Bihormonal artificial pancreas system based on switching model predictive control

Automated regulation of blood glucose concentration has become a daily challenge for diabetes mellitus. Normally, there are two main hormones for glucose control, i.e. glucagon and insulin. In accordance with whether glucagon is infused, the artificial pancreas systems can be categorized into two types: unihormone type (single insulin infusion) and bihormone type (both insulin and glucagon infusion). However, the reported studies about bihormone AP system were very scarce since the research time is not long and the research content is very complicated. In this study, an optimal switching control method was proposed to design the logic of whether insulin or glucagon subsystems should be active, where the delivery rates of insulin as well as glucagon were designed by using model predictive control (MPC). The in silico results prove that the proposed bihormonal AP systems can perform outstandingly in respect of lowing the risk of hypoglycemia, smoothing the glucose level, and robustness with regard to insulin/glucagon sensitivity variations.