The Impact of Model-based Therapeutics on Glucose Control in an Intensive Care Unit

Abstract

-This paper investigates the impact of fast parameter identification methods, which do not require any forward simulations, on model-based glucose control, using retrospective data in the Christchurch Hospital Intensive Care Unit. The integral-based identification method has been previously clinically validated and extensively applied in a number of biomedical applications; and is a crucial element in the presented model-based therapeutics approach. Common non-linear regression and gradient descent approaches are too computationally intense and not suitable for the glucose control applications presented. The main focus in this paper is on better characterizing and understanding the importance of the integral in the formulation and the effect it has on modelbased drug therapy control. As a comparison, a potentially more natural derivative formulation which has the same computation speed advantages is investigated, and is shown to go unstable with respect to modelling error which is always present clinically. The integral method remains robust.

Model-Based Design of Experiments in the Presence of Continuous Measurement Systems

“…Measurements are available with a maximum relative error of 15% on the readings. This high noise level is consistent with the accuracy of recent CGMSs . Uncertainty on the prior estimation of the metabolic parameters is taken into account during the design phase assuming an initial guess on model parameters of θ 0 = [0.0235 0.0189 0.92 E -5] T , corresponding to a relative deviation of 40% from the true metabolic parameters θ describing the physiology of the diabetic subject.…”

Online redesign of clinical tests for the identification of type 1 diabetes models in the presence of continuous glucose monitoring systems

“…It is assumed that the "synthetic" glucose measurements (obtained by simulation of HWM at Θ) are modelled as follows: a. discrete sampling: blood glucose measurements are corrupted by Gaussian noise with zero mean and a constant relative deviation of 0.03; b. CGMS: following Hann et al (2009), the measurements are supposed to be affected by a 7% low frequency modelling error added to a 18 % Gaussian mean error:…”