, if their behavior is symmetric in terms of influence in glycemia.

Improving Glucose Prediction Accuracy in Physically Active Adolescents With Type 1 Diabetes

BACKGROUND

Physical activity presents a significant challenge for glycemic control in individuals with type 1 diabetes. As accurate glycemic predictions are key to successful automated decision-making systems (eg, artificial pancreas, AP), the inclusion of additional physiological variables in the estimation of the metabolic state may improve the glucose prediction accuracy during exercise.

METHODS

Predictor-based subspace identification is applied to a dynamic glucose prediction model including heart rate measurements along with variables representing the carbohydrate consumption and insulin boluses. To demonstrate the improvement in prediction ability due to the additional heart rate variable, the performance of the proposed modeling technique is evaluated with (SID-HR) and without heart rate (SID-2) as an additional input using experimental data involving adolescents at ski camp. Furthermore, the performance of the proposed approach is compared to that of the metabolic state observer (MSO) model currently used in the University of Virginia AP algorithm.

RESULTS

The addition of heart rate in the subspace-based model (SID-HR) yields a statistically significant improvement in the root-mean-square error compared to the SID-2 model (P < .001) and the standard MSO (P < .001). Furthermore, the SID-HR model performed favorably in comparison to the SID-2 and MSO models after accounting for its increased complexity.

CONCLUSIONS

Directly considering the effects of physical activity levels on glycemic dynamics through the inclusion of heart rate as an additional input variable in the glucose dynamics model improves the glucose prediction accuracy. The proposed methodology could improve exercise-informed model-based predictive control algorithms in artificial pancreas systems.

Short-term prediction of glucose in type 1 diabetes using kernel adaptive filters

This study aims at presenting a nonlinear, recursive, multivariate prediction model of the subcutaneous glucose concentration in type 1 diabetes. Nonlinear regression is performed in a reproducing kernel Hilbert space, by either the fixed budget quantized kernel least mean square (QKLMS-FB) or the approximate linear dependency kernel recursive least-squares (KRLS-ALD) algorithm, such that a sparse model structure is accomplished. A multivariate feature set (i.e., subcutaneous glucose, food carbohydrates, insulin regime and physical activity) is used and its influence on short-term glucose prediction is investigated. The method is evaluated using data from 15 patients with type 1 diabetes in free-living conditions. In the case when all the input variables are considered: (i) the average root mean squared error (RMSE) of QKLMS-FB increases from 13.1 mg dL^-1 (mean absolute percentage error (MAPE) 6.6%) for a 15-min prediction horizon (PH) to 37.7 mg dL^-1 (MAPE 20.8%) for a 60-min PH and (ii) the RMSE of KRLS-ALD, being predictably lower, increases from 10.5 mg dL^-1 (MAPE 5.2%) for a 15-min PH to 31.8 mg dL^-1 (MAPE 18.0%) for a 60-min PH. Multivariate data improve systematically both the regularity and the time lag of the predictions, reducing the errors in critical glucose value regions for a PH ≥ 30 min. Graphical abstract ᅟ.

An ARIMA Model With Adaptive Orders for Predicting Blood Glucose Concentrations and Hypoglycemia

The continuous glucose monitoring system is an effective tool, which enables the users to monitor their blood glucose (BG) levels. Based on the continuous glucose monitoring (CGM) data, we aim at predicting future BG levels so that appropriate actions can be taken in advance to prevent hyperglycemia or hypoglycemia. Due to the time-varying nonstationarity of CGM data, verified by Augmented Dickey-Fuller test and analysis of variance, an autoregressive integrated moving average (ARIMA) model with an adaptive identification algorithm of model orders is proposed in the prediction framework. Such identification algorithm adaptively determines the model orders and simultaneously estimates the corresponding parameters using Akaike Information Criterion and least square estimation. A case study is conducted with the CGM data of diabetics under daily living conditions to analyze the prediction performance of the proposed model together with the early hypoglycemic alarms. Results show that the proposed model outperforms the adaptive univariate model and ARIMA model.

