Modular closed-loop control of diabetes

Abstract

Modularity plays a key role in many engineering systems, allowing for plug-and-play integration of components, enhancing flexibility and adaptability, and facilitating standardization. In the control of diabetes, i.e., the so-called "artificial pancreas," modularity allows for the step-wise introduction of (and regulatory approval for) algorithmic components, starting with subsystems for assured patient safety and followed by higher layer components that serve to modify the patient's basal rate in real time. In this paper, we introduce a three-layer modular architecture for the control of diabetes, consisting in a sensor/pump interface module (IM), a continuous safety module (CSM), and a real-time control module (RTCM), which separates the functions of insulin recommendation (postmeal insulin for mitigating hyperglycemia) and safety (prevention of hypoglycemia). In addition, we provide details of instances of all three layers of the architecture: the APS© serving as the IM, the safety supervision module (SSM) serving as the CSM, and the range correction module (RCM) serving as the RTCM. We evaluate the performance of the integrated system via in silico preclinical trials, demonstrating 1) the ability of the SSM to reduce the incidence of hypoglycemia under nonideal operating conditions and 2) the ability of the RCM to reduce glycemic variability.

Fig. 1

Modular architecture for closed-loop control of diabetes.

Fig. 2

SSM safety supervision algorithm.

Fig. 3

CV GA with level lines labeled according to performance index. The dots represent virtual patients while the red line is the so-called calibration curve for the patient #100. The scenario refers to the virtual protocol discussed in Section V-A. The numbers of patients in the regions A, B, C, and D are reported at the top of the plot.

Fig. 4

Median (continuous line) and the 2.5% (dashed) and 97.5% (dashed) percentiles of the glucose time series are plotted for Scenario 3 (insulin sensitivity perturbation) subject to either FF (thin, red line) or +RCM (heavy, blue line) control strategy. Target (70–180 mg/dL) and tight target (80–140 mg/dL) limits are shown in black (solid and dotted, respectively). The variability reduction due to the +RCM system (i.e., FF with both the RCM and SSM) is apparent, along with the system’s ability to prevent hypoglycemic episodes. (In the 10:00 am–2:00 pm time frame the in silico patients use conventional (FF) therapy, and, then, at 2:00 pm they switch as appropriate to the closed-loop +RCM strategy.) The outcome metrics reported in Table I are computed beginning 2 h after closed-loop control, i.e., at 4:00 pm (green line).

Fig. 5

Boxplots of plasma glucose in the four scenarios. In nominal conditions (Scenario 1) the three control strategies, FF, +SSM, +RCM, show similar performance. In the perturbed scenarios, variability increases under the FF strategy, with substantial risk of hypoglycemia in Scenarios 3 and 4. Safety is restored by the SSM module at the cost of an increase of glucose levels, especially in Scenario 3. A further improvement is introduced by the addition of the closed-loop action of the RCM module in the fully integrated strategy, reducing variability in all nonnominal scenarios.

Fig. 6

Performance of the FF, +SSM, and +RCM control strategies as measured by the time in target range of 70–180 (mg/dL). Note that in Scenario 4 FF performance is improved with the addition of the SSM, and a further improvement over both FF and SSM is observed in Scenarios 2–4 after the inclusion of RCM.

Fig. 7

Performance of the FF, +SSM, and +RCM control strategies as measured by the time in the “tight” target range of 80–140 (mg/dL). Comparing FF to +SSM in Scenarios 1–3 the hypoglycemia-reducing tendency of the SSM actually leads to a reduction of the time spent in the tight target range. However, with the addition of the RCM, the fully integrated system improves over FF in all scenarios.

Fig. 8

CVGA plot for 100 virtual patients under the FF and +SSM strategies. Note that in nominal conditions (Scenario 1) both strategies control the patients well. However, in the nonnominal Scenarios 2–4, the FF strategy fails to avoid low-glucose values. In particular, in Scenario 3 (Scenario 4) there are ten (31) patients suffering from hypoglycemic events. On the contrary, SSM is very effective in preventing hypoglycemia in that no hypo event occurs in any scenario.

Fig. 9

CVGA plot for 100 virtual patients under the +SSM and +RCM strategies. As observed in Fig. 8, the SSM effectively prevents hypoglycemia. Here, it can be seen that, in all scenarios, the RCM increases the number of patients in region A (optimal regulation). This is achieved without jeopardizing safety as only one patient is subject to hypoglycemia in Scenario 4.