Search

A typical artificial pancreas depends only on the continuous glucose monitoring (CGM) value for insulin dosing. However, both the insulin infusion and the glucose sensing are subject to time delays and slow dynamics. An automated and reliable meal onset information could enhance the control outcome of artificial pancreas by making it possible to infuse insulin earlier and thereby avoid large postprandial glucose excursions. In this study we employ abdominal sounds recorded in two healthy volunteers with a condenser microphone and propose an automated approach for meal onset detection from abdominal sounds. We use the Mel-frequency cepstral coefficients (MFCCs) and wavelet entropy extracted from the abdominal sounds as features. These features are fed to a simple feed forward neural network for discriminating meal from no-meal abdominal sounds. This approach detects meal onset with an average delay of 4.3 minutes in our limited number of subjects. More importantly, it provides lesser response delay than the state-of-the-art CGM based approach, which achieved a response delay ranging from 30- 40 minutes. The preliminary results indicate that the proposed abdominal sound-based approach may provide early meal onset information. This can be exploited in an artificial pancreas through allowable earlier meal insulin boluses, resulting in improved glycemic control. © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Accurate continuous glucose monitoring (CGM) is essential for fully automated glucose control in diabetes mellitus type 1. State-of-the-art glucose control systems automatically regulate the basal insulin infusion. Users still need to manually announce meals to dose the prandial insulin boluses. An automated meal detection could release the user and improve the glucose regulation. In this study, patterns in the postprandial CGM data are exploited for meal detection. Binary classifiers are trained to recognize the postprandial pattern in horizons of the estimated glucose rate of appearance and in CGM data. The appearance rate is determined by moving horizon estimation (MHE) based on a simple model. Linear discriminant analysis (LDA) is used for classification. The proposed method is compared to methods that detect meals when thresholds are violated. Diabetes care data from twelve free-living pediatric patients was downloaded during regular screening. Experts identified meals and their start by retrospective evaluation. The classification was tested by cross-validation. Compared to the threshold-based methods, LDA showed higher sensitivity to meals with a low rate of false alarms. Classifying horizons outperformed the other methods also with respect to time of detection. The onset of meals can be detected by pattern recognition based on estimated model states and consecutive CGM measurements. No individual tuning is necessary. This makes the method easily adopted in the clinical practice. © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

In classical approaches for an artificial pancreas, continuous glucose monitoring (CGM) is the only measured variable used for insulin dosing and additional control functions. The CGM values are subject to time delays and slow dynamics between blood and the sensing location. These time lags compromise the controller’s performance in maintaining (near to) normal glucose levels. Meal information could enhance the control outcome. However, meal announcement by the user is not reliable, and it takes 30 min to 40 min from meal onset until a meal is detected by methods based on CGM. In this pilot study, the use of bowel sounds for meal detection was investigated. In particular, we focused on whether bowel sounds change qualitatively during or shortly after meal ingestion. After fasting for at least 4 h, 11 healthy volunteers ingested a lunch meal at their usual time. Abdominal sound was recorded by a condenser microphone that was attached to the right upper quadrant of the abdomen by medical tape. Features that describe the power distribution over the frequency spectrum were extracted and used for classification by support vector machines. These classifiers were trained in a leave-one-out cross-validation scheme. Meals could be detected on average 10 min (std: 4.4 min) after they had started. Half of these were detected without false alarms. This shows that abdominal sound monitoring could provide an early meal detection. Further studies should investigate this possibility on a larger population in more general settings., Spectrogram of a recording in the training set. Windows with a length of 20 s and overlap of 10 s were analyzed using spectrogram function in MATLAB. The meal started at minute 51.

Glucose-insulin metabolism models are useful tools for research on diabetes, in development of diabetes-related medical devices like artificial pancreas systems, insulin pumps and continuous glucose monitors, and may also play a role in personalized decision support tools for people with diabetes. Such models are often highly nonlinear with many parameters that are person dependent. An example is the model used in the UVa/Padova T1DM simulator, which has a large number of states and parameters. It is desirable to be able to personalize such models through parameter identification based on limited glucose, meal and insulin data obtainable from free-living settings, as opposed to clinical research settings that have traditionally been required. In this paper we use the UVa-Padova T1DM simulator model in a case study to investigate observability of the model under different measurements, and the identifiability of its parameters as a function of the model's inputs and outputs. Structural identifiability is discussed and briefly investigated using the nonlinear Observability Rank Condition. Practical identifiability is discussed and investigated using sensitivity and Fisher information matrix analysis. We show how such analyses can be used to guide model reduction for improved identifiability, or to select the most proper subset of parameters to estimate. © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

