Neural Network based Glucose - Insulin Metabolism Models for Children with Type 1 Diabetes

Abstract

In this paper two models for the simulation of glucose-insulin metabolism of children with Type 1 diabetes are presented. The models are based on the combined use of Compartmental Models (CMs) and artificial Neural Networks (NNs). Data from children with Type 1 diabetes, stored in a database, have been used as input to the models. The data are taken from four children with Type 1 diabetes and contain information about glucose levels taken from continuous glucose monitoring system, insulin intake and food intake, along with corresponding time. The influences of taken insulin on plasma insulin concentration, as well as the effect of food intake on glucose input into the blood from the gut, are estimated from the CMs. The outputs of CMs, along with previous glucose measurements, are fed to a NN, which provides short-term prediction of glucose values. For comparative reasons two different NN architectures have been tested: a Feed-Forward NN (FFNN) trained with the back-propagation algorithm with adaptive learning rate and momentum, and a Recurrent NN (RNN), trained with the Real Time Recurrent Learning (RTRL) algorithm. The results indicate that the best prediction performance can be achieved by the use of RNN.

Convolutional Recurrent Neural Networks for Glucose Prediction

“…In addition, we want to compare our algorithm with other approaches in the literature. Using dataset generated from simulators, our algorithm is better than the results of RMSE = 18.78mg/dL [34] and RMSE = 13.65mg/dL in [35]. For several other works, it is difficult to evaluate the RMSE through direct comparison due to the unavailability of codes, parameters of the models, and the benchmark datasets.…”

Inverse optimal neural control for a class of discrete‐time nonlinear positive systems

“…In addition, many of the results known for conventional system identification are applicable to the NN-based identification as well [29]. The use of NN for glucose modeling has been considered in different recent works [20,25,40,41]. The types and architectures of the NN employed in these works range from simple feedforward NNs [25,40,41] to sophisticated artificial ones [20].…”

Recent Trends in Blood Glucose Control Studies

“…Dalla Man et al [12] developed a detailed simulation model (Figure 1a) of the glucose-insulin system for healthy and type 2 diabetics considering meal rate of appearance, endogenous glucose production, glucose utilization and in-sulin secretion. Mougiakakou et al [13] developed a model ( Figure 1b) for type 1 diabetics including meal intake that consists of compartmental models of insulin kinetics and glucose absorption, and neural networks representing glucose kinetics. Recently, it has been understood that not only the carbohydrate amount in a particular meal but also glucose absorption rate influence blood glucose excursion.…”

“…Hence, mathematical models that represent glucose-insulin metabolism including meal digestion and absorption have also been developed [12,13] to realize an appropriate control for the postprandial state. Dalla Man et al [12] developed a detailed simulation model (Figure 1a) of the glucose-insulin system for healthy and type 2 diabetics considering meal rate of appearance, endogenous glucose production, glucose utilization and in-sulin secretion.…”