Blood Glucose Prediction Research Articles

Non-invasive prediction of blood glucose trends during hypoglycemia

Over the last four decades, there has been a pursuit for a non-invasive solution for glucose measurement, but there is not yet any viable product released. Of the many sensor modalities tried, the combination of electrical and optical measurement is among the most promising for continuous measurements. Although non-invasive prediction of exact glucose levels may seem futile, prediction of their trends may be useful for certain applications. Hypoglycemia is the most serious of the acute complications in type-1 diabetes highlighting the need for a reliable alarm, but little is known about the performance of this technology in predicting hypoglycemic glucose levels and associated trends. We aimed to assess such performance on the way to develop a multisensor system for detection of hypoglycemia, based on near-infrared (NIR), bioimpedance and skin temperature measurements taken during hypoglycemic and euglycemic glucose clamps in 20 subjects with type-1 diabetes. Performance of blood glucose prediction was assessed by global partial least squares and neural network regression models using repeated double cross-validation. Best trend prediction was obtained by including all measurements in a neural network model. Prediction of glucose level was inaccurate for threshold-based detection of hypoglycemia, but the trend predictions may provide useful information in a multisensor system. Comparing NIR and bioimpedance measurements, NIR seems to be the main predictor of blood glucose while bioimpedance may act as correction for individual confounding properties.

Unsupervised calibration for noninvasive glucose-monitoring devices using mid-infrared spectroscopy

Noninvasive, glucose-monitoring technologies using infrared spectroscopy that have been studied typically require a calibration process that involves blood collection, which renders the methods somewhat invasive. We develop a truly noninvasive, glucose-monitoring technique using mid-infrared spectroscopy that does not require blood collection for calibration by applying domain adaptation (DA) using deep neural networks to train a model that associates blood glucose concentration with mid-infrared spectral data without requiring a training dataset labeled with invasive blood sample measurements. For realizing DA, the distribution of unlabeled spectral data for calibration is considered through adversarial update during training networks for regression to blood glucose concentration. This calibration improved the correlation coefficient between the true blood glucose concentrations and predicted blood glucose concentrations from 0.38 to 0.47. The result indicates that this calibration technique improves prediction accuracy for mid-infrared glucose measurements without any invasively acquired data.

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.

Systematic map and review of predictive techniques in diabetes self-management

Data mining (DM) provides powerful tools to extract knowledge from large volumes of data offering valuable information to decision making. The extracted knowledge can be used for predictive and/or descriptive purposes. DM has been successfully used in different subfields of eHealth such us cardiology, oncology and endocrinology. This paper deals with the use of DM predictive techniques in diabetes self-management (DSM).To the best of our knowledge, neither a systematic map nor a systematic review have yet been performed with a focus on the use of DM predictive techniques in DSM. Thus, the aim of this study is to classify and review primary studies investigating DM predictive techniques in DSM by summarizing and analyzing knowledge with respect to: year and source of publication, type of diabetes, clinical tasks, DM predictive techniques, and the performance of the predictive techniques used. A total of 38 papers published between 2000 and April 2017 were therefore selected and analyzed accordingly to address six research questions. The results show that Type 1 Diabetes Mellitus (T1DM) is largely the type of diabetes that is most concerned by the studies and the prediction of blood glucose is the most investigated clinical task. Moreover, artificial neural networks were the most frequently used predictive technique which along with autogressive models, yield highest accuracy rates.

The fasting blood glucose and long non-coding RNA SNHG8 predict poor prognosis in patients with gastric carcinoma after radical gastrectomy.

This prospective study sought to evaluate the prediction of fasting blood glucose and long non-coding RNA (lncRNA) SNHG8 for the risk of gastric carcinoma mortality. A total of 217 gastric carcinoma patients underwent radical gastrectomy were included during 2012-16. The final follow-up was finished in January 2017. The aggregate hazard ratio(HR) demonstrated that poor prognosis of gastric carcinoma was associated with fasting blood glucose (HR= 1.29, P=0.037), SNHG8 expression(HR = 1.10, P= 0.009), positive distant metastasis(HR = 2.99, P= 0.020), EBV positive (HR = 3.40, P=0.002), and tumor size more than 5.0 cm (HR = 3.36, P= 0.005). In survival analysis, elevated fasting blood glucose (P =0.007) and high SNHG8 expression (P =0.007) were significantly associated with shorter survival times in gastric cancer. Significant multiplicative interaction was shown between fasting blood glucose and SNHG8 expression (chi-squared=7.81, Pmultiplicative =0.005), without statistical additive interaction. Fasting blood glucose and SNHG8 expression could predict poor prognosis after radical gastrectomy. LncRNA SNHG8 could be applied as a novel epigenetic molecular target in gastric carcinoma.

