A classification-based fault detection method for Continuous glucose monitoring (CGM)

Abstract

Diabetes is a group of metabolic diseases which is harm to human health seriously. Continuous glucose monitoring (CGM) plays an important role in the treatment of diabetes. However, false information of glucose, caused by the faults of CGM, will reduce the treatment effect, and even lead diabetic patients to severe risks. So far, there are some notable attempts in the research of CGM fault detection. But the common fault detection methods of CGM have low accuracy due to the complexity of glucose fluctuation, including the change of correlation, nonlinearity and so on. In this paper, we propose an algorithm based on iterative PCA, to classify the glucose, ensuring that the glucose data with the same class are linear and have similar correlation. In addition, a PCA-based monitoring chart with naive Bayesian classifier is used to detect the faults of CGM on line. The simulation results show that the sensitivity of classified fault detection method is greater than the method without classification, which means that the classified fault detection method has a promising future in the field of CGM.