Development of a multi-parametric model predictive control algorithm for insulin delivery in type 1 diabetes mellitus...

Area of Science:

• Biomedical Engineering
• Control Systems Engineering
• Endocrinology

Background:

• Type 1 diabetes mellitus (T1DM) requires continuous glucose monitoring and insulin management.
• Existing insulin delivery systems face challenges with efficiency, robustness, and user burden.
• Model predictive control (MPC) offers potential for advanced glycemic control.

Purpose of the Study:

• To develop a computationally efficient, robust, and user-friendly multi-parametric model predictive control (mpMPC) algorithm for subcutaneous insulin delivery in T1DM.
• To investigate system identification methods using ambulatory impulse response tests and readily available clinical parameters.
• To incorporate safety constraints using clinical data to minimize hypoglycemia.

Main Methods:

• Developed and simulated a multi-parametric model predictive control (mpMPC) algorithm for subcutaneous insulin delivery.
• Utilized the UVa/Padova T1DM simulator for system identification and closed-loop simulations.
• Assessed controller robustness through subject/model mismatch scenarios and noise injection.
• Incorporated clinical parameters as safety constraints and controller model inputs.

Main Results:

• A second-order-plus-time-delay transfer function model accurately represented system dynamics (RMSE 26 mg/dL).
• The mpMPC algorithm maintained a low risk of hypoglycemia for 90% of simulated subjects without carbohydrate intake information.
• Low-order linear models with clinical parameters proved sufficient for effective glycemic control.

Conclusions:

• Clinically meaningful parameters and low-order models are adequate for MPC-based glycemic control in T1DM.
• Integrating clinical knowledge as safety constraints and controller models reduces hypoglycemia and improves glycemic management.
• The proposed mpMPC algorithm is compact and suitable for implementation on electronic devices.