Prediction of Adverse Glycemic Events From Continuous Glucose Monitoring Signal

The most important objective of any diabetes therapy is to maintain the blood glucose concentration within the euglycemic range, avoiding or at least mitigating critical hypo/hyperglycemic episodes. Modern continuous glucose monitoring (CGM) devices bear the promise of providing the patients with an increased and timely awareness of glycemic conditions as these get dangerously near to hypo/hyperglycemia. The challenge is to detect, with reasonable advance, the patterns leading to risky situations, allowing the patient to make therapeutic decisions on the basis of future (predicted) glucose concentration levels. We underline that a technically sound performance comparison of the approaches proposed in recent years has yet to be done, thus it is unclear which one is preferred. The aim of this study is to fill this gap by carrying out a comparative analysis among the most common methods for glucose event prediction. Both regression and classification algorithms have been implemented and analyzed, including static and dynamic training approaches. The dataset consists of 89 CGM time series measured in diabetic subjects for 7 subsequent days. Performance metrics, specifically defined to assess and compare the event-prediction capabilities of the methods, have been introduced and analyzed. Our numerical results show that a static training approach exhibits better performance, in particular when regression methods are considered. However, classifiers show some improvement when trained for a specific event category, such as hyperglycemia, achieving performance comparable to the regressors, with the advantage of predicting the events sooner.

Model-Fusion-Based Online Glucose Concentration Predictions in People with Type 1 Diabetes

Accurate predictions of glucose concentrations are necessary to develop an artificial pancreas (AP) system for people with type 1 diabetes (T1D). In this work, a novel glucose forecasting paradigm based on a model fusion strategy is developed to accurately characterize the variability and transient dynamics of glycemic measurements. To this end, four different adaptive filters and a fusion mechanism are proposed for use in the online prediction of future glucose trajectories. The filter fusion mechanism is developed based on various prediction performance indexes to guide the overall output of the forecasting paradigm. The efficiency of the proposed model fusion based forecasting method is evaluated using simulated and clinical datasets, and the results demonstrate the capability and prediction accuracy of the data-based fusion filters, especially in the case of limited data availability. The model fusion framework may be used in the development of an AP system for glucose regulation in patients with T1D.

A Novel Classification Indicator of Type 1 and Type 2 Diabetes in China

Because of the differences of treatment, it is extremely important to classify the types of diabetes, especially for the diagnosis made by clinician. In this study, we proposed a novel scheme calculating an indicator of classifying diabetes, which contains two stages: the first is a model of feature extraction, 17 features are automatically extracted from the curve of glucose concentration acquired by continuous glucose monitoring system (CGM); the second is a model of diabetes parameter regression based on an ensemble learning algorithm named double-Class AdaBoost. 1050 curves of glucose concentration of type 1 and type 2 diabetics were acquired at the Department of Endocrinology in People's Hospital of Zhengzhou University China, and an upper threshold μ was set to 7 mmol/L, 8 mmol/L, 9 mmol/L, 10 mmo/L, and 11 mmol/L respectively according to the guideline of WHO. The experiments show that the coincidence rate of our scheme and clinical diagnosis is 90.3%. The novel indicator extends the criteria in diagnosing types of diabetes and provides doctors with a scalar to classify diabetes of type 1 and type 2.

Kernel-based adaptive learning improves accuracy of glucose predictive modelling in type 1 diabetes: A proof-of-concept study

This study aims at demonstrating the need for nonlinear recursive models to the identification and prediction of the dynamic glucose system in type 1 diabetes. Nonlinear regression is performed in a reproducing kernel Hilbert space, by the Approximate Linear Dependency Kernel Recursive Least Squares (KRLS-ALD) algorithm, such that a sparse model structure is accomplished. The method is evaluated on seven people with type 1 diabetes in free-living conditions, where a change in glycaemic dynamics is forced by increasing the level of physical activity in the middle of the observational period. The univariate input allows for short-term (≤30 min) predictions with KRLS-ALD reaching an average root mean square error of 15.22±5.95 mgdL^-1 and an average time lag of 17.14±2.67 min for an horizon of 30 min. Its performance is considerably better than that of time-invariant (regularized) linear regression models.

