Extremum Seeking Control for Personalized Zone Adaptation in Model Predictive Control for Type 1 Diabetes

Area of Science:

• Biomedical Engineering
• Control Systems
• Endocrinology

Background:

• Zone model predictive control (MPC) is an established closed-loop strategy for type 1 diabetes (T1D) glucose regulation.
• Individualized glycemic control remains a challenge due to patient variability and lifestyle factors.

Purpose of the Study:

• To develop a universal, model-free optimization scheme for adapting the blood glucose target zone for individual T1D patients.
• To minimize the clinical glycemic risk index, relative regularized glycemic penalty index (rrGPI), through adaptive zone modulation.

Main Methods:

• Employed extremum seeking control (ESC) to adapt the upper and lower bounds of the blood glucose target zone.
• Utilized sinusoidal modulation and demodulation of rrGPI readings for gradient estimation.
• Incorporated a decaying feedback gain and vanishing dither signal to enhance robustness against uncertainties.

Main Results:

• In-silico trials demonstrated that personalized optimized zones were achieved within one week.
• The proposed adaptive zone method outperformed the conventional fixed zone [80, 140] mg/dL for both announced and unannounced meals.
• The developed method exhibited strong robustness against real-life uncertainties in glycemic control.

Conclusions:

• Model-free adaptive zone optimization offers a promising approach for personalized T1D management.
• This method enhances glycemic control efficacy and robustness compared to fixed target zones.
• The rrGPI-based adaptation strategy provides an effective mechanism for individualizing T1D treatment.