Classification-based deep neural network vs mixture density network models for insulin sensitivity prediction problem.

Model-based glycemic control (GC) protocols are used to treat stress-induced hyperglycaemia in intensive care units (ICUs). The STAR (Stochastic-TARgeted) glycemic control protocol - used in clinical practice in several ICUs in New Zealand, Hungary, Belgium, and Malaysia - is a model-based GC protocol using a patient-specific, model-based insulin sensitivity to describe the patient's actual state. Two neural network based methods are defined in this study to predict the patient's insulin sensitivity parameter: a classification deep neural network and a Mixture Density Network based method. Treatment data from three different patient cohorts are used to train the network models. Accuracy of neural network predictions are compared with the current model- based predictions used to guide care. The prediction accuracy was found to be the same or better than the reference. The authors suggest that these methods may be a promising alternative in model-based clinical treatment for patient state prediction. Still, more research is needed to validate these findings, including in-silico simulations and clinical validation trials.

Design of Luenberger observer for glucose-insulin control via mathematica.

Many articles dealing with insulin-glucose control have been published in the last decades, and they mostly assumed that all the system state variables are available for feedback. However, this is not usually the case, or they are not so cheap in practice as blood glucose measurements are. In this paper the use of the reduced-order estimator (also known as the Luenberger observer) is considered in symbolic form employing Polynomial Control System Application of Mathematica for the three-state minimal Bergman model, [1], as this can be used to reconstruct those state variables that are hard to be recovered directly from the system outputs: remote compartment insulin and plasma insulin. Nonlinear closed loop simulations with H(2)/H(infinity) control (disturbance rejection LQ method) showed that the observer, which is faster than the system itself, can provide a very good state recovery performance.

Robust blood-glucose control using Mathematica.

A robust control design on frequency domain using Mathematica is presented for regularization of glucose level in type I diabetes persons under intensive care. The method originally proposed under Mathematica by Helton and Merino, --now with an improved disturbance rejection constraint inequality--is employed, using a three-state minimal patient model. The robustness of the resulted high-order linear controller is demonstrated by nonlinear closed loop simulation in state-space, in case of standard meal disturbances and is compared with H infinity design implemented with the mu-toolbox of Matlab. The controller designed with model parameters represented the most favorable plant dynamics from the point of view of control purposes, can operate properly even in case of parameter values of the worst-case scenario.