Capteur de glucose pour un pancréas artificiel : quel défi ?

Continuous glucose monitoring is a mandatory component of an artificial pancreas performing closed-loop insulin therapy. Glucose sensor data provides the basic information for the computing by the control algorithm of the needed rate of insulin infusion from the pump in order to reach the objective, either a blood glucose level target, or a predefined safe glucose range. Hence, the emitted sensor signal that is received by the control algorithm is of utmost importance because it will drive the command of insulin delivery from the pump. The improvement of structural and functional components of glucose sensors during the last decade has allowed consistent gains in precision and accuracy of commonly used sensors, i.e. of enzymatic type and located in subcutaneous tissue. When these sensors are connected in a closed-loop system, the key-elements will be the analysis and the treatment of sensor signal in order to reduce variable noise, improve calibration against blood glucose and anticipate by prediction blood glucose variations. The identification of events that impair the signal becomes even more important with the evolution toward closed-loop use in free-life conditions. Two of them require a specific attention: signal interruption, and artifacts related to sensor compression.