Optimising Personalised Medical Insights by Introducing a Scalable Health Informatics Application for Sensor Data Extraction, Preprocessing, and Analysis.

Wearable sensors, among other informatics solutions, are readily accessible to enable noninvasive remote monitoring in healthcare. While providing a wealth of data, the wide variety of such sensing systems and the differing implementations of the same or similar sensors by different developers complicate comparisons of collected data. An online application as a platform technology that provides uniform methods for analysing balance data is presented as a case study. The development of balance problems is common in neurodegenerative conditions, leading to falls and a reduced quality of life. While balance can be assessed using, for example, perturbation tests, sensors offer a more quantitative and scalable way. Researchers can adjust the platform to integrate the sensors of their choice or upload data and then preprocess, featurise, analyse, and visualise them. This eases performing comparative analyses across the sensors and datasets through a reduction of heterogeneity and facilitates easy integration of machine learning and other advanced data analytics, thereby targeting personalising medical insights.

The use of event-related potentials in the investigation of cognitive performance in people with Multiple Sclerosis: Systematic review.

A biomarker of cognition in Multiple Sclerosis (MS) that is independent from the response of people with MS (PwMS) to test questions would provide a more holistic assessment of cognitive decline. One suggested method involves event-related potentials (ERPs). This systematic review tried to answer five questions about the use of ERPs in distinguishing PwMS from controls: which stimulus modality, which experimental paradigm, which electrodes, and which ERP components are most discriminatory, and whether amplitude or latency is a better measure. Our results show larger pooled effect sizes for visual stimuli than auditory stimuli, and larger pooled effect sizes for latency measurements than amplitude measurements. We observed great heterogeneity in methods and suggest that future research would benefit from more uniformity in methods and that results should be reported for the individual subtypes of PwMS. With more standardised methods, ERPs have the potential to be developed into a clinical tool in MS.

Evaluating Effects of Resting-State Electroencephalography Data Pre-Processing on a Machine Learning Task for Parkinson's Disease.

Resting-state electroencephalography pre-processing methods in machine learning studies into Parkinson's disease classification vary widely. Here three separate data sets were pre-processed to four different stages to investigate the effects on evaluation metrics, using power features from six regions-of-interest, Random Forest Classifiers for feature selection, and Support Vector Machines for classification. This showed muscle artefact inflated evaluation metrics, and alpha and theta band features produced the best results when fully pre-processing data.

Evaluating methods of oversampling and averaging resting-state electroencephalography data in classifying Parkinson's disease.

Collecting resting-state electroencephalography (RSEEG) data is time-consuming and data sets are therefore often small. Because many machine learning (ML) algorithms work better with ample data, researchers looking to use RSEEG and ML to develop diagnostic models have used oversampling methods that may seem to contradict averaging methods used in conventional electroencephalography (EEG) research to improve the signal-to-noise ratio. Using eyes open (EO) and eyes closed (EC) recordings from 3 different research groups, we investigated the effect of different averaging and oversampling methods on classification metrics when classifying people with Parkinson's disease (PD) and controls. Both EC and EO recordings were used due to differences found between these methods. Our results indicated that grouping 58 electrodes into regions-of-interest (ROI) based on anatomical location is preferable to using single electrodes. Furthermore, although recording EO data led to slightly better classification, the number of data points for each participant was reduced and recordings for three participants entirely lost during pre-processing due to a higher level of artefacts than in the EC data.Clinical relevance- RSEEG is a potential biomarker for the diagnosis and prognostication of PD, but for RSEEG to have clinical relevance, it is necessary to establish which averaging and oversampling of data most reliably segregates the classes for people with PD and controls. We found that using of ROIs and EC data performed the best, as EO data was often contaminated with artefacts.

Jointly Predicting Postprandial Hypoglycemia and Hyperglycemia Using Continuous Glucose Monitoring Data in Type 1 Diabetes.

