Estimation of parameters for plasma glucose regulation in type‐2 diabetics in presence of meal. Issue 1 (1st February 2018)

• Record Type: Journal Article
• Authors: Biswas, Prova; Sutradhar, Ashoke; Datta, Pallab
• Is Part Of: IET systems biology. Volume 12:Issue 1(2018)
• Page Start: 18
• Page End: 25
• Publication Date: 2018-02-01
• Subjects: blood -- diseases -- biochemistry -- parameter estimation -- biological tissues -- liver -- Bayes methods -- nonlinear filters -- Kalman filters -- drugs -- drug delivery systems -- medical signal processing; automated insulin delivery system -- closed‐loop controller -- hyperglycaemic patients -- drug dose -- root mean square error -- cubature quadrature KF -- Kalman filter -- type II diabetes mellitus -- process noise -- Bayesian nonlinear filter -- glucose‐insulin homoeostasis model -- interstitial fluid -- liver -- portal vein -- insulin mass -- biological tissues -- intestine -- stomach -- glucose mass -- meal intake -- type‐2 diabetics -- plasma glucose regulation -- parameter estimation; Systems biology -- Periodicals; Cell physiology -- Periodicals; Biological systems -- Mathematical models -- Periodicals; Genetics -- Mathematical models -- Periodicals; Computational biology -- Periodicals; 573
• DOI: 10.1049/iet-syb.2017.0036
• Languages: English
• ISSNs: 1751-8849

Abstract

Abstract : In this study, the authors propose a methodology for the estimation of glucose masses in stomach (both in solid and liquid forms), intestine, plasma and tissue; insulin masses in portal vein, liver, plasma and interstitial fluid using only plasma glucose measurement. The proposed methodology fuses glucose–insulin homoeostasis model (in the presence of meal intake) and plasma glucose measurement with a Bayesian non‐linear filter. Uncertainty of the model over individual variations has been incorporated by adding process noise to the homoeostasis model. The estimation is carried out over 24 h for the healthy people as well as a type II diabetes mellitus patients. In simulation, the estimator follows the truth accurately for both the cases. Moreover, the performances of two non‐linear filters, namely the unscented Kalman filter (KF) and cubature quadrature KF are compared in terms of root mean square error. The proposed methodology will be helpful in future to: (i) observe a patient's insulin–glucose profile, (ii) calculate drug dose for any hyperglycaemic patients and (iii) develop a closed‐loop controller for automated insulin delivery system.