Non-linear dynamic modeling of glucose in type 1 diabetes with kernel adaptive filters

We propose a non-linear recursive solution to the problem of short-term prediction of glucose in type 1 diabetes. The Fixed Budget Quantized Kernel Least Mean Square (QKLMS-FB) algorithm is employed to construct a univariate model of subcutaneous glucose concentration, which: (i) handles nonlinearities by transforming the input space into a high-dimensional Reproducing Kernel Hilbert Space and, (ii) finds a sparse solution by retaining a representative subset of the training input vectors. The dataset comes from the continuous multi-day recordings of 15 type 1 patients in free-living conditions. QKLMS-FB produces an average root mean squared error of 18.66±3.19 mg/dl for a prediction horizon of 30 min with 82.04% of hypoglycemic readings and 93.30% of hyperglycemic ones being classified as clinically accurate or with benign errors. The effect of the prediction horizon is more evident in the hypoglycemic range.

How Much Is Short-Term Glucose Prediction in Type 1 Diabetes Improved by Adding Insulin Delivery and Meal Content Information to CGM Data? A Proof-of-Concept Study

BACKGROUND

In type 1 diabetes (T1D) management, short-term glucose prediction can allow to anticipate therapeutic decisions when hypo/hyperglycemia is imminent. Literature prediction methods mainly use past continuous glucose monitoring (CGM) readings. Sophisticated algorithms can use information on insulin delivered and meal carbohydrate (CHO) content. The quantification of how much insulin and CHO information improves glucose prediction is missing in the literature and is investigated, in an open-loop setting, in this proof-of-concept study.

METHODS

We adopted a versatile literature prediction methodology able to utilize a variety of inputs. We compared predictors that use (1) CGM; (2) CGM and insulin; (3) CGM and CHO; and (4) CGM, insulin, and CHO. Data of 15 T1D subjects in open-loop setup were used. Prediction was evaluated via absolute error and temporal gain focusing on meal/night periods. The relative importance of each individual input of the predictor was evaluated with a sensitivity analysis.

RESULTS

For a prediction horizon (PH) ≥ 30 minutes, insulin and CHO information improves prediction accuracy of 10% and double the temporal gain during the 2 hours following the meal. During the night the 4 methods did not give statistically different results. When PH ≥ 45 minutes, the influence of CHO information on prediction is 5-fold that of insulin.

CONCLUSIONS

In an open-loop setting, with PH ≥ 30 minutes, information on CHO and insulin improves short-term glucose prediction in the 2-hour time window following a meal, but not during the night. CHO information improves prediction significantly more than insulin.

Real-Time Model-Based Fault Detection of Continuous Glucose Sensor Measurements

OBJECTIVE

Faults in subcutaneous glucose concentration readings with a continuous glucose monitoring (CGM) may affect the computation of insulin infusion rates that can lead to hypoglycemia or hyperglycemia in artificial pancreas control systems for patients with type 1 diabetes (T1D).

METHODS

Multivariable statistical monitoring methods are proposed for detection of faults in glucose concentration values reported by a subcutaneous glucose sensor. A nonlinear first principle glucose/insulin/meal dynamic model is developed. An unscented Kalman filter is used for state and parameter estimation of the nonlinear model. Principal component analysis models are developed and used for detection of dynamic changes. K-nearest neighbor classification algorithm is used for diagnosis of faults. Data from 51 subjects are used to assess the performance of the algorithm.

RESULTS

The results indicate that the proposed algorithm works successfully with 84.2% sensitivity. Overall, 155 (out of 184) of the CGM failures are detected with a 2.8-min average detection time.

CONCLUSION

A novel algorithm that integrates data-driven and model-based methods is developed. The proposed method is able to detect CGM failures with a high rate of success.

SIGNIFICANCE

The proposed fault detection algorithm can decrease the effects of faults on insulin infusion rates and reduce the potential for hypo- or hyperglycemia for patients with T1D.