The development of continuous glucose monitoring (CGM) systems has enabled people with type 1 diabetes mellitus (T1DM) to track their glucose trajectory in real-time and inspired research in personalised glucose prediction. In this paper, our aim is to predict postprandial abnormal-glycemia events. Different from prior research which focuses on hypoglycemia only, we make the first attempt to establish our problem as the joint prediction of hyperglycemia and hypoglycemia. On this basis, we propose a machine learning model that learns from the pattern of 1 hour past glucose and makes predictions for the two tasks simultaneously using a unified backbone. Key benefits of our methodology include 1) requiring only the CGM sequence as the input, thus making it more widely applicable than other counterparts using extra inputs such as the nutrition details, and 2) minimising the computational cost as the two tasks are unified into a single model. Our experiments on the openly available OhioT1DM dataset achieve state-of-the-art performance (Matthew's correlation coefficient of 0.61 for hyperglycemia and 0.48 for hypoglycemia). To encourage further study, we release our codes at https://github.com/r-cui/PostprandialHyperHypoPrediction under the MIT license.

High Accuracy Protein Identification: Fusion of Solid-State Nanopore Sensing and Machine Learning.

Proteins are arguably one of the most important class of biomarkers for health diagnostic purposes. Label-free solid-state nanopore sensing is a versatile technique for sensing and analyzing biomolecules such as proteins at single-molecule level. While molecular-level information on size, shape, and charge of proteins can be assessed by nanopores, the identification of proteins with comparable sizes remains a challenge. Here, solid-state nanopore sensing is combined with machine learning to address this challenge. The translocations of four similarly sized proteins is assessed using amplifiers with bandwidths (BWs) of 100 kHz and 10 MHz, the highest bandwidth reported for protein sensing, using nanopores fabricated in <10 nm thick silicon nitride membranes. F-values of up to 65.9% and 83.2% (without clustering of the protein signals) are achieved with 100 kHz and 10 MHz BW measurements, respectively, for identification of the four proteins. The accuracy of protein identification is further enhanced by classifying the signals into different clusters based on signal attributes, with F-value and specificity of up to 88.7% and 96.4%, respectively, for combinations of four proteins. The combined use of high bandwidth instruments, advanced clustering and machine learning methods allows label-free identification of proteins with high accuracy.

The role of machine learning in developing non-magnetic resonance imaging based biomarkers for multiple sclerosis: a systematic review.

BACKGROUND: Multiple sclerosis (MS) is a neurological condition whose symptoms, severity, and progression over time vary enormously among individuals. Ideally, each person living with MS should be provided with an accurate prognosis at the time of diagnosis, precision in initial and subsequent treatment decisions, and improved timeliness in detecting the need to reassess treatment regimens. To manage these three components, discovering an accurate, objective measure of overall disease severity is essential. Machine learning (ML) algorithms can contribute to finding such a clinically useful biomarker of MS through their ability to search and analyze datasets about potential biomarkers at scale. Our aim was to conduct a systematic review to determine how, and in what way, ML has been applied to the study of MS biomarkers on data from sources other than magnetic resonance imaging. METHODS: Systematic searches through eight databases were conducted for literature published in 2014-2020 on MS and specified ML algorithms. RESULTS: Of the 1, 052 returned papers, 66 met the inclusion criteria. All included papers addressed developing classifiers for MS identification or measuring its progression, typically, using hold-out evaluation on subsets of fewer than 200 participants with MS. These classifiers focused on biomarkers of MS, ranging from those derived from omics and phenotypical data (34.5% clinical, 33.3% biological, 23.0% physiological, and 9.2% drug response). Algorithmic choices were dependent on both the amount of data available for supervised ML (91.5%; 49.2% classification and 42.3% regression) and the requirement to be able to justify the resulting decision-making principles in healthcare settings. Therefore, algorithms based on decision trees and support vector machines were commonly used, and the maximum average performance of 89.9% AUC was found in random forests comparing with other ML algorithms. CONCLUSIONS: ML is applicable to determining how candidate biomarkers perform in the assessment of disease severity. However, applying ML research to develop decision aids to help clinicians optimize treatment strategies and analyze treatment responses in individual patients calls for creating appropriate data resources and shared experimental protocols. They should target proceeding from segregated classification of signals or natural language to both holistic analyses across data modalities and clinically-meaningful differentiation of disease.

