Denoising of Continuous Glucose Monitoring Signal with Adaptive SG Filter

Abstract

This paper aims at simulating an intelligent Savitzky–Golay (SG) filter which can automatically adjust its parameters in accordance with the changes in sampling or cut-off frequency. The filter is implemented on signals obtained from Continuous Glucose Monitoring (CGM) device. The Continuous Glucose Monitoring devices can indicate any life-threatening (hyperglycemic or hypoglycemic) event, so that a preventive action may be taken to control blood glucose. The accuracy of these devices is generally not very good due to factors like sensor electronics and movement artefacts. In the present research work, data obtained from GlucoSim simulator is used, which is an educational software to simulate glucose and insulin levels in blood and their dynamics in healthy or diabetic (Type-1) individuals. Genetic Algorithm and Particle Swarm Optimization techniques are used to tune the parameters of Savitzky–Golay filter leading to GA-SGF and PSO-SGF filters. Continuous Glucose Monitoring signal is denoised using the adaptive Savitzky–Golay filters. It is observed from the results that PSO-SGF provides fast and efficient filtering.