A Personalized Healthcare Monitoring System for Diabetic Patients by Utilizing BLE-Based Sensors and Real-Time Data Processing.

Current technology provides an efficient way of monitoring the personal health of individuals. Bluetooth Low Energy (BLE)-based sensors can be considered as a solution for monitoring personal vital signs data. In this study, we propose a personalized healthcare monitoring system by utilizing a BLE-based sensor device, real-time data processing, and machine learning-based algorithms to help diabetic patients to better self-manage their chronic condition. BLEs were used to gather users’ vital signs data such as blood pressure, heart rate, weight, and blood glucose (BG) from sensor nodes to smartphones, while real-time data processing was utilized to manage the large amount of continuously generated sensor data. The proposed real-time data processing utilized Apache Kafka as a streaming platform and MongoDB to store the sensor data from the patient. The results show that commercial versions of the BLE-based sensors and the proposed real-time data processing are sufficiently efficient to monitor the vital signs data of diabetic patients. Furthermore, machine learning–based classification methods were tested on a diabetes dataset and showed that a Multilayer Perceptron can provide early prediction of diabetes given the user’s sensor data as input. The results also reveal that Long Short-Term Memory can accurately predict the future BG level based on the current sensor data. In addition, the proposed diabetes classification and BG prediction could be combined with personalized diet and physical activity suggestions in order to improve the health quality of patients and to avoid critical conditions in the future.

Designing an Internet-of-Things (IoT) and sensor-based in-home monitoring system for assisting diabetes patients: iterative learning from two case studies

ABSTRACTThe ageing of the global population is creating a crisis in chronic disease management. In the USA, 29 million people (or 9.3% of the population) suffer from the chronic disease of diabetes; according to the WHO, globally around 200 million people are diabetic. Left unchecked, diabetes can lead to acute and long-term complications and ultimately death. Diabetes prevalence tends to be the highest among those aged 65 and older (nearly 20.6%), a population which often lacks the cognitive resources to deal with the daily self-management regimens. In this paper, we discuss the design and implementation of an Internet-of-Things (IoT) and wireless sensor system which patients use in their own homes to capture daily activity, an important component in diabetes management. Following Fogg’s 2009 persuasion theory, we mine the activity data and provide motivational messages to the subjects with the intention of changing their activity and dietary behaviour. We introduce a novel idea called “persuasive sensing” and report results from two home implementations that show exciting promise. With the captured home monitoring data, we also develop analytic models that can predict blood glucose levels for the next day with an accuracy of 94%. We conclude with lessons learned from these two home case studies and explore design principles for creating novel IoT systems.

Near-Infrared Spectroscopic Method for Noninvasive Blood Glucose Prediction without Using Multivariate Analyses

A near-infrared spectroscopy method for noninvasive blood glucose prediction without using multivariate analyses is proposed and examined. Near-infrared diffuse reflection spectra with a wavelength range from 1400 to 1850 nm were measured at the skin tissue using an optical fiber probe. The difference spectra obtained by subtracting the initially measured spectrum from the other measured spectra were reconstructed from the spectra of four major components in the human skin (water, protein, glucose, and fat) and the spectrum of a scattering equivalent component called baseline, based on the modified Beer’s law. Then the blood glucose contents were predicted from the content index of the glucose component appearing in calculation of the reconstructed spectrum. We assumed that one of the origins of the errors in the blood glucose prediction using near-infrared spectroscopy was the similarity of the shapes of the absorption spectrum between glucose and baseline. To distinguish the glucose contribution from those of baseline and fat in the difference spectra, an imaginary component combining baseline and fat was introduced by considering that both the change in the fat contribution and the generation of baseline originated from the change in scattering in the skin. The imaginary component enabled us to reduce the errors in the blood glucose prediction. In contrast to the conventional methods using multivariate analyses, the calculation process of the proposed method is not so called "a black box." Thus, it is easy to estimate the origins of the changes and contributions of the components in the difference spectra. The proposed method may become a useful tool for realization of noninvasive blood glucose prediction. Disclosure K. Maruo: None. Y. Yamada: None.