The aim of this paper is to achieve a model for prediction of cerebral palsy based on motion data of young infants. The prediction is formulated as a classification problem to assign each of the infants to one of the healthy or with cerebral palsy groups. Unlike formerly proposed features that are mostly defined in the time domain, this study proposes a set of features derived from frequency analysis of infants' motions. Since cerebral palsy affects the variability of the motions, and frequency analysis is an intuitive way of studying variability, suggested features are suitable and consistent with the nature of the condition. In the current application, a well-known problem, few subjects and many features, was initially encountered. In such a case, most classifiers get trapped in a suboptimal model and, consequently, fail to provide sufficient prediction accuracy. To solve this problem, a feature selection method that determines features with significant predictive ability is proposed. The feature selection method decreases the risk of false discovery and, therefore, the prediction model is more likely to be valid and generalizable for future use. A detailed study is performed on the proposed features and the feature selection method: the classification results confirm their applicability. Achieved sensitivity of 86%, specificity of 92% and accuracy of 91% are comparable with state-of-the-art clinical and expert-based methods for predicting cerebral palsy.

A method for preprocessing a time series of glucose measurements based on Kalman smoothing is presented. Given a glucose data time series that may be irregularly sampled, the method outputs an interpolated time series of glucose estimates with mean and variance. The method can provide homogenization of glucose data collected from different devices by using separate measurement noise parameters for differing glucose measurement equipment. We establish a link between the ISO 15197 standard and the measurement noise variance used by the Kalman smoother for Self Monitoring of Blood Glucose (SMBG) measurements. The method provides phaseless smoothing, and it can automatically correct errors in the original datasets like small fallouts and erroneous readings when surrounding data allows. The estimated variance can be used for deciding at which times the data are trustworthy. The method can be used as a preprocessing step in many kinds of glucose data processing and analysis tasks, such as computing the Mean Absolute Relative Deviation (MARD) between measurement systems, or estimating the plasma-to-interstital fluid glucose dynamics of continuous glucose monitor (CGM) or Flash Glucose Monitor (FGM) signals. The method is demonstrated on SMBG and FGM glucose data from a clinical study. © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Robot-assisted therapy is an emerging form of rehabilitation treatment for motor recovery of the upper limbs after neurological injuries such as stroke or spinal cord injury. Robotic rehabilitation devices have the potential to reduce the physical strain put on therapists due to the high-effort one-to-one interactions between the therapist and patient involving repetitive high-intensity movements to restore arm and hand functions. Numerous custom robotic devices have been developed in recent years to aid in physical rehabilitation of stroke patients, but most commercially available systems are high-cost devices because of low production volumes and high development costs. In this paper, we analyse the safety and functionality of the UR5 collaborative industrial robot from Universal Robots equipped with an external force/torque sensor in a real-time control system for typical rehabilitation exercises. The aim of the paper is to show that a new class of general-purpose industrial robots designed for human-robot collaboration may prove a viable alternative to custom designs. Experiments show that robotic rehabilitation of the upper limbs using a standard industrial robot manipulator UR5 may be feasible. Results have the potential to make robotic rehabilitation more available as a high-quality therapeutic treatment for more patients. © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Author postprint As a step toward enabling snake robots to move in cluttered environments, this paper proposes a control strategy that combines environment adaptation with directional control in order to achieve straight line path following control in environments with obstacles. Moreover, the paper presents the design of a mechanical snake robot with tactile sensing capabilities that allow the robot to sense its environment. Experimental results are presented where the snake robot is successfully propelled through different obstacle environments with the proposed control strategy.

Cerebral palsy (CP) is a perinatally acquired nonprogressive brain damage resulting in motor impairment affecting mobility and posture. Early identification of infants with CP is desired to target early interventions and follow-up. During early infancy, distinct motion patterns occur which are highly predictive for later disability. These motor patterns can be observed and recorded. In this paper, a method to predict later CP based on early video recordings of the infants' spontaneous movements, applying optical flow and statistical pattern recognition, is presented. We extract motion information from video recordings of young infants using a total variation related optical flow method. By using wavelet analysis features from motion trajectories of points initiated on a regular grid were extracted and classified using a support vector machine.

This is the authors accepted and refereed manuscript to the article. In a recent paper, the authors have proposed a control strategy for a snake robot during obstacle-aided locomotion. In this paper, experimental results are presented where the controller is shown to successfully maintain the forward propulsion of a physical snake robot in a course with different obstacle configurations. This is the authors accepted and refereed manuscript to the article. ("(c) 2011 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works.

This paper investigates the controllability and stability properties of a planar snake robot influenced by anisotropic viscous ground friction. An analysis of the model shows that any asymptotically stabilizing control law for the robot to an equilibrium point must be time-varying. Furthermore, the analysis shows that the snake robot (with four links) is strongly accessible from almost any equilibrium point, except for certain singular configurations, and that the robot does not satisfy sufficient conditions for small-time local controllability (STLC). Averaging theory is employed to model the average velocity of the snake robot during lateral undulation. It is proven that the average velocity will converge exponentially fast to a steady state velocity, and an analytical expression for calculating the steady state velocity of the robot with an arbitrary number of links is presented. The paper presents simulation results that support the theoretical findings. © 2010 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.