A closed-loop artificial pancreas using model predictive control and a sliding meal size estimator.

Area of Science:

• Biomedical Engineering
• Endocrinology
• Control Systems

Background:

• Artificial pancreas systems aim to automate glucose control in diabetes management.
• Accurate meal detection and insulin dosing are critical challenges in closed-loop systems.
• Existing systems often require manual meal input, leading to potential errors.

Purpose of the Study:

• To develop and evaluate a comprehensive closed-loop artificial pancreas strategy.
• To create a meal detection and size estimation algorithm for unannounced meals.
• To improve glucose regulation by integrating advanced control features and safety mechanisms.

Main Methods:

• Development of a meal detection and size estimation algorithm.
• Incorporation of insulin-on-board constraints using a pharmacodynamic model.
• Implementation of a supervisory pump shut-off feature to prevent hypoglycemia.
• Utilizing model predictive control (MPC) with a validated MPC model.
• Testing in a simulated clinical trial with adolescent and adult subjects.

Main Results:

• The integrated artificial pancreas system with the new meal estimation algorithm achieved a daily mean glucose of 138 mg/dl for adolescents and 132 mg/dl for adults.
• This represents a substantial improvement compared to the MPC-only approach (159 mg/dl and 145 mg/dl, respectively).
• Performance comparison demonstrated the superiority of the newly proposed meal size estimation algorithm over previous methods.

Conclusions:

• The developed artificial pancreas strategy significantly improves glycemic control in simulated diabetic populations.
• The novel meal detection and estimation algorithm is effective in managing glucose levels during unannounced meals.
• This integrated system offers a promising advancement for automated diabetes management.