Meal Detection in Patients With Type 1 Diabetes: A New Module for the Multivariable Adaptive Artificial Pancreas Control System

Abstract

A novel meal-detection algorithm is developed based on continuous glucose measurements. Bergman’s minimal model is modified and used in an unscented Kalman filter for state estimations. The estimated rate of appearance of glucose is used for meal detection. Data from nine subjects are used to assess the performance of the algorithm. The results indicate that the proposed algorithm works successfully with high accuracy. The average change in glucose levels between the meals and the detection points is 16(±9.42) [mg/dl] for 61 successfully detected meals and snacks. The algorithm is developed as a new module of an integrated multivariable adaptive artificial pancreas control system. Meal detection with the proposed method is used to administer insulin boluses and prevent most of post-prandial hyperglycemia without any manual meal announcements. A novel meal bolus calculation method is proposed and tested with the UVA/Padova simulator. The results indicate significant reduction in hyperglycemia.

Meal detection based on non-individualized moving horizon estimation and classification

“…The current estimate at each time step corresponds to the last value in the horizon. In a study using UKF, meals of various glucose content are detected if the estimated glucose rate of appearance exceeded 2 mg/dL/min [8]. This threshold is used after adjusting the units to 2 mg/dL/min · V g = 3.4 mg/kg/min.…”

“…(8) This corresponds to an insulin sensitivity that is changing from its nominal value by ±50% within 6 hours, which appears realistic [19].…”

“…The KF requires linear models whereas extended (EKF) and unscented Kalman filters (UKF) are designed for the use of non-linear models which may even describe the glucose dynamics in the human body. The UKF using an extended version of Bergman's minimal model [7] has been proposed for meal detection where not the estimated glucose concentration but rather the estimated glucose rate of appearance in plasma is checked against a threshold [8]. Defining the thresholds is in any case a trade-off between sensitivity and specificity.…”

Embedded Control in Wearable Medical Devices: Application to the Artificial Pancreas

“…A health monitoring system is introduced in [44] as a safety layer in the AP device system. A meal detection algorithm either based on the calculation of the glucose rate of change [45,46] or model based [47] is proposed to prevent postprandial hyperglycemia for the case of unannounced meals. A low-glucose suspend system includes a threshold-suspend feature of sensor-augmented insulin pumps that forces the pump to shut off when CGM values drop below a certain threshold [48].…”

Simultaneous model predictive control and moving horizon estimation for blood glucose regulation in Type 1 diabetes

“…For example, it would be useful or even necessary to consider meal announcement or detection for clinical trials; however, even when meal announcement or detection is available, a controller must be able to handle the more difficult case of unannounced meals, and meal boluses and other feed-forward boluses hide or prevent the action of the control law, making it difficult (or irrelevant) to contrast different control laws. We leave the preparation of this new control and estimation method for clinical deployment (including the incorporation of such techniques as meal detection 30 ) as a future work. It is thus important to clarify that the simulation results presented here (for MPC/LO and for MPC/MHE) are not a representative of expected results in trials since these would require the additional abovementioned safety features.…”