A model-based approach to synthesizing insulin infusion pump usage parameters for diabetic patients

We present a model-based approach to synthesizing insulin infusion pump usage parameters against varying meal scenarios and physiological conditions. Insulin infusion pumps are commonly used by type-1 diabetic patients to control their blood glucose levels. The amounts of insulin to be infused are calculated based on parameters such as insulin-to-carbohydrate ratios and correction factors that need to be calibrated carefully for each patient. Frequent and careful calibration of these parameters is essential for avoiding complications such as hypoglycemia and hyperglycemia. In this paper, we propose to synthesize optimal parameters for meal bolus calculation starting from models of the patient's insulin-glucose regulatory system and the infusion pump. Various off-the-shelf global optimization techniques are used to search for parameter values that minimize a penalty function defined over the predicted glucose sensor readings. The penalty function 'rewards' glucose levels that lie within the prescribed ranges and 'penalizes' the occurrence of hypoglycemia and hyperglycemia. We evaluate our approach using a model of the insulin-glucose regulatory system proposed by Dalla Man et al. Using this model, we compare various strategies for optimizing pump usage parameters for a virtual population of in-silico patients.