Diabetes is a chronic disease posing significant threats to global public health, where precision blood glucose monitoring serves as a cornerstone of effective disease management. While invasive blood sampling techniques remain prevalent in clinical practice, their inherent drawbacks—including patient discomfort and potential infection risks— have positioned non-invasive glucose detection as a major focus of scientific research. This study is based on near-infrared spectroscopy technology (780- 2500nm) to systematically analyze the characteristic absorption of hydrogen-containing functional groups, such as C-H and O-H in glucose molecules. The detection wavelength selection (1500-1800nm) and detection site (fingertip) were optimized, and the applicability of diffuse reflection and transmission detection modes was compared. In response to the complex and easily interfered characteristics of near-infrared spectral signals, an innovative BP neural network algorithm is introduced to construct a prediction model. By utilizing its powerful nonlinear mapping ability and adaptive learning characteristics, the accuracy of blood glucose concentration prediction is effectively improved. The research results indicate that this method has the advantages of noninvasiveness, convenience, and low cost, but it also faces technical challenges such as individual differences and environmental interference.
J Wang (Thu,) studied this question.