A contraction theory approach to observer-based controller design for glucose regulation in type 1 diabetes with intra-patient variability

TL;DR

This paper introduces a contraction theory-based observer controller for type 1 diabetes management, effectively handling intra-patient variability and meal disturbances without additional safety algorithms or insulin sensors.

Contribution

It presents a novel observer-based control method using contraction analysis for glucose regulation in T1D, addressing variability and sensor limitations.

Findings

Achieved over 73% time in target glucose range across virtual patients.

Reduced hyperglycemia episodes without feed-forward meal compensation.

Validated effectiveness under realistic multi-meal scenarios.

Abstract

While the Artificial Pancreas is effective in regulating the blood glucose in the safe range of 70-180 mg/dl in type 1 diabetic patients, the high intra-patient variability, as well as exogenous meal disturbances, poses a serious challenge. The existing control algorithms thus require additional safety algorithms and feed-forward actions. Moreover, the unavailability of insulin sensors in Artificial Pancreas makes this task more difficult. In the present work, a subcutaneous model of type 1 diabetes (T1D) is considered for observer-based controller design in the framework of contraction analysis. A variety of realistic multiple-meal scenarios for three virtual T1D patients have been investigated with +30 % and -30 % of parametric variability. The average time spent by the three T1D patients is found to be 77 %, 73 % and 76 %, respectively. A significant reduction in the time spent in hyperglycemia (>180 mg/dl) is achieved without any feed-forward action for meal compensation.