Search results

(1 - 2 of 2)
MODELING GLUCOSE-INSULIN DYNAMICS AND AUTOMATED BLOOD GLUCOSE REGULATION IN PATIENTS WITH TYPE 1 DIABETES
Title
MODELING GLUCOSE-INSULIN DYNAMICS AND AUTOMATED BLOOD GLUCOSE REGULATION IN PATIENTS WITH TYPE 1 DIABETES
Creator
Oruklu, Meriyan
Date
2012-11-06, 2012-12
Description
Estimation of future glucose concentrations is a crucial task for diabetes management. Predicted glucose values can be used for early...
Show moreEstimation of future glucose concentrations is a crucial task for diabetes management. Predicted glucose values can be used for early hypoglycemic/hyperglycemic alarms or for adjustment of insulin amount/rate. In the first part of this thesis, reliable subject-specific glucose concentration prediction models are developed using a patient’s continuous glucose monitoring (CGM) data. CGM technologies provide glucose readings at a high frequency and consequently detailed insight into a patient’s glucose variation. Time-series analyses are utilized to develop low-order linear models from a patient’s own CGM data. Glucose prediction models are integrated with recursive identification and change detection methods, which enable dynamical adaptation of the model to inter-/intra-subject variability and glycemic disturbances. Two separate patient data sets collected under hospitalized (disturbance-free) and normal daily life conditions are used to validate the univariate glucose prediction algorithm developed. Prediction performance is evaluated in terms of prediction error metrics and Clarke error grid analysis (CG-EGA). The long-term complications of diabetes can be reduced by controlling the blood glucose concentrations within normoglycemic limits. In the second part of this thesis, the subject-specific modeling algorithm developed in part one is integrated with a control algorithm for closing the glucose regulation loop for patients with type 1 diabetes. An adaptive control algorithm is developed to keep a patient’s glucose concentrations within normoglycemic range and dynamically respond to glycemic challenges with automated subcutaneous insulin infusion. A model-based control strategy is used to calculate the required insulin infusion rate, while the model parameters are recursively identified at each sampling step. The closed-loop algorithm is designed for the subcutaneous route for both glucose sensing and insulin delivery. xii It accounts for the slow insulin absorption from the adipose tissue and the time-delay between blood and subcutaneous glucose concentrations. The performance of the control algorithm developed is demonstrated on two simulated patient populations to provide effective blood glucose regulation in response to multiple meal challenges with a simultaneous challenge on a patient’s insulin sensitivity. Physical activity and emotional stimuli such as stress are known to have a significant effect on a patient’s whole-body fuel metabolism. In the third part of this thesis, the univariate time-series models developed from recent glucose concentration history are extended to include additional information on a patient’s physical and emotional condition. Physiological measurements from a multi-sensor body monitor are used to supplement a patient’s CGM data and develop multivariate glucose prediction models. The prediction performance of the multivariate algorithm developed is evaluated on data collected from patients with type 2 diabetes, and a real life implementation of the algorithm is demonstrated for early (i.e., 30 min in advance) hypoglycemia detection. Finally, the control algorithm developed in part two is extended to utilize the glucose profiles predicted by the multivariate patient model. The multivariate closedloop algorithm is tested with two clinical experiments performed on a patient with type 1 diabetes during a high intensity exercise followed by a carbohydrate-rich meal challenge. The algorithm acquires the patient’s CGM and armband (body monitor) data every 10 min, and accordingly calculates the required basal insulin infusion rate. Insulin is administered in a fully automated manner without any food or activity announcements (e.g., no information on meal/exercise size or time). None of the algorithms developed in this thesis require any patient specific tailoring or prior experimental data before implementation. They are also designed to function in a fully automated manner and do not require any disturbance announcexiii ments or manual inputs. Therefore, they are good candidates for installation on a portable ambulatory device used in a patient’s home environment for his/her diabetes management.
