Dynamic model with time varying delay for type 1 diabetes mellitus identified by using expectation maximization algorithm

• Published in: Chinese Control Conference pp. 9376 - 9381
• Main Authors: Zeng, Fanmao, Wang, Youqing
• Format: Conference Proceeding Journal Article
• Language: English
• Published: TCCT 01.07.2016
• Subjects: Algorithms; autoregressive with exogenous inputs (ARX); clinical trial data; Delay; Delays; Diabetes; Dynamic models; EM algorithm; Glucose; Heuristic algorithms; Hidden Markov models; Insulin; Mathematical model; Mathematical models; Maximization; model identification; Patients; Sugar; time-varying delay; type 1 diabetes mellitus (T1DM)
• ISSN: 1934-1768
• DOI: 10.1109/ChiCC.2016.7554848

Model identification and glucose prediction for patients with type 1 diabetes mellitus (T1DM) have drawn much attention over the past decade. The rate of insulin absorption of patients with T1DM usually changes over time. This study collected the clinical data from two patients with T1DM in China-Japan Friendship Hospital. In all previous studies concerning dynamic model identification for T1DM, it was assumed that the delay of insulin absorption is constant for each subject. However, it is not true, the delay can be varying at every sample. In this study, an autoregressive with exogenous inputs (ARX) model was used to describe the dynamics of glucose-insulin and the varying delay is modeled as a Markov Chain., The expectation maximization (EM) algorithm was used to identify the parameters in the model. Comparison of the accuracy of identification and glucose prediction using the EM algorithm and conventional least square algorithm shows the practicability and efficacy of the EM algorithm.