The Phantom Experiment And Spectral Correction Method Of Positional Floating Reference For Non-Invasive Blood Glucose Sensing With Near-Infrared Spectroscopy

Non-invasive blood glucose sensing with near-infrared (NIR) spectroscopy is an international frontier topics. The background interference from human body that affect the accuracy and stability of NIR glucose monitoring has become an important adverse factor. Aiming at this problem, the thesis focuses on simulation analysis, theory derivation and experimental verification around the positional floating reference method.Based on Monte Carlo simulation method, the features of the positional floating reference point of glucose is studies in three-layer skin model in this paper, and the results confirm that the positional floating reference point in the model does exist. Focused on the optical properties and skin structure, the impact of these parameters to the positional floating reference point is described, and the variation range of the floating reference point is given. The result could provide the foundation to the probe design in vivo.Four kinds of interferences in the NIR glucose sensing are discussed. And the typical data processing methods are introduced for spectral correction to eliminate the interference noise. The modeling methods for NIR glucose sensing are present. Based on these, the theoretical derivation of the reference spectral correction method is implemented, and it is proved theoretically that the method can eliminate the background noise effectively in vivo.According to the results of Monte Carlo simulation, a custom-built fiber probe with two collected rings is designed for a NIR sening system. A experiment is implemented in 5% Intralipid glucose solutions using the sensing system. Based on the collected reference spectrum and measurement spectrum, the reference spectral correction method is investigated. Compared with the measurement spectrum model, the root square error of prediction (RMSEP) of the model by the reference spectrum correlation method is decreased by 35%, averagely. The results verify the positional floating reference method can eliminate the background noise significantly.