A Machine-Learning Model Accurately Predicts Projected Blood Glucose

Background: Clinical guidelines do not specify the frequency of self-monitoring of blood glucose (SMBG) for the ∼85% of people with type 2 diabetes (T2D) not on intensive insulin therapy. For these individuals, insurance covers a limited amount of testing supplies, and blood glucose (BG) checks are infrequent. With limited BG data, people may not know how foods, activity, and other factors affect their BG levels. Other, non-SMBG-dependent, approaches are needed to illustrate how self-care choices affect BG. Objectives: 1.) Use a large dataset to train a machine learning model 2.) Use the model and minimal data inputs to predict BG values at variable times 3.) Test the accuracy of the prediction. Methods: The One Drop | Mobile app collected 1,923,416 BG measurements from 14,706 people with noninsulin treated T2D. Contextual information (CI), when available, included demographics, health metrics (e.g., weight, A1c), and self-care. Inputs to each BG prediction included a prior BG and available CI. The model did not distinguish whether BGs with similar CI were from the same or different users. Forecast horizons were set by the time since prior BG and varied from 10 minutes to several days. Machine learning algorithms to predict BG values were trained and vetted on BGs entered prior to Sept. 2017 (83% of all BGs). BGs (17%) entered from Sept.-Nov. 2017 were held out and predicted. Results: Users were 59% male, with 80% from North America, 9% from Europe, and 11% from elsewhere; 50% were diagnosed with T2D in the past 3 years. The median and mean absolute error of holdout predictions were 14.2 and 21.3 mg/dL respectively, with 91% of predictions within +/-50 mg/dL. Discussion: A machine learning model with minimal inputs accurately predicts future BG values in noninsulin treated people with T2D. Whether people modify their behavior (e.g., eating less carbohydrates) after knowing a planned behavior (e.g., a 60 carbohydrate gram meal) produces an undesirable BG (&amp;gt;140 mg/dL postprandial) is an empirical question worth investigating. Disclosure D.R. Goldner: Consultant; Self; One Drop. C.Y. Osborn: Employee; Self; One Drop. Research Support; Self; MannKind Corporation. Stock/Shareholder; Self; One Drop. L.E. Sears: Employee; Self; Informed Data Systems Inc.. Research Support; Self; MannKind Corporation. B. Huddleston: Employee; Self; One Drop. J. Dachis: Stock/Shareholder; Self; One Drop. Board Member; Self; One Drop. Employee; Self; One Drop.

Predicting Future Glucose Fluctuations Using Machine Learning and Wearable Sensor Data

Predicting blood glucose values (BG) using machine learning (ML) algorithms and data fusion techniques. There has been a recent explosion of interest in BG prediction due to its application in the development of insulin regulating algorithms for the Artificial Pancreas Project. During this study, we measured the predictive accuracy of a software system designed to predict glucose behaviour using step-count and heart-rate data in addition to BG-insulin dynamics. The software was tested in a blinded pilot study at BC Children's Hospital for 9 type 1 diabetic children. Using continuous glucose monitors (CGM) and fitness wearables (Fitbit), the software aggregated 60-days of continuous data from each participant. The data from the first 30-days of the study was used to train the algorithem. The trained algorithm was then used to make predictions every 5mins for the next 30days. On average, the software was able to predict user's future glucose values with 93% accuracy rate for 60-mins ahead of time. Although encouraging, the algorithm required further testing. We have since released the app under the commercial name "DiaBits" for more testing and further data collection. Disclosure A. Hayeri: Other Relationship; Self; Dexcom, Inc., Fitbit, Inc.. Research Support; Self; BC Children's Hospital - Vancouver Canada. Other Relationship; Self; University Of British Columbia.

In vivo noninvasive blood glucose detection using near-infrared spectrum based on the PSO-2ANN model.

