Study On Theories And Experiments Of The Background Correction For Non-Invasive Blood Glucose Sensing With Near-Infrared Spectroscopy

Noninvasive measurement of human blood glucose concentration with near-infrared spectroscopy has attracted worldwide interest and is of great practical importance. However, a valid noninvasive method hasn't come into clinical realization yet. The main obstacle lies in the fact that the weak glucose specific signal is unfortunately covered by the complicated and variational physiological background. In this thesis, background correction methods are studied in order to eliminate the interference caused by the background variations, and then, further improve the validity of NIR noninvasive blood glucose concentration sensing technique.The feasibility of using skin diffusion spectroscopy to detect blood glucose concentration is analyzed theoretically and validated experimentally. Through Monte Carlo stimulation, it has been demonstrated that the measurement precision can reach 5.2mg/dl based on the custom-built NIR spectroscopy. Simultaneously the spectra of rabbit skin and vessel are collected for the first time in the animal experiments. High correlation between skin and vessel diffusion spectra is observed, which further demonstrates the feasibility of the NIR noninvasive blood glucose sensing .The background correction methods are studied. The distribution of main background noises is analyzed from two aspects: instruments and measuring conditions. To eliminate the instrumental drift, the background correction method, called"Sandwich", is adopted. The experimental result shows that the predicting precision toward glucose concentration has been improved by 44.7% in the glucose solution. To avoid the influence from the variation of sample characteristics, the background correction method by subtracting the nearest reference spectrum can effectively improve the measuring precision in the in vitro experiments. The predicting precision has improved by 25.9% and 40.1% respectively in the glucose aqueous solution and the Intralipid-2% solution.A novel background correction method based on reference wavelength is proposed. Since the intensity of diffusion light is insensitive to the glucose concentration variation at the reference wavelength, it has the potential to be used as the internal reference to eliminate the complicated physiological background variation in the in vivo human blood glucose concentration measurement. The existence of the reference wavelength has been proved in the absorptive and turbid media. Then, the background correction formula related to reference wavelength method is deduced and applied into the in vitro experiments. The predicting precision has improved 50.3% and 40.3% in the albumin sample and Intralipd-2% solution respectively.A projection to background space based background correction method based on the projection in background space is brought forward. Those wavelengths, which are sensitive to the variation of the background signals, are stretched to form a background space, and the analytical objective is projected on the background space, which can decreased the influence of noise and extract useful glucose signal. The characteristics of wavelength were analyzed from the aspects of the instrumental hardware and the measuring condition based on the two-dimensional correlation theory in non-invasive blood glucose sensing with near-infrared diffusion spectroscopy. And then the background space and corresponding arithmetic are defined with the theory of the net analyte signal. The prediction precision of glucose concentration has improved 25.8% and 26.6% respectively in the glucose aqueous solution and the Intralipid-2% solution used this projection arithmetic.