Empirical Dynamic Model Identification for Blood-Glucose Dynamics in Response to Physical Activity

In this paper, the dynamic response of blood glucose concentration in response to physical activity of people with Type 1 Diabetes Mellitus (T1DM) is captured by subspace identification methods. Activity (input) and subcutaneous blood glucose measurements (output) are employed to construct a personalized prediction model through semi-definite programming. The model is calibrated and subsequently validated with non-overlapping data sets from 15 T1DM subjects. This preliminary clinical evaluation reveals the underlying linear dynamics between blood glucose concentration and physical activity. These types of models can enhance our capabilities of achieving tighter blood glucose control and early detection of hypoglycemia for people with T1DM.

Physical Activity Capture Technology With Potential for Incorporation Into Closed-Loop Control for Type 1 Diabetes

Physical activity is an important determinant of glucose variability in type 1 diabetes (T1D). It has been incorporated as a nonglucose input into closed-loop control (CLC) protocols for T1D during the last 4 years mainly by 3 research groups in single center based controlled clinical trials involving a maximum of 18 subjects in any 1 study. Although physical activity data capture may have clinical benefit in patients with T1D by impacting cardiovascular fitness and optimal body weight achievement and maintenance, limited number of such studies have been conducted to date. Clinical trial registries provide information about a single small sample size 2 center prospective study incorporating physical activity data input to modulate closed-loop control in T1D that are seeking to build on prior studies. We expect an increase in such studies especially since the NIH has expanded support of this type of research with additional grants starting in the second half of 2015. Studies (1) involving patients with other disorders that have lasted 12 weeks or longer and tracked physical activity and (2) including both aerobic and resistance activity may offer insights about the user experience and device optimization even as single input CLC heads into real-world clinical trials over the next few years and nonglucose input is introduced as the next advance.

Multi-model data fusion to improve an early warning system for hypo-/hyperglycemic events

Correct predictions of future blood glucose levels in individuals with Type 1 Diabetes (T1D) can be used to provide early warning of upcoming hypo-/hyperglycemic events and thus to improve the patient's safety. To increase prediction accuracy and efficiency, various approaches have been proposed which combine multiple predictors to produce superior results compared to single predictors. Three methods for model fusion are presented and comparatively assessed. Data from 23 T1D subjects under sensor-augmented pump (SAP) therapy were used in two adaptive data-driven models (an autoregressive model with output correction - cARX, and a recurrent neural network - RNN). Data fusion techniques based on i) Dempster-Shafer Evidential Theory (DST), ii) Genetic Algorithms (GA), and iii) Genetic Programming (GP) were used to merge the complimentary performances of the prediction models. The fused output is used in a warning algorithm to issue alarms of upcoming hypo-/hyperglycemic events. The fusion schemes showed improved performance with lower root mean square errors, lower time lags, and higher correlation. In the warning algorithm, median daily false alarms (DFA) of 0.25%, and 100% correct alarms (CA) were obtained for both event types. The detection times (DT) before occurrence of events were 13.0 and 12.1 min respectively for hypo-/hyperglycemic events. Compared to the cARX and RNN models, and a linear fusion of the two, the proposed fusion schemes represents a significant improvement.

Hypo- and Hyperglycemic Alarms: Devices and Algorithms

Soon after the discovery that insulin regulates blood glucose by Banting and Best in 1922, the symptoms and risks associated with hypoglycemia became widely recognized. This article reviews devices to warn individuals of impending hypo- and hyperglycemia; biosignals used by these devices include electroencephalography, electrocardiography, skin galvanic resistance, diabetes alert dogs, and continuous glucose monitors (CGMs). While systems based on other technology are increasing in performance and decreasing in size, CGM technology remains the best method for both reactive and predictive alarming of hypo- or hyperglycemia.

Evaluation of short-term predictors of glucose concentration in type 1 diabetes combining feature ranking with regression models

Glucose concentration in type 1 diabetes is a function of biological and environmental factors which present high inter-patient variability. The objective of this study is to evaluate a number of features, which are extracted from medical and lifestyle self-monitoring data, with respect to their ability to predict the short-term subcutaneous (s.c.) glucose concentration of an individual. Random forests (RF) and RReliefF algorithms are first employed to rank the candidate feature set. Then, a forward selection procedure follows to build a glucose predictive model, where features are sequentially added to it in decreasing order of importance. Predictions are performed using support vector regression or Gaussian processes. The proposed method is validated on a dataset of 15 type diabetics in real-life conditions. The s.c. glucose profile along with time of the day and plasma insulin concentration are systematically highly ranked, while the effect of food intake and physical activity varies considerably among patients. Moreover, the average prediction error converges in less than d/2 iterations (d is the number of features). Our results suggest that RF and RReliefF can find the most informative features and can be successfully used to customize the input of glucose models.

