Improved Generalized Predictive Control Algorithm for Blood Glucose Control of Type 1 Diabetes

Abstract

Artificial pancreas (AP) is an important treatment for patients with Type 1 diabetes (T1D). The control algorithm adopted in an AP system determines its reliability and accuracy. The generalized predictive control (GPC) is a representative adaptive control algorithm and has been widely applied to AP systems. However, we found that the traditional GPC controller does not work well for adolescents with T1D because of their high-fluctuating blood glucose and high insulin resistance. Here, we propose an improved GPC algorithm with an adaptive reference glucose trajectory and an adaptive softening factor. The slopes of the reference trajectory and the value of softening factor are calculated real-time on the basis of the blood glucose concentration (BGC) variations. In silico testing was done using the US Food and Drug Administration (FDA) approved virtual patient software T1D mellitus. The BGC trace and density of 20 patient-subjects (10 adults and 10 adolescents) were recorded. Results showed that the average BGC percentage within the target regions (70-180 mg/dL) of the tests with adaptive reference glucose trajectory and softening factor for adolescents (0.93 ± 0.07) was significantly higher than that of the traditional GPC algorithm tests (0.88 ± 0.11), suggesting that the control quality of the blood glucose of adolescents is significantly improved with our GPC algorithm. Therefore, our improved GPC controller is effective and should have a good applicability in AP systems.

Artificial Organs 2019: A year in review

“…Wenping Liu et al of the Guangdong Food and Drug Vocational College, Guangzhou, China reported a generalized predictive control (GPC) algorithm for glucose control to control for high fluctuating blood glucose and high insulin resistance as occurs with adolescents. Their GPC algorithm has an adaptive reference glucose trajectory and an adaptive softening factor.…”

In-silico Testing of a Generalized Predictive Control Based Artificial Pancreas for Child Patients with Type 1 Diabetes

“…A GPC controller is designed and applied for blood glucose regulation in our previous research [16]. Insulin injection rate is computed by minimizing J in the following function.…”

“…α denotes the weight of ( ), also known as softening factor. The minimization of function (1) gives the optimal control action (insulin infusion rate) and its solution process is introduced in our previous research in detail [16].…”

“…The scenario was demonstrated as below [16]. Each test was run for 3 days (72 hours), and each child was provided 3 meals a day.…”

“…Its performance was tested by using the UVA/Padova type 1 diabetes mellitus simulator (T1DMS) approved by the US FDA, involving 10 adults and 10 adolescents. Test results showed that it realized the effective control of BGCs of those 20 in-silico subjects [16]. Here, four GPC controller, including the traditional GPC controller, the GPC+ARGT controller, the GPC+ASF controller and the GPC+ARGT+ASF controller, were tested with 10 in-silico children by using the UVA/Padova T1DMS.…”