Real-time detection of Glucose Sensor and Insulin Pump Faults in an Artificial Pancreas.

Abstract

Sensors for real-time continuous glucose monitoring (CGM) and pumps for continuous subcutaneous insulin infusion (CSII) have opened new scenarios for Type 1 diabetes treatment, including the development of an Artificial Pancreas, a minimally invasive device for automated glycemic control via insulin infusion modulation. However, permanent or temporary faults of these two components, such as compression artifacts and pump-catheter occlusions, may expose diabetic patients to severe risks and strongly affect the efficacy of an automated treatment. In Facchinetti et al. [2013], a fault detection method was proposed, simultaneously exploiting CGM and CSII data streams and individualized models of glucose-insulin interaction. The method was assessed during night-time, a simple yet practically relevant case study where meals do not perturb glucose-insulin dynamics. In this contribution we extend the method from nighttime to whole day, facing the challenge of the meals and taking advantage of meal information commonly provided by the patient to the system. To this aim, the patient-specific model identified includes now meal as a further input of the system. The efficacy of the method is tested using the UVA/Padova Type 1 diabetic patient simulator.