Embedded Cubature Kalman Filter for Glucose and Insulin Concentration Estimation Using Noisy Glucose Sensor Data and Multiple Meal Disturbances

Abstract

This letter introduces an embedded cubature Kalman filter (ECKF) based on the fifth-degree embedded cubature rule for estimating unmeasured/hidden concentrations of plasma insulin (PIC) and interstitial insulin (IIC). The design of a robust and intelligent controller for blood glucose (BG) regulation requires estimates of PIC and IIC. The nonavailability of insulin sensors necessitates the use of a mathematical model to estimate PIC and IIC. We have integrated Bergman's minimal model of dynamic glucose-insulin relations with the glucose sensor measurement. The dynamic model and stochastic process that accounts for fluctuations in BG levels were integrated into ECKF to estimate PIC and IIC resulting from multiple meal disturbances. The root mean square error results demonstrate the improved estimation accuracy of the proposed method. © 2017 IEEE.