Jump neural network for real-time prediction of glucose concentration

Prediction of the future value of a variable is of central importance in a wide variety of fields, including economy and finance, meteorology, informatics, and, last but not least important, medicine. For example, in the therapy of Type 1 Diabetes (T1D), in which, for patient safety, glucose concentration in the blood should be maintained in a defined normoglycemic range, the ability to forecast glucose concentration in the short-term (with a prediction horizon of around 30 min) might be sufficient to reduce the incidence of hypoglycemic and hyperglycemic events. Neural Network (NN) approaches are suitable for prediction purposes because of their ability to model nonlinear dynamics and handle in their inputs signals coming from different domains. In this chapter we illustrate the design of a jump NN glucose prediction algorithm that exploits past glucose concentration data, measured in real-time by a minimally invasive continuous glucose monitoring (CGM) sensor, and information on ingested carbohydrates, supplied by the patient himself or herself. The methodology is assessed by tuning the NN on data of ten T1D individuals and then testing it on a dataset of ten different subjects. Results with a prediction horizon of 30 min show that prediction of glucose concentration in T1D via NN is feasible and sufficiently accurate. The average time anticipation obtained is compatible with the generation of preventive hypoglycemic and hyperglycemic alerts and the improvement of artificial pancreas performance.

Multivariable adaptive identification and control for artificial pancreas systems

A constrained weighted recursive least squares method is proposed to provide recursive models with guaranteed stability and better performance than models based on regular identification methods in predicting the variations of blood glucose concentration in patients with Type 1 Diabetes. Use of physiological information from a sports armband improves glucose concentration prediction and enables earlier recognition of the effects of physical activity on glucose concentration. Generalized predictive controllers (GPC) based on these recursive models are developed. The performance of GPC for artificial pancreas systems is illustrated by simulations with UVa-Padova simulator and clinical studies. The controllers developed are good candidates for artificial pancreas systems with no announcements from patients.

An integrated multivariable artificial pancreas control system

The objective was to develop a closed-loop (CL) artificial pancreas (AP) control system that uses continuous measurements of glucose concentration and physiological variables, integrated with a hypoglycemia early alarm module to regulate glucose concentration and prevent hypoglycemia. Eleven open-loop (OL) and 9 CL experiments were performed. A multivariable adaptive artificial pancreas (MAAP) system was used for the first 6 CL experiments. An integrated multivariable adaptive artificial pancreas (IMAAP) system consisting of MAAP augmented with a hypoglycemia early alarm system was used during the last 3 CL experiments. Glucose values and physical activity information were measured and transferred to the controller every 10 minutes and insulin suggestions were entered to the pump manually. All experiments were designed to be close to real-life conditions. Severe hypoglycemic episodes were seen several times during the OL experiments. With the MAAP system, the occurrence of severe hypoglycemia was decreased significantly (P < .01). No hypoglycemia was seen with the IMAAP system. There was also a significant difference (P < .01) between OL and CL experiments with regard to percentage of glucose concentration (54% vs 58%) that remained within target range (70-180 mg/dl). Integration of an adaptive control and hypoglycemia early alarm system was able to keep glucose concentration values in target range in patients with type 1 diabetes. Postprandial hypoglycemia and exercise-induced hypoglycemia did not occur when this system was used. Physical activity information improved estimation of the blood glucose concentration and effectiveness of the control system.