The Potential of Current Noninvasive Wearable Technology for the Monitoring of Physiological Signals in the Management of Type 1 Diabetes: Literature Survey.

BACKGROUND: Monitoring glucose and other parameters in persons with type 1 diabetes (T1D) can enhance acute glycemic management and the diagnosis of long-term complications of the disease. For most persons living with T1D, the determination of insulin delivery is based on a single measured parameter-glucose. To date, wearable sensors exist that enable the seamless, noninvasive, and low-cost monitoring of multiple physiological parameters. OBJECTIVE: The objective of this literature survey is to explore whether some of the physiological parameters that can be monitored with noninvasive, wearable sensors may be used to enhance T1D management. METHODS: A list of physiological parameters, which can be monitored by using wearable sensors available in 2020, was compiled by a thorough review of the devices available in the market. A literature survey was performed using search terms related to T1D combined with the identified physiological parameters. The selected publications were restricted to human studies, which had at least their abstracts available. The PubMed and Scopus databases were interrogated. In total, 77 articles were retained and analyzed based on the following two axes: the reported relations between these parameters and T1D, which were found by comparing persons with T1D and healthy control participants, and the potential areas for T1D enhancement via the further analysis of the found relationships in studies working within T1D cohorts. RESULTS: On the basis of our search methodology, 626 articles were returned, and after applying our exclusion criteria, 77 (12.3%) articles were retained. Physiological parameters with potential for monitoring by using noninvasive wearable devices in persons with T1D included those related to cardiac autonomic function, cardiorespiratory control balance and fitness, sudomotor function, and skin temperature. Cardiac autonomic function measures, particularly the indices of heart rate and heart rate variability, have been shown to be valuable in diagnosing and monitoring cardiac autonomic neuropathy and, potentially, predicting and detecting hypoglycemia. All identified physiological parameters were shown to be associated with some aspects of diabetes complications, such as retinopathy, neuropathy, and nephropathy, as well as macrovascular disease, with capacity for early risk prediction. However, although they can be monitored by available wearable sensors, most studies have yet to adopt them, as opposed to using more conventional devices. CONCLUSIONS: Wearable sensors have the potential to augment T1D sensing with additional, informative biomarkers, which can be monitored noninvasively, seamlessly, and continuously. However, significant challenges associated with measurement accuracy, removal of noise and motion artifacts, and smart decision-making exist. Consequently, research should focus on harvesting the information hidden in the complex data generated by wearable sensors and on developing models and smart decision strategies to optimize the incorporation of these novel inputs into T1D interventions.

Integrating Multiple Inputs Into an Artificial Pancreas System: Narrative Literature Review.