PH.D in Chemical and Biological Engineering, December 2012
Show less
MODELING GLUCOSE-INSULIN DYNAMICS AND AUTOMATED BLOOD GLUCOSE REGULATION IN PATIENTS WITH TYPE 1 DIABETES
Title
MODELING GLUCOSE-INSULIN DYNAMICS AND AUTOMATED BLOOD GLUCOSE REGULATION IN PATIENTS WITH TYPE 1 DIABETES
Creator
Oruklu, Meriyan
Date
2012-11-06, 2012-12
Description
Estimation of future glucose concentrations is a crucial task for diabetes management. Predicted glucose values can be used for early...
Show moreEstimation of future glucose concentrations is a crucial task for diabetes management. Predicted glucose values can be used for early hypoglycemic/hyperglycemic alarms or for adjustment of insulin amount/rate. In the first part of this thesis, reliable subject-specific glucose concentration prediction models are developed using a patient’s continuous glucose monitoring (CGM) data. CGM technologies provide glucose readings at a high frequency and consequently detailed insight into a patient’s glucose variation. Time-series analyses are utilized to develop low-order linear models from a patient’s own CGM data. Glucose prediction models are integrated with recursive identification and change detection methods, which enable dynamical adaptation of the model to inter-/intra-subject variability and glycemic disturbances. Two separate patient data sets collected under hospitalized (disturbance-free) and normal daily life conditions are used to validate the univariate glucose prediction algorithm developed. Prediction performance is evaluated in terms of prediction error metrics and Clarke error grid analysis (CG-EGA). The long-term complications of diabetes can be reduced by controlling the blood glucose concentrations within normoglycemic limits. In the second part of this thesis, the subject-specific modeling algorithm developed in part one is integrated with a control algorithm for closing the glucose regulation loop for patients with type 1 diabetes. An adaptive control algorithm is developed to keep a patient’s glucose concentrations within normoglycemic range and dynamically respond to glycemic challenges with automated subcutaneous insulin infusion. A model-based control strategy is used to calculate the required insulin infusion rate, while the model parameters are recursively identified at each sampling step. The closed-loop algorithm is designed for the subcutaneous route for both glucose sensing and insulin delivery. xii It accounts for the slow insulin absorption from the adipose tissue and the time-delay between blood and subcutaneous glucose concentrations. The performance of the control algorithm developed is demonstrated on two simulated patient populations to provide effective blood glucose regulation in response to multiple meal challenges with a simultaneous challenge on a patient’s insulin sensitivity. Physical activity and emotional stimuli such as stress are known to have a significant effect on a patient’s whole-body fuel metabolism. In the third part of this thesis, the univariate time-series models developed from recent glucose concentration history are extended to include additional information on a patient’s physical and emotional condition. Physiological measurements from a multi-sensor body monitor are used to supplement a patient’s CGM data and develop multivariate glucose prediction models. The prediction performance of the multivariate algorithm developed is evaluated on data collected from patients with type 2 diabetes, and a real life implementation of the algorithm is demonstrated for early (i.e., 30 min in advance) hypoglycemia detection. Finally, the control algorithm developed in part two is extended to utilize the glucose profiles predicted by the multivariate patient model. The multivariate closedloop algorithm is tested with two clinical experiments performed on a patient with type 1 diabetes during a high intensity exercise followed by a carbohydrate-rich meal challenge. The algorithm acquires the patient’s CGM and armband (body monitor) data every 10 min, and accordingly calculates the required basal insulin infusion rate. Insulin is administered in a fully automated manner without any food or activity announcements (e.g., no information on meal/exercise size or time). None of the algorithms developed in this thesis require any patient specific tailoring or prior experimental data before implementation. They are also designed to function in a fully automated manner and do not require any disturbance announcexiii ments or manual inputs. Therefore, they are good candidates for installation on a portable ambulatory device used in a patient’s home environment for his/her diabetes management.
PH.D in Chemical and Biological Engineering, December 2012
Show less