Personalized State-space Modeling of Glucose Dynamics for Type 1 Diabetes Using Continuously Monitored Glucose, Insulin Dose, and Meal Intake: An Extended Kalman Filter Approach

An essential component of any artificial pancreas is on the prediction of blood glucose levels as a function of exogenous and endogenous perturbations such as insulin dose, meal intake, and physical activity and emotional tone under natural living conditions. In this article, we present a new data-driven state-space dynamic model with time-varying coefficients that are used to explicitly quantify the time-varying patient-specific effects of insulin dose and meal intake on blood glucose fluctuations. Using the 3-variate time series of glucose level, insulin dose, and meal intake of an individual type 1 diabetic subject, we apply an extended Kalman filter (EKF) to estimate time-varying coefficients of the patient-specific state-space model. We evaluate our empirical modeling using (1) the FDA-approved UVa/Padova simulator with 30 virtual patients and (2) clinical data of 5 type 1 diabetic patients under natural living conditions. Compared to a forgetting-factor-based recursive ARX model of the same order, the EKF model predictions have higher fit, and significantly better temporal gain and J index and thus are superior in early detection of upward and downward trends in glucose. The EKF based state-space model developed in this article is particularly suitable for model-based state-feedback control designs since the Kalman filter estimates the state variable of the glucose dynamics based on the measured glucose time series. In addition, since the model parameters are estimated in real time, this model is also suitable for adaptive control.

Physical activity measured by physical activity monitoring system correlates with glucose trends reconstructed from continuous glucose monitoring

BACKGROUND

In type 1 diabetes mellitus (T1DM), physical activity (PA) lowers the risk of cardiovascular complications but hinders the achievement of optimal glycemic control, transiently boosting insulin action and increasing hypoglycemia risk. Quantitative investigation of relationships between PA-related signals and glucose dynamics, tracked using, for example, continuous glucose monitoring (CGM) sensors, have been barely explored.

SUBJECTS AND METHODS

In the clinic, 20 control and 19 T1DM subjects were studied for 4 consecutive days. They underwent low-intensity PA sessions daily. PA was tracked by the PA monitoring system (PAMS), a system comprising accelerometers and inclinometers. Variations on glucose dynamics were tracked estimating first- and second-order time derivatives of glucose concentration from CGM via Bayesian smoothing. Short-time effects of PA on glucose dynamics were quantified through the partial correlation function in the interval (0, 60 min) after starting PA.

RESULTS

Correlation of PA with glucose time derivatives is evident. In T1DM, the negative correlation with the first-order glucose time derivative is maximal (absolute value) after 15 min of PA, whereas the positive correlation is maximal after 40-45 min. The negative correlation between the second-order time derivative and PA is maximal after 5 min, whereas the positive correlation is maximal after 35-40 min. Control subjects provided similar results but with positive and negative correlation peaks anticipated of 5 min.

CONCLUSIONS

Quantitative information on correlation between mild PA and short-term glucose dynamics was obtained. This represents a preliminary important step toward incorporation of PA information in more realistic physiological models of the glucose-insulin system usable in T1DM simulators, in development of closed-loop artificial pancreas control algorithms, and in CGM-based prediction algorithms for generation of hypoglycemic alerts.

Signal processing algorithms implementing the "smart sensor" concept to improve continuous glucose monitoring in diabetes

Glucose readings provided by current continuous glucose monitoring (CGM) devices still suffer from accuracy and precision issues. In April 2013, we proposed a new conceptual architecture to deal with these problems and render CGM sensors algorithmically smarter, which consists of three modules for denoising, enhancement, and prediction placed in cascade to a commercial CGM sensor. The architecture was assessed on a data set consisting of 24 type 1 diabetes patients collected in four clinical centers of the AP@home Consortium (a European project of 7th Framework Programme funded by the European Committee). This article, as a companion to our prior publication, illustrates the technical details of the algorithms and of the implementation issues.

A glucose model based on support vector regression for the prediction of hypoglycemic events under free-living conditions

BACKGROUND

The prevention of hypoglycemic events is of paramount importance in the daily management of insulin-treated diabetes. The use of short-term prediction algorithms of the subcutaneous (s.c.) glucose concentration may contribute significantly toward this direction. The literature suggests that, although the recent glucose profile is a prominent predictor of hypoglycemia, the overall patient's context greatly impacts its accurate estimation. The objective of this study is to evaluate the performance of a support vector for regression (SVR) s.c. glucose method on hypoglycemia prediction.

