Theoretical And Experimental Study On Effective Extraction Of Glucose Specific Signal In NIR Spectroscopy

Near-infrared noninvasive human blood glucose detecting has increasingly become an interesting and challenging field in the area of biomedical engineering. However, the detecting precision at the clinical level has not yet been achieved. The critical obstacle lies in how to extract the glucose-specific signal that is unfortunately swamped within a complex physiological system. Thus, this paper holds a theoretical and experimental study on effective extraction of glucose specific signal in NIR spectroscopy.In this paper, by analyzing the extraction mechanism of glucose concentration signal, the background deducting and the optimum background selection criterion are proposed to increase the specificity of glucose-specific signal. Theoretical deduction demonstrates that the optimum background should bear similar optical properties with the sample solution. The corresponding in-vitro validating experiment results show that the predicting precision in glucose solution, plasma and 2% intralipid has improved by 17%, 25.9% and 40%, respectively, when the optimum background is used. The identical background selection criterion is applied into the in-vivo animal trial. When baseline spectra are used as background, the predicting precision toward glucose concentration has increased by 58.26%. This satisfactory result can be used as an important reference for the human in-vivo study.The reference wavelength based glucose-specific signal extraction is studied in this paper. According to the water displacement effect upon dissolution of glucose, the existence of the reference wavelength under which the absorption coefficient is insensitive to the changes of glucose concentration is put forward theoretically. At this wavelength, the small fluctuation of the transmitted light intensity can be regarded totally as the consequence of background noises. Thus, it can be used to eliminate the noises in the sample spectra. Both the theoretical and laboratorial results show that the reference wavelength appears at 1525nm. Based on this, an effective method for extracting the glucose-specific signal is proposed and the corresponding validating experiments are constructed in complicated non-scattering samples. Using the reference wavelength method, the predicting precision in albumin and plasma sample has increased by 50% and 43.3%, respectively. The experimental results indicate that the reference wavelength method can effectively eliminate the influence of various noises on the glucose-specific signal extraction, and thus can remarkably improve the measuring precision in noninvasive near-infrared glucose detecting.