Control-Oriented Model With Intra-Patient Variations for an Artificial Pancreas.

Area of Science:

• Biomedical Engineering
• Control Systems
• Metabolic Modeling

Background:

• Type 1 Diabetes Mellitus (T1DM) management requires accurate glucose-insulin system modeling.
• Existing models face challenges with intra-patient variability and inter-subject differences.
• Control-oriented models are crucial for developing effective glucose regulation strategies.

Purpose of the Study:

• To develop a novel low-order, control-oriented model for T1DM glucose regulation.
• To capture nonlinear dynamics and intra-patient insulin sensitivity variations.
• To improve upon previous T1DM control models for enhanced personalization and robustness.

Main Methods:

• Utilized the UVA/Padova metabolic simulator to derive a low-order model.
• Incorporated intra-patient variability in insulin sensitivity (S_I).
• Enabled model personalization using a priori patient information for Linear Parameter Varying (LPV) controller design.
• Evaluated model performance against a previous control-oriented model.

Main Results:

• The proposed model accurately represents subcutaneous glucose dynamics from insulin input.
• It effectively captures nonlinear system behavior and daily S_I variations.
• Demonstrated superior performance compared to prior T1DM control-oriented models in open- and closed-loop simulations.
• Showed reduced differences compared to the UVA/Padova simulator.

Conclusions:

• The developed low-order model offers improved accuracy and personalization for T1DM glucose regulation.
• Its structure is suitable for advanced LPV controller design.
• This advancement holds potential for more robust and reliable glycemia control systems in T1DM patients.