Virtual type 1 diabetic patient for feedback control systems

Research and development of an external wearable artificial pancreas, although improved by recent technology achievements, still require simulation tools like the mathematical model introduced in the present study. The aim was to realize a mathematical model of insulin and glucose metabolism in type I diabetes and to test its effectiveness for development and assessment of control algorithms. A Proportional-Derivative-2nd derivative (PDD2) control algorithm was tested on a “virtual patient” constituted by the previously developed mathematical model simulating the patient response to a standard meal (90 g of carbohydrates, then increased and decreased). PDD2 controller performances were compared with the results of standard therapy (preprandial bolus) in continuous subcutaneous insulin infusion therapy (CSII). Area under the curve of glycaemic response (QG) and hyperglicemia duration (Durat) were chosen as performance indexes. In response to medium-light meals PDD2 reduced 35% of Durat and 25% of QG in respect to standard therapy. The average values of model estimated parameters on a cluster of patients give a virtual patient that supports the design and testing of control algorithms, improving the expectations and reducing the risk of experimental activities. PDD2 seems a very promising control strategy that can improve post-prandial behaviour in respect to standard therapy.