BACKGROUND: Type 1 diabetes (T1D) is a chronic autoimmune disease in which a deficiency in insulin production impairs the glucose homeostasis of the body. Continuous subcutaneous infusion of insulin is a commonly used treatment method. Artificial pancreas systems (APS) use continuous glucose level monitoring and continuous subcutaneous infusion of insulin in a closed-loop mode incorporating a controller (or control algorithm). However, the operation of APS is challenging because of complexities arising during meals, exercise, stress, sleep, illnesses, glucose sensing and insulin action delays, and the cognitive burden. To overcome these challenges, options to augment APS through integration of additional inputs, creating multi-input APS (MAPS), are being investigated. OBJECTIVE: The aim of this survey is to identify and analyze input data, control architectures, and validation methods of MAPS to better understand the complexities and current state of such systems. This is expected to be valuable in developing improved systems to enhance the quality of life of people with T1D. METHODS: A literature survey was conducted using the Scopus, PubMed, and IEEE Xplore databases for the period January 1, 2005, to February 10, 2020. On the basis of the search criteria, 1092 articles were initially shortlisted, of which 11 (1.01%) were selected for an in-depth narrative analysis. In addition, 6 clinical studies associated with the selected studies were also analyzed. RESULTS: Signals such as heart rate, accelerometer readings, energy expenditure, and galvanic skin response captured by wearable devices were the most frequently used additional inputs. The use of invasive (blood or other body fluid analytes) inputs such as lactate and adrenaline were also simulated. These inputs were incorporated to switch the mode of the controller through activity detection, directly incorporated for decision-making and for the development of intermediate modules for the controller. The validation of the MAPS was carried out through the use of simulators based on different physiological models and clinical trials. CONCLUSIONS: The integration of additional physiological signals with continuous glucose level monitoring has the potential to optimize glucose control in people with T1D through addressing the identified limitations of APS. Most of the identified additional inputs are related to wearable devices. The rapid growth in wearable technologies can be seen as a key motivator regarding MAPS. However, it is important to further evaluate the practical complexities and psychosocial aspects associated with such systems in real life.

A Significance Assessment of Diabetes Diagnostic Biomarkers Using Machine Learning.

Diabetes can be diagnosed by either Fasting Plasma Glucose or Hemoglobin A1c. The aim of our study was to explore the differences between the two criteria through the development of a machine learning based diabetes diagnostic algorithm and analysing the predictive contribution of each input biomarker. Our study concludes that fasting insulin is predictive of diabetes defined by FPG, but not by HbA1c. Besides, 28 other fasting blood biomarkers were not significant predictors of diabetes.

The Use of Event-Related Potentials and Machine Learning to Improve Diagnostic Testing and Prediction of Disease Progression in Parkinson's Disease.

Current tests of disease status in Parkinson's disease suffer from high variability, limiting their ability to determine disease severity and prognosis. Event-related potentials, in conjunction with machine learning, may provide a more objective assessment. In this study, we will use event-related potentials to develop machine learning models, aiming to provide an objective way to assess disease status and predict disease progression in Parkinson's disease.

Model-Free Machine Learning in Biomedicine: Feasibility Study in Type 1 Diabetes.

Although reinforcement learning (RL) is suitable for highly uncertain systems, the applicability of this class of algorithms to medical treatment may be limited by the patient variability which dictates individualised tuning for their usually multiple algorithmic parameters. This study explores the feasibility of RL in the framework of artificial pancreas development for type 1 diabetes (T1D). In this approach, an Actor-Critic (AC) learning algorithm is designed and developed for the optimisation of insulin infusion for personalised glucose regulation. AC optimises the daily basal insulin rate and insulin:carbohydrate ratio for each patient, on the basis of his/her measured glucose profile. Automatic, personalised tuning of AC is based on the estimation of information transfer (IT) from insulin to glucose signals. Insulin-to-glucose IT is linked to patient-specific characteristics related to total daily insulin needs and insulin sensitivity (SI). The AC algorithm is evaluated using an FDA-accepted T1D simulator on a large patient database under a complex meal protocol, meal uncertainty and diurnal SI variation. The results showed that 95.66% of time was spent in normoglycaemia in the presence of meal uncertainty and 93.02% when meal uncertainty and SI variation were simultaneously considered. The time spent in hypoglycaemia was 0.27% in both cases. The novel tuning method reduced the risk of severe hypoglycaemia, especially in patients with low SI.

GoCARB in the Context of an Artificial Pancreas.

