摘要：AbstractThis investigation presents a smart artificial pancreas (AP) for treating Type 1 Diabetes Mellitus (T1DM) in elderly which simultaneously automates the insulin administration but also diet recommender system using implicit carbohydrate (CHO) measurements. Three main components of the AP are: (i) long-term model of physiological dynamics, (ii) model predictive controller and (iii) a diet recommender algorithm which uses implicit CHO measurements. We first show that long-term dynamics are important for capturing the food influences on blood glucose levels and to maintain within admissive bands. The diet recommender algorithm fuses the insulin infusion information of the MPC, long-term model, average CHO and its variations to recommend diet and its pattern. The proposed artificial pancreas with diet recommender system is illustrated using studies conducted on elderly patients with T1DM based on clinical trials conducted at Jothydev’s Diabetes Research Centre, Trivandrum, India. Our studies shows that the proposed AP not only automates the insulin infusion but also provides a recommender system for diet.