MATERIALS AND METHODS

We extend our SVR model to predict separately the nocturnal events during sleep and the non-nocturnal (i.e., diurnal) ones over 30-min and 60-min horizons using information on recent glucose profile, meals, insulin intake, and physical activities for a hypoglycemic threshold of 70 mg/dL. We also introduce herein additional variables accounting for recurrent nocturnal hypoglycemia due to antecedent hypoglycemia, exercise, and sleep. SVR predictions are compared with those from two other machine learning techniques.

RESULTS

The method is assessed on a dataset of 15 patients with type 1 diabetes under free-living conditions. Nocturnal hypoglycemic events are predicted with 94% sensitivity for both horizons and with time lags of 5.43 min and 4.57 min, respectively. As concerns the diurnal events, when physical activities are not considered, the sensitivity is 92% and 96% for a 30-min and 60-min horizon, respectively, with both time lags being less than 5 min. However, when such information is introduced, the diurnal sensitivity decreases by 8% and 3%, respectively. Both nocturnal and diurnal predictions show a high (>90%) precision.

CONCLUSIONS

Results suggest that hypoglycemia prediction using SVR can be accurate and performs better in most diurnal and nocturnal cases compared with other techniques. It is advised that the problem of hypoglycemia prediction should be handled differently for nocturnal and diurnal periods as regards input variables and interpretation of results.

Hypoglycemia Early Alarm Systems Based On Multivariable Models

Hypoglycemia is a major challenge of artificial pancreas systems and a source of concern for potential users and parents of young children with Type 1 diabetes (T1D). Early alarms to warn the potential of hypoglycemia are essential and should provide enough time to take action to avoid hypoglycemia. Many alarm systems proposed in the literature are based on interpretation of recent trends in glucose values. In the present study, subject-specific recursive linear time series models are introduced as a better alternative to capture glucose variations and predict future blood glucose concentrations. These models are then used in hypoglycemia early alarm systems that notify patients to take action to prevent hypoglycemia before it happens. The models developed and the hypoglycemia alarm system are tested retrospectively using T1D subject data. A Savitzky-Golay filter and a Kalman filter are used to reduce noise in patient data. The hypoglycemia alarm algorithm is developed by using predictions of future glucose concentrations from recursive models. The modeling algorithm enables the dynamic adaptation of models to inter-/intra-subject variation and glycemic disturbances and provides satisfactory glucose concentration prediction with relatively small error. The alarm systems demonstrate good performance in prediction of hypoglycemia and ultimately in prevention of its occurrence.

An early warning system for hypoglycemic/hyperglycemic events based on fusion of adaptive prediction models

INTRODUCTION

Early warning of future hypoglycemic and hyperglycemic events can improve the safety of type 1 diabetes mellitus (T1DM) patients. The aim of this study is to design and evaluate a hypoglycemia/hyperglycemia early warning system (EWS) for T1DM patients under sensor-augmented pump (SAP) therapy.

METHODS

The EWS is based on the combination of data-driven online adaptive prediction models and a warning algorithm. Three modeling approaches have been investigated: (i) autoregressive (ARX) models, (ii) auto-regressive with an output correction module (cARX) models, and (iii) recurrent neural network (RNN) models. The warning algorithm performs postprocessing of the models' outputs and issues alerts if upcoming hypoglycemic/hyperglycemic events are detected. Fusion of the cARX and RNN models, due to their complementary prediction performances, resulted in the hybrid autoregressive with an output correction module/recurrent neural network (cARN)-based EWS.

RESULTS

The EWS was evaluated on 23 T1DM patients under SAP therapy. The ARX-based system achieved hypoglycemic (hyperglycemic) event prediction with median values of accuracy of 100.0% (100.0%), detection time of 10.0 (8.0) min, and daily false alarms of 0.7 (0.5). The respective values for the cARX-based system were 100.0% (100.0%), 17.5 (14.8) min, and 1.5 (1.3) and, for the RNN-based system, were 100.0% (92.0%), 8.4 (7.0) min, and 0.1 (0.2). The hybrid cARN-based EWS presented outperforming results with 100.0% (100.0%) prediction accuracy, detection 16.7 (14.7) min in advance, and 0.8 (0.8) daily false alarms.