BACKGROUND: In an artificial pancreas (AP), the meals are either manually announced or detected and their size estimated from the blood glucose level. Both methods have limitations, which result in suboptimal postprandial glucose control. The GoCARB system is designed to provide the carbohydrate content of meals and is presented within the AP framework. METHOD: The combined use of GoCARB with a control algorithm is assessed in a series of 12 computer simulations. The simulations are defined according to the type of the control (open or closed loop), the use or not-use of GoCARB and the diabetics' skills in carbohydrate estimation. RESULTS: For bad estimators without GoCARB, the percentage of the time spent in target range (70-180 mg/dl) during the postprandial period is 22.5% and 66.2% for open and closed loop, respectively. When the GoCARB is used, the corresponding percentages are 99.7% and 99.8%. In case of open loop, the time spent in severe hypoglycemic events (<50 mg/dl) is 33.6% without the GoCARB and is reduced to 0.0% when the GoCARB is used. In case of closed loop, the corresponding percentage is 1.4% without the GoCARB and is reduced to 0.0% with the GoCARB. CONCLUSION: The use of GoCARB improves the control of postprandial response and glucose profiles especially in the case of open loop. However, the most efficient regulation is achieved by the combined use of the control algorithm and the GoCARB.

Glycemic Variability Is Associated with Frequency of Blood Glucose Testing and Bolus: Post Hoc Analysis Results from the ProAct Study.

INTRODUCTION: The ProAct study has shown that a pump switch to the Accu-Chek(®) Combo system (Roche Diagnostics Deutschland GmbH, Mannheim, Germany) in type 1 diabetes patients results in stable glycemic control with significant improvements in glycated hemoglobin (HbA1c) in patients with unsatisfactory baseline HbA1c and shorter pump usage time. PATIENTS AND METHODS: In this post hoc analysis of the ProAct database, we investigated the glycemic control and glycemic variability at baseline by determination of several established parameters and scores (HbA1c, hypoglycemia frequency, J-score, Hypoglycemia and Hyperglycemia Indexes, and Index of Glycemic Control) in participants with different daily bolus and blood glucose measurement frequencies (less than four day, four or five per day, and more than five per day, in both cases). The data were derived from up to 299 patients (172 females, 127 males; age [mean±SD], 39.4±15.2 years; pump treatment duration, 7.0±5.2 years). RESULTS: Participants with frequent glucose readings had better glycemic control than those with few readings (more than five readings per day vs. less than four readings per day: HbA1c, 7.2±1.1% vs. 8.0±0.9%; mean daily blood glucose, 151±22 mg/dL vs. 176±30 mg/dL; percentage of readings per month >300 mg/dL, 10±4% vs. 14±5%; percentage of readings in target range [80-180 mg/dL], 59% vs. 48% [P<0.05 in all cases]) and had a lower glycemic variability (J-score, 49±13 vs. 71±25 [P<0.05]; Hyperglycemia Index, 0.9±0.5 vs. 1.9±1.2 [P<0.05]; Index of Glycemic Control, 1.9±0.8 vs. 3.1±1.6 [P<0.05]; Hypoglycemia Index, 0.9±0.8 vs. 1.2±1.3 [not significant]). Frequent self-monitoring of blood glucose was associated with a higher number of bolus applications (6.1±2.2 boluses/day vs. 4.5±2.0 boluses/day [P<0.05]). Therefore, a similar but less pronounced effect on glycemic variability in favor of more daily bolus applications was observed (more than five vs. less than four bolues per day: J-score, 57±17 vs. 63±25 [not significant]; Hypoglycemia Index, 1.0±1.0 vs. 1.5±1.4 [P<0.05]; Hyperglycemia Index, 1.3±0.6 vs. 1.6±1.1 [not significant]; Index of Glycemic Control, 2.3±1.1 vs. 3.1±1.7 [P<0.05]). CONCLUSIONS: Pump users who perform frequent daily glucose readings have a better glycemic control with lower glycemic variability.

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.

Personalized tuning of a reinforcement learning control algorithm for glucose regulation.

