Evolutionary LTI models of Glucose-Carbohidrate-Insulin interaction

Abstract

A key issue for a proper control and treatment of diabetes patients is to have an accurate model of the interaction between glucose, carbohidrates (CHO) and exogenous insulin infusions. There are two main modeling trends. The rst trend is focused on the physiology and pharmacokinetics of the glucoregulatory system in order to provide realistic in-silico patients for developing further medical studies. The second trend is focused on models that are to be employed in the so called articial pancreas (AP). Currently comercialized APs implement predictive stategies based on a linear and time-invariant (LTI) models described by a single-input single-output (SISO) transfer function; which is usually identied using in-silico patients. The most recent approach to personalized models consists of testing the actual patient with CHO intakes and insulin bolus for registering the body response in terms of glucose and insulin rate appearence in blood. Considering these responses as inputs instead of intakes and insulin infusions recorded in a log simplies the model considerably; but it requires to visit the hospital and to follow the test protocol.

J. Manuel Colmenar

Associate Professor

My research interests are focused on metaheuristics applied to optimization problems. I have worked on different combinatorial optimization problems applying trajectorial algorithms such us GRASP or VNS. Besides, I am very interested in applications of Grammatical Evolution, specifically in model and prediction domain, as alternative to machine learning approaches.