Multi-level Supervision and Modification of Artificial Pancreas Control System

Area of Science:

• Biomedical Engineering
• Control Systems
• Endocrinology

Background:

• Artificial pancreas (AP) systems automate blood glucose regulation for type 1 diabetes (T1D) management using continuous glucose monitoring (CGM), a controller, and an insulin pump.
• Model-based controllers in AP systems are sensitive to model inaccuracies caused by metabolic changes or equipment performance issues, potentially leading to erroneous insulin dosing.
• Sensor errors and signal loss can significantly impact AP system performance and patient safety.

Purpose of the Study:

• To develop and evaluate a multi-level supervision and controller modification (ML-SCM) module for enhancing AP system performance and safety.
• To improve the accuracy of insulin infusion rates and broaden the safety range of blood glucose concentration (BGC) in T1D patients using AP systems.

Main Methods:

• Developed an ML-SCM module with three supervision levels: sample, period, and day.
• Implemented online controller performance assessment, sensor error detection, and signal reconciliation at the sample level.
• Tuned the controller more aggressively at the period level and adjusted controller parameters at the day level based on CGM data.

Main Results:

• Simulations with 30 subjects in the UVa/Padova metabolic simulator demonstrated improved AP performance with the ML-SCM module.
• A clinical experiment confirmed the module's effectiveness in a real-world setting.
• The AP system integrated with the ML-SCM module achieved a safer BGC distribution and more appropriate insulin infusion rate suggestions.

Conclusions:

• The ML-SCM module effectively supervises and retunes AP controllers, enhancing system reliability and safety.
• This advanced supervision strategy leads to better glycemic control outcomes for individuals with T1D.
• The ML-SCM module represents a significant advancement in AP technology for diabetes management.