BACKGROUND: To improving the nursing level of diabetics, it is necessary to develop noninvasive blood glucose method.OBJECTIVE: In order to reduce the number of the near-infrared signal, consider the nonlinear relationship between the blood glucose concentration and near-infrared signal, and correct the individual difference and physiological glucose dynamic, 2 artificial neural networks (2ANN) combined with particle swarm optimization (PSO), named as PSO-2ANN, is proposed.METHOD: Two artificial neural networks (ANNs) are employed as the basic structure of the PSO-ANN model, and the weight coefficients of the two ANNs which represent the difference of individual and daily physiological rule are optimized by particle swarm optimization (PSO).RESULTS: Clarke error grid shows the blood glucose predictions are distributed in regions A and B, Bland-Altman analysis show that the predictions and measurements are in good agreement.CONCLUSIONS: The PSO-2ANN model is a nonlinear calibration strategy with accuracy and robustness using 1550-nm spectroscopy, which can correct the individual difference and physiological glucose dynamics.

Is type 1 diabetes a chaotic phenomenon?

A database of ten type 1 diabetes patients wearing a continuous glucose monitoring device has enabled to record their blood glucose continuous variations every minute all day long during fourteen consecutive days. These recordings represent, for each patient, a time series consisting of 1 value of glycaemia per minute during 24 h and 14 days, i.e., 20,160 data points. Thus, while using numerical methods, these time series have been anonymously analyzed. Nevertheless, because of the stochastic inputs induced by daily activities of any human being, it has not been possible to discriminate chaos from noise. So, we have decided to keep only the 14 nights of these ten patients. Then, the determination of the time delay and embedding dimension according to the delay coordinate embedding method has allowed us to estimate for each patient the correlation dimension and the maximal Lyapunov exponent. This has led us to show that type 1 diabetes could indeed be a chaotic phenomenon. Once this result has been confirmed by the determinism test, we have computed the Lyapunov time and found that the limit of predictability of this phenomenon is nearly equal to half the 90 min sleep-dream cycle. We hope that our results will prove to be useful to characterize and predict blood glucose variations.

A Test Bench for Performance Validation of an Artificial Pancreas System

A test bench for performance validation of an artificial pancreas system is described. The test bench allows simulating an increase or a decrease in blood glucose by automated mixing of glucose-containing test solution. It also provides validation of the blood glucose prediction algorithm. The test bench consists of an electrical subsystem and a hydraulic subsystem that are controlled from PC user interface. It provides blood glucose measurement and prediction with an error of <5% and the accuracy of insulin dosing of 0.05 μL.

Development and Evaluation of a Mobile Personalized Blood Glucose Prediction System for Patients With Gestational Diabetes Mellitus.

BackgroundPersonalized blood glucose (BG) prediction for diabetes patients is an important goal that is pursued by many researchers worldwide. Despite many proposals, only a few projects are dedicated to the development of complete recommender system infrastructures that incorporate BG prediction algorithms for diabetes patients. The development and implementation of such a system aided by mobile technology is of particular interest to patients with gestational diabetes mellitus (GDM), especially considering the significant importance of quickly achieving adequate BG control for these patients in a short period (ie, during pregnancy) and a typically higher acceptance rate for mobile health (mHealth) solutions for short- to midterm usage.ObjectiveThis study was conducted with the objective of developing infrastructure comprising data processing algorithms, BG prediction models, and an appropriate mobile app for patients’ electronic record management to guide BG prediction-based personalized recommendations for patients with GDM.MethodsA mobile app for electronic diary management was developed along with data exchange and continuous BG signal processing software. Both components were coupled to obtain the necessary data for use in the personalized BG prediction system. Necessary data on meals, BG measurements, and other events were collected via the implemented mobile app and continuous glucose monitoring (CGM) system processing software. These data were used to tune and evaluate the BG prediction model, which included an algorithm for dynamic coefficients tuning. In the clinical study, 62 participants (GDM: n=49; control: n=13) took part in a 1-week monitoring trial during which they used the mobile app to track their meals and self-measurements of BG and CGM system for continuous BG monitoring. The data on 909 food intakes and corresponding postprandial BG curves as well as the set of patients’ characteristics (eg, glycated hemoglobin, body mass index [BMI], age, and lifestyle parameters) were selected as inputs for the BG prediction models.ResultsThe prediction results by the models for BG levels 1 hour after food intake were root mean square error=0.87 mmol/L, mean absolute error=0.69 mmol/L, and mean absolute percentage error=12.8%, which correspond to an adequate prediction accuracy for BG control decisions.ConclusionsThe mobile app for the collection and processing of relevant data, appropriate software for CGM system signals processing, and BG prediction models were developed for a recommender system. The developed system may help improve BG control in patients with GDM; this will be the subject of evaluation in a subsequent study.

