Closed-loop MPC-based artificial pancreas: Handling circadian variability of insulin sensitivity

SERENO, JUAN; FERRAMOSCA, ANTONIO; ABUIN, PABLO; GONZÁLEZ, ALEJANDRO HERNÁN

Buenos Aires

Congreso; XXVIIº Congreso Argentino de Control Automático; 2020

Closed-loop glycemic control algorithms havedemonstrated the ability to improve glucose regulation in patientswith type 1 diabetes mellitus (T1DM), both in silico and clinicalresearch trials. However, many of the proposed control strategieshave been developed under time-invariant linear models, withoutconsidering the well-know intraday parametric variability. Inthis work, a recently published pulsatile Zone Model PredictiveControl (pZMPC) is analyzed under a real observed pattern ofintraday insulin sensitivity (SI ), according to the latest updates ofthe FDA-approved UVA/Padova simulator. To this end, nominalinformation of the postprandial insulin sensitivity is explicitlyincluded in the control-oriented model to be taken into accountin the predictions. The proposed control strategy is evaluatedin silico under the FDA-approved UVA/Padova simulator, theperformance analysis is done using statistical metrics and itis compared with the invariant version of the aforementionedcontroller. Results show a significant improvement in the analyzedmetrics when the time-variant control-oriented modelis considered in the predictions, which justifies the proposedincrement in the controller complexity.