A Closed-Loop Artificial Pancreas Based on Risk Management:

摘要: Background:Control algorithms that regulate blood glucose (BG) levels in individuals with type 1 diabetes mellitus face several fundamental challenges. Two of these are the asymmetric risk clinical complications associated low and high irreversibility insulin action when using only insulin. Both nonlinearities force a controller to be more conservative uncertainties high. We developed novel extended model predictive (EMPC) explicitly addresses two challenges.Method:Our extensions control (MPC) operate three ways. First, they minimize combined hypoglycemia hyperglycemia. Second, integrate effect prediction into risk. Third, understand future actions will vary if measurements fall above or below predictions. Using University Virginia/Padova Simulator, we compared our against optimized versions proport...