Empirical Modeling for Glucose Control in Critical Care and Diabetes

摘要: Empirical Volterra series models of glucose–insulin behavior were identified from input–output data provided by a physiologically-based nonlinear patient model. While completely accurate output prediction, clinical limitations constrained the structure to linear plus diagonal coefficients. In absence measurement noise, ability this structurally model capture process dynamics was very good. The addition Gaussian distributed having variance 10 mg2/dL2, significantly degraded coefficient estimates, but projection onto Laguerre basis (with subsequent expansion space for analysis) excellent noise-filtering and predictions requiring only 164 measurements identify 121 Linear predictive control algorithms developed models. mild performance enhancement in rejecting 50 g glucose challenge. With noise matching above characteristics, benefits lost. superiority empirical open-loop did not translate into closed-loop presence noise. control, with filter effects proper tuning, best combination robustness uncertainty (both model) case study.