CONCLUSION

Combined use of cARX and RNN models for the development of an EWS outperformed the single use of each model, achieving accurate and prompt event prediction with few false alarms, thus providing increased safety and comfort.

Multivariable adaptive closed-loop control of an artificial pancreas without meal and activity announcement

BACKGROUND

Accurate closed-loop control is essential for developing artificial pancreas (AP) systems that adjust insulin infusion rates from insulin pumps. Glucose concentration information from continuous glucose monitoring (CGM) systems is the most important information for the control system. Additional physiological measurements can provide valuable information that can enhance the accuracy of the control system. Proportional-integral-derivative control and model predictive control have been popular in AP development. Their implementations to date rely on meal announcements (e.g., bolus insulin dose based on insulin:carbohydrate ratios) by the user. Adaptive control techniques provide a powerful alternative that do not necessitate any meal or activity announcements.

MATERIALS AND METHODS

Adaptive control systems based on the generalized predictive control framework are developed by extending the recursive modeling techniques. Physiological signals such as energy expenditure and galvanic skin response are used along with glucose measurements to generate a multiple-input-single-output model for predicting future glucose concentrations used by the controller. Insulin-on-board (IOB) is also estimated and used in control decisions. The controllers were tested with clinical studies that include seven cases with three different patients with type 1 diabetes for 32 or 60 h without any meal or activity announcements.

RESULTS

The adaptive control system kept glucose concentration in the normal preprandial and postprandial range (70-180 mg/dL) without any meal or activity announcements during the test period. After IOB estimation was added to the control system, mild hypoglycemic episodes were observed only in one of the four experiments. This was reflected in a plasma glucose value of 56 mg/dL (YSI 2300 STAT; Yellow Springs Instrument, Yellow Springs, OH) and a CGM value of 63 mg/dL).

CONCLUSIONS

Regulation of blood glucose concentration with an AP using adaptive control techniques was successful in clinical studies, even without any meal and physical activity announcement.

Reduction of number and duration of hypoglycemic events by glucose prediction methods: a proof-of-concept in silico study

BACKGROUND

Hypoglycemia prevention is one of the major challenges in diabetes research. Recently, it has been suggested that continuous glucose monitoring (CGM)-based short-term glucose prediction algorithms could be exploited to generate alerts when hypoglycemia is forecasted, allowing the patient to take appropriate countermeasures to avoid/mitigate the event. However, quantifying the potential benefits of prediction in terms of reduction of number/duration of hypoglycemia requires an in silico assessment that is the object of the present article.

MATERIALS AND METHODS

Data for 50 virtual subjects were generated by using the University of Virginia/Padova type 1 diabetes simulator (54-h monitoring), made more credible by adding realistic measurement noise and perturbations of meals and insulin injections. CGM was assumed to be well calibrated. Occurrence and duration of hypoglycemic events were compared in three scenarios: (1) hypoglycemia was not recognized and not dealt with; (2) 15 g of carbohydrates was ingested when CGM crossed the hypoglycemia threshold; or (3) 15 g of carbohydrates was ingested when the 30-min ahead-of-time CGM prediction crossed the hypoglycemia threshold. The effectiveness of alerts was investigated also in the case of delayed/absent ingestion of carbohydrates.

RESULTS

In Scenario 1, each virtual subject spent 17.7% of the time in the hypoglycemic range, with a median of four events of 120 min in the 54-h period monitored. In Scenario 2, the time spent in hypoglycemia was reduced to 4.7% (four events of 40 min). In Scenario 3, the time spent in hypoglycemia was further reduced to 1.2% (one event of 15 min). Absent/delayed patient's responses to alerts slightly increase these percentages, but improvements remain significant.

CONCLUSIONS

This in silico proof-of-concept study demonstrates that using predicted rather than measured CGM allows a significant reduction of the number of hypoglycemic events and the time spent in hypoglycemic range both by 75%, stimulating further research and clinical investigation on the generation of preventive hypoglycemic alerts exploiting glucose prediction methods.