INTELLIGENT ARTIFICIAL PANCREAS INCORPORATED WITH MEAL DETECTION SYSTEM AND RECOMMENDER SYSTEM FOR TYPE-1 DIABETES SELF-MANAGEMENT

Open Access
- Author:
- Xie, Jinyu
- Graduate Program:
- Mechanical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- December 12, 2016
- Committee Members:
- Qian Wang, Dissertation Advisor/Co-Advisor
Qian Wang, Committee Chair/Co-Chair
Alok Sinha, Committee Member
Bo Cheng, Committee Member
Peter Molenaar, Outside Member - Keywords:
- Diabetes
Type 1
Meal Detection
Estimation
Recommender System
Adaptive Control
System Identification
Insulin
Exercise
Kalman Filter - Abstract:
- Type 1 Diabetes is a metabolic disease characterized by hyperglycemia resulting from defects in insulin secretion. In absence of a cure, insulin therapy is necessary for Type-1 Diabetes patients. Current insulin pump therapy is an open-loop insulin delivery system that yields to burdensome self monitoring and management process with suboptimal glucose control and high risk of fatal hypoglycemia events. In this dissertation, a closed-loop glucose sensor augmented insulin delivery system (Artificial Pancreas) is proposed. The novelty of the proposed Artificial Pancreas system lies in 1. It models both the meal effect and physical activity effect on glucose dynamics, and learns the patient parameters online starting with an initial guess; 2. It leverages rich physiological information (meal announcements and heart rates indicating physical activities) to make comprehensive decisions on the insulin dosage; 3. It detects uninformed meals and estimates the meal sizes, according to which controller delivers insulin boluses to compensate the meal-induced glucose excursion; 4. It serves as an expert system providing patient-specific dietary (meal/snack sizes) and exercise (target heart rates in exercise) recommendations to minimize the clinical risks. The performance of the proposed system is evaluated both in a Food & Drug Administration (FDA) approved simulator (in place of animal experiments) and a real patient data set collected under free-living conditions (offline evaluation).