Noise Reduction in Continuous Blood Glucose Sensor using Physiology based Kalman Filter for Artificial Pancreatic System

Abstract

Continuous Glucose Monitoring (CGM) is an essential module of Artificial Pancreatic system. The efficiency of this system in maintaining the blood sugar level is of high significance for glucose control in T1DM patients. often, the sensor data is corrupted by random noises such as zero-mean white Gaussian noise (WGN). Conventional noise rejection techniques such as FIR and IIR digital filter, Moving Average (MA) filter are discussed in the literature. More recently, Kalman Filtering (KF) has been broadly studied for glucose sensor measurement. However, the glucose-insulin (GI) dynamics were modelled as double integrated random walk model rather than a physiological model in noise filter. Moreover, the relation of noise reduction scheme on controller action is not reported in the literature. Therefore, this work proposes a physiology model-based noise filter scheme for glucose sensor noise reduction. Moreover, the effects of noise reduction system on PID controller action is also studied here. The simulation results show that the suggested physiology based KF showed improved performance in noise reduction and also leads to effective glucose control.