סמינר מחקר בחוג להנדסה ביו רפואית

Dr. Amjad H. Abu-Rmileh

University of Girona – Spain

Control and modeling techniques for closed-loop artificial pancreas in type 1 diabetes

The artificial pancreas is an automated closed-loop control system that is applied for glucose regulation in patients with type 1 diabetes mellitus, who suffer severe lack of insulin production. The artificial pancreas has the potential to reduce the frequency and severity of the diabetes complications and improve the patients’ quality of life. However, a closed-loop system is not yet commercially available, and blood glucose control is still a challenging problem in biomedical engineering.

We present different control strategies, for the closed-loop artificial pancreas, that are based on Model Predictive Control (MPC) and Sliding Mode Control (SMC). Multiple MPC with linear models and gain scheduling, and SMC with linear and nonlinear models, have been developed. The proposed control strategies combine more than one linear/nonlinear control and modeling approaches in one structure. The main idea behind such combined approaches is to make use of the virtues of each approach while reducing the effects of their drawbacks. The control strategies are tested and validated in simulations (in-silico validation). For the in-silico testing, two mathematical models are used, simulating patients with type 1 diabetes. The control strategies are tested under different conditions, such as the presence of meal disturbance, patient variability, sensor errors and time-delay.

Then, we propose the use of frequency domain techniques to analyze and compare the response of the two frequently used models of the glucose regulation system. The analysis provides insight into the underlying linear dynamics, the nature and level of model nonlinearity, and the effect of nonlinear behavior on the linear modeling approach.

.