Simulation studies on neural predictive control of glucose using the subcutaneous route

Area of Science:

• Biomedical Engineering
• Control Systems
• Computational Biology

Background:

• Diabetes management requires precise glucose regulation.
• Subcutaneous glucose monitoring and insulin delivery present unique control challenges.
• Existing control strategies often struggle with physiological variability and time delays.

Purpose of the Study:

• To develop and evaluate a novel closed-loop glucose control strategy.
• To integrate neural networks with nonlinear model predictive control for enhanced performance.
• To assess the feasibility of subcutaneous glucose sensing and insulin infusion for automated control.

Main Methods:

• Utilized a radial basis function neural network for off-line system identification of a Nonlinear Auto Regressive model with eXogenous inputs (NARX) model.
• Employed nonlinear model predictive control (NPC) for glucose regulation based on the identified NARX model.
• Conducted simulation studies using a comprehensive glucose regulation model to evaluate the control strategy.

Main Results:

• The system identification successfully created a parsimonious NARX model of the glucoregulatory system.
• The developed controller demonstrated stable glucose regulation under significant noise and variable time delays.
• Simulation results indicate the potential for closed-loop glucose control using subcutaneous measurements and insulin administration.

Conclusions:

• Closed-loop glucose control is achievable with subcutaneous glucose monitoring and insulin infusion.
• The proposed hybrid control strategy shows promise for managing glucose levels.
• Limitations in subcutaneous insulin absorption necessitate continued patient involvement, particularly during meal times.