Accurate prediction of continuous blood glucose based on support vector regression and differential evolution algorithm

Type 1 diabetes (T1D) is a chronic disease requiring patients to know their blood glucose values in order to ensure blood glucose levels as close to normal as possible. Hence, the ability to predict blood glucose levels is of a great interest for clinical researchers. In this sense, the literature is rich with several solutions that can predict blood glucose levels. Unfortunately, these methods require the patient to specific their daily activities: meal intake, insulin injection and emotional factors, which can be error prone. To reduce this burden on the patent, this work proposes to use only continuous glucose monitoring (CGM) data to predict blood glucose levels independently of other factors. To support this, support vector regression (SVR) and differential evolution (DE) algorithms were investigated. The proposed method is validated using real CGM data of 12 patients. The obtained average of root mean square error (RMSE) was 9.44, 10.78, 11.82 and 12.95mg/dL for prediction horizon (PH) respectively equal to 15, 30, 45 and 60min. The results of the present study and comparison with some previous works show that the proposed method holds promise. The SVR based on DE algorithm achieved high prediction accuracy while being robustness, automatic, and requiring no human intervention.

Getting Daily Blood Glucose Tracks Using Clinical Protocols of the DirecNet Database

This article describes a method for forming daily blood glucose (BG) tracks for diabetes mellitus patients using protocols taken from the DirecNet database. The obtained daily tracks can be used for testing mathematical models of BG regulation, verifying BG prediction algorithms, and analyzing individual trends in patient’s BG dynamics.

A Model for Diabetic Blood Glucose Prediction Based on Electroencephalography Signals Using Deep Learning

A Model for Diabetic Blood Glucose Prediction Based on Electroencephalography Signals Using Deep Learning

Blood glucose prediction of type I diabetic patient based on linear model with lag weighted regularization

Blood glucose prediction of type I diabetic patient based on linear model with lag weighted regularization

Effect of physiological parameters on glucose microcirculation compartmental model in glucose monitoring

ABSTRACTA glucose microcirculation compartmental model (GMCM) for predicting blood glucose concentrations from the reference interstitial fluid (ISF) glucose concentrations was proposed before. It was used to compensate the delay error in ISF in order to improve the prediction accuracy and to facilitate the clinical application of continuous blood glucose monitoring. In this paper, many physiological parameters in GMCM reflecting delay and transport properties were discussed in details. The significance and influence of the parameters, diffusion distance (x), pressure gradient-driven transport ratio (RStarling), blood flow velocity (F), glucose capillary permeability (P), glucose diffusion coefficient (D), on the model were simulated and analysed. One group of blood (ISF) glucose values was set as input for the model with one varied parameter and other fixed parameters to predict ISF (blood) glucose concentrations. The results verified that F and P have much influence on the model and others play more minor roles. And some have positive influence on the model, like P, while others have a reverse effect, like F. All parameters can be modulated for different species or different physiological status during the model application.

Personalized blood glucose prediction: A hybrid approach using grammatical evolution and physiological models.

The large patient variability in human physiology and the effects of variables such as exercise or meals challenge current prediction modeling techniques. Physiological models are very precise but they are typically complex and specific physiological knowledge is required. In contrast, data-based models allow the incorporation of additional inputs and accurately capture the relationship between these inputs and the outcome, but at the cost of losing the physiological meaning of the model. In this work, we designed a hybrid approach comprising physiological models for insulin and grammatical evolution, taking into account the clinical harm caused by deviations from the target blood glucose by using a penalizing fitness function based on the Clarke error grid. The prediction models were built using data obtained over 14 days for 100 virtual patients generated by the UVA/Padova T1D simulator. Midterm blood glucose was predicted for the 100 virtual patients using personalized models and different scenarios. The results obtained were promising; an average of 98.31% of the predictions fell in zones A and B of the Clarke error grid. Midterm predictions using personalized models are feasible when the configuration of grammatical evolution explored in this study is used. The study of new alternative models is important to move forward in the development of alarm-and-control applications for the management of type 1 diabetes and the customization of the patient’s treatments. The hybrid approach can be adapted to predict short-term blood glucose values to detect continuous glucose-monitoring sensor errors and to estimate blood glucose values when the continuous glucose-monitoring system fails to provide them.