Estimation of plasma insulin concentration under glycemic variability using nonlinear filtering techniques

Area of Science:

• Biomedical Engineering
• Control Systems Engineering
• Endocrinology

Background:

• Artificial pancreas systems aim to optimize insulin delivery for type 1 diabetes management.
• Accurate estimation of plasma insulin concentration is crucial for closed-loop control, but direct measurement is unavailable.
• Kalman filter variants are commonly used for estimating unmeasured states like insulin levels.

Purpose of the Study:

• To investigate the impact of glycemic variability on the accuracy of insulin concentration estimation using Kalman filter-based methods.
• To evaluate the performance of Extended (EKF), Cubature (CKF), and Unscented (UKF) Kalman filters under varying levels of glucose fluctuations.

Main Methods:

• A deterministic glucose-insulin interaction model was combined with a stochastic process for glycemic fluctuations.
• EKF, CKF, and UKF configurations were employed to estimate plasma insulin concentration.
• Simulations were conducted to assess filter performance across different levels of glycemic variability.

Main Results:

• Insulin state estimation accuracy was acceptable under conditions of low glycemic variability.
• As glycemic variability increased, the performance of EKF, CKF, and UKF significantly degraded.
• Large nonlinearities introduced by high glucose variability compromised the filters' ability to accurately estimate insulin.

Conclusions:

• Current Kalman filter-based insulin estimation methods may be insufficient for individuals with high glycemic variability.
• The accuracy of artificial pancreas control is potentially compromised by unaddressed glucose fluctuations.
• Further research is needed to develop robust estimation algorithms that account for significant inter- and intra-patient variability in glucose-insulin dynamics.