UC Santa Barbara

Type 1 diabetes mellitus (T1DM) is a chronic disease characterized by the body’s inability to produce insulin, leading to chronically high blood glucose (BG) concentrations. T1DM is treated by frequent self-administration of insulin based on BG measurements; however, there is a fine line between too little and too much insulin, and an overdose can lead to a dangerous drop in BG. The artificial pancreas (AP), consisting of a glucose sensor, an insulin pump, and a feedback control algorithm, will replace self-treatment by automatically calculating and delivering insulin dosages based on continuous glucose measurements. Many iterations of the AP utilize commercially available subcutaneous (SC) insulin pumps and glucose sensors, but these devices introduce physiological limitations that make control difficult.

In this work, we present a clinical evaluation of an AP that uses SC devices, as well as an investigation of the intraperitoneal (IP) space as an alternative site for insulin delivery and glucose sensing to improve AP performance. Our results show that glucose sensors placed in the IP space have a lower time constant than SC sensors, allowing the controller to respond more quickly to BG disturbances. Similarly, insulin delivered through the IP space has faster pharmacokinetic and pharmacodynamic characteristics than SC insulin. Based on models of the sensing and actuation dynamics, a proportional-integral-derivative control algorithm with anti-reset windup protection was designed for the IP-IP route and evaluated on 10 simulated T1DM subjects. Using the IP-IP route led to a more robust controller that provided excellent control during the simulation studies. Our results support the development of a fully implantable AP that will operate within the IP space to safely and effectively control BG levels.