Noninvasive blood glucose detection system based on dynamic spectrum and “M+N″ theory

Abstract

Noninvasive detection of blood glucose is an urgent problem to be solved in the prevention, treatment and management of diabetes. However, in the system of noninvasive detection of blood glucose content by spectral method, due to individual differences, interference in the measurement system and the fact that the content of blood glucose is three orders of magnitude smaller than the most abundant component in blood, the spectral signal of blood glucose component is weak, and the problem has not been solved so far. Dynamic spectrum (DS) method can theoretically eliminate the influence of individual differences and changes in measurement conditions, and “M + N″ theory comprehensively considers the internal and external factors of spectral quantitative analysis of complex solution components, and provides some strategies to suppress the errors caused by multiple factors in spectral quantitative analysis of complex solutions. Therefore, this paper proposes a system for non-invasive quantitative analysis of human blood glucose based on DS and “M + N″ theory. In this paper, feasible methods are used in each step of the system to effectively reduce the influence of noise and interference factors, so that the model has good prediction performance. The correlation coefficient and root mean square error of calibration set were 0.9539 and 0.3965 (mmol/L), respectively. The correlation coefficient and root mean square error of prediction set were 0.9542 and 0.7305 (mmol/L), respectively. The system of noninvasive quantitative analysis of blood glucose proposed in this paper is effective and feasible, which not only makes great progress in noninvasive detection of human blood glucose, but also has application value for realizing noninvasive real-time detection of human blood components.