Considering Plasma Insulin Concentrations in Adaptive Model Predictive Control for Artificial Pancreas Systems

Area of Science:

• Biomedical Engineering
• Control Systems
• Endocrinology

Background:

• Type 1 diabetes management requires precise insulin dosing to prevent hypoglycemia.
• Existing control algorithms often lack personalization and real-time adaptation.
• Overdose-induced hypoglycemia remains a significant risk in automated insulin delivery.

Purpose of the Study:

• To develop and evaluate an adaptive model predictive control (MPC) algorithm for type 1 diabetes management.
• To integrate a personalized plasma insulin concentration (PIC) estimator with adaptive models.
• To improve blood glucose concentration (BGC) regulation by adapting control constraints online.

Main Methods:

• Personalized PIC estimation to characterize temporal dynamics of PIC and BGC.
• Adaptive MPC algorithm incorporating PIC for insulin dose computation.
• Online adaptation of control problems and constraints based on individual metabolic state.

Main Results:

• The adaptive MPC algorithm effectively controlled BGC without manual meal information.
• The system avoided excessive insulin administration, mitigating hypoglycemia risk.
• Simulations showed an average of 71.14% of time spent within the target BGC range of [70, 180] mg/dL.

Conclusions:

• Adaptive MPC with integrated PIC estimation offers a promising approach for personalized glucose regulation in type 1 diabetes.
• The algorithm demonstrates robust performance in maintaining BGC within the target range while preventing hypoglycemia.
• This method enhances automated insulin delivery systems by providing on-line adaptation and constraint management.