Noninvasive blood glucose monitor via multi-sensor fusion and its clinical evaluation

Abstract Blood glucose monitoring is vital for controlling the complications of diabetes. We propose a noninvasive blood glucose monitor featuring multi-sensor fusion, which functions well for rapid yet accurate blood glucose monitoring and screening of diabetes. The characteristic parameters of diabetics were extracted by applying one-factor analysis of variance, and a classification model was established via principal components analysis and discriminant analysis to identify whether the subjects suffer from diabetes. In light of the challenges for noninvasive blood glucose monitoring modelling brought by different pathogeneses, disease processes and complicated complications of different diabetics, the subjects were classified into different categories via unsupervised K-means clustering algorithm and modelled for each category separately for expanded coverage and improved accuracy. The accuracy was clinically evaluated in 254 volunteers with type II diabetes. The results in the Parkes error grid were as follows: 58.33% in Zone A, 39.43% in Zone B and 2.24% in Zone C, with a correlation coefficient of 0.69 and a root mean squared error of 2.67 mmol/L. In 2019, the monitor was certified by the National Medical Products Administration of China as the first Class III Medical Device.