Artificial pancreas is in the forefront of research towards the automatic insulin infusion for patients with type 1 diabetes. Due to the high inter- and intra-variability of the diabetic population, the need for personalized approaches has been raised. This study presents an adaptive, patient-specific control strategy for glucose regulation based on reinforcement learning and more specifically on the Actor-Critic (AC) learning approach. The control algorithm provides daily updates of the basal rate and insulin-to-carbohydrate (IC) ratio in order to optimize glucose regulation. A method for the automatic and personalized initialization of the control algorithm is designed based on the estimation of the transfer entropy (TE) between insulin and glucose signals. The algorithm has been evaluated in silico in adults, adolescents and children for 10 days. Three scenarios of initialization to i) zero values, ii) random values and iii) TE-based values have been comparatively assessed. The results have shown that when the TE-based initialization is used, the algorithm achieves faster learning with 98%, 90% and 73% in the A+B zones of the Control Variability Grid Analysis for adults, adolescents and children respectively after five days compared to 95%, 78%, 41% for random initialization and 93%, 88%, 41% for zero initial values. Furthermore, in the case of children, the daily Low Blood Glucose Index reduces much faster when the TE-based tuning is applied. The results imply that automatic and personalized tuning based on TE reduces the learning period and improves the overall performance of the AC algorithm.

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.

An Actor-Critic based controller for glucose regulation in type 1 diabetes.

A novel adaptive approach for glucose control in individuals with type 1 diabetes under sensor-augmented pump therapy is proposed. The controller, is based on Actor-Critic (AC) learning and is inspired by the principles of reinforcement learning and optimal control theory. The main characteristics of the proposed controller are (i) simultaneous adjustment of both the insulin basal rate and the bolus dose, (ii) initialization based on clinical procedures, and (iii) real-time personalization. The effectiveness of the proposed algorithm in terms of glycemic control has been investigated in silico in adults, adolescents and children under open-loop and closed-loop approaches, using announced meals with uncertainties in the order of ±25% in the estimation of carbohydrates. The results show that glucose regulation is efficient in all three groups of patients, even with uncertainties in the level of carbohydrates in the meal. The percentages in the A+B zones of the Control Variability Grid Analysis (CVGA) were 100% for adults, and 93% for both adolescents and children. The AC based controller seems to be a promising approach for the automatic adjustment of insulin infusion in order to improve glycemic control. After optimization of the algorithm, the controller will be tested in a clinical trial.

Real-time adaptive models for the personalized prediction of glycemic profile in type 1 diabetes patients.

BACKGROUND: Prediction of glycemic profile is an important task for both early recognition of hypoglycemia and enhancement of the control algorithms for optimization of insulin infusion rate. Adaptive models for glucose prediction and recognition of hypoglycemia based on statistical and artificial intelligence techniques are presented. METHODS: We compared an autoregressive (AR) model using only glucose information, an AR model with external insulin input (ARX), and an artificial neural network (ANN) using both glucose and insulin information. Online adaptive models were used to account for the intra- and inter-subject variability of the population with diabetes. The evaluation of the predictive ability included prediction horizons (PHs) of 30 min and 45 min. RESULTS: The AR model presented root mean square error (RMSE) values of 14.0-21.6 mg/dL and correlation coefficients (CCs) of 0.92-0.95 for PH=30 min and 23.2-35.9 mg/dL and 0.79-0.87, respectively, for PH=45 min. The respective values for the ARX models were slightly better (PH=30 min, 13.3-18.8 mg/dL and 0.94-0.96; PH=45 min, 22.8-29.4 mg/dL and 0.83-0.88). For the ANN, the RMSE values ranged from 2.8 to 6.3 mg/dL, and the CC was 0.99 for all cases and PHs. The sensitivity of hypoglycemia prediction was 78% for AR, 81% for ARX, and 96% for ANN for PH=30 min and 65%, 67%, and 95%, respectively, for PH=45 min. The corresponding specificities were 96%, 96%, and 99% for PH=30 min and 93%, 93%, and 99% for PH=45 min. CONCLUSIONS: The ANN appears to be more appropriate for the prediction of glucose profile based on glucose and insulin data.