摘要：This paper develops a novel approach to data-driven optimization of insulin pump treatment parameters in Type 1 Diabetes (T1D). In this approach, records of continuous glucose monitoring (CGM), insulin delivery, and meal records are used (i) to retrospectively estimate samples of the disturbance process that is responsible for daily variability in blood glucose and (ii) to optimize the parameters of functional insulin therapy (i.e. the patient’s basal rate, correction factor, and carbohydrate ratio profiles) against the ensemble of estimated disturbance process samples. We illustrate the proposed methodology through retrospective application to data collected in a 30-day field study of patients with T1D, as well as through in silico pre-clinical trials using the FDA-accepted Virginia / Padova Type 1 Simulator.
关键词：Metabolic Engineering and Systems BiologyModel Based ControlModelling and System Identification
