温度变化是影响近红外无创血糖测量精度的主要因素之一。 为降低温度变化对近红外漫反射光谱的影响, 提出了一种基于温度不敏感源-探测器距离的测量方法, 即在漫反射光强对人体组织温度变化不敏感的源-探测器距离处进行光谱测量。 利用Monte Carlo方法模拟了温度为30～40 ℃、 葡萄糖浓度为0～300 mmol·L-1的皮肤组织在多个源-探测器距离处的漫反射光强。 根据模拟结果, 分析了人体皮肤组织模型中温度不敏感源-探测器距离的存在性及其受葡萄糖浓度变化的影响; 比较了1 000 nm处温度恒定和温度变动时, 不同源-探测器距离处漫反射光强与葡萄糖浓度的相关性; 进一步地, 利用六个波长(1 000, 1 050, 1 100, 1 150, 1 350和1 410 nm)下的温度不敏感源-探测器距离及其他距离处的漫反射光强, 建立了葡萄糖的偏最小二乘(PLS)模型, 并比较了这些模型在温度恒定和温度变动时的预测精度。 结果表明, 在1 000～1 440 nm范围内, 人体存在温度不敏感源-探测器距离, 且葡萄糖浓度变化对该距离的影响可以忽略不计; 当组织温度变化时, 温度不敏感源-探测器距离处的漫反射光强与葡萄糖浓度的相关性及建模效果均明显优于其他源-探测器距离, 基本接近样品温度恒定时的情况。 研究表明, 基于温度不敏感源-探测器距离的测量方法能有效降低温度变化对漫反射光强的影响, 有望提高近红外漫反射无创血糖测量的精度。

The variation of temperature is one of the main interference factors that affect the accuracy of near-infrared (NIR) non-invasive blood glucose sensing. In order to reduce the influence of temperature variation on NIR diffuse reflectance, a measurement method based on temperature insensitive source-detector separation was proposed in this paper, i. e., the temperature insensitive source-detector separation, whose diffuse reflectance was insensitive to temperature variations, was chosen to obtain the spectrum. The Monte Carlo method was used to simulate the diffuse reflectance at multiple source-detector separations with temperatures of 30～40 ℃ and glucose concentrations of 0～300 mmol·L-1. According to the simulation results, the existence of temperature insensitive source-detector separation in human skin tissue model and the influence of glucose concentration on the separation were analyzed. At wavelength of 1 000 nm, the correlation between the diffuse reflectance and glucose concentration at different source-detector separations under random temperature and constant temperature were compared. Furthermore, the partial least square (PLS) model of glucose were established using the diffuse reflectance at temperature insensitive source-detector separations and other separations under six wavelengths (1 000, 1 050, 1 100, 1 150, 1 350 and 1 410 nm), and the prediction accuracy of the model under random temperature and constant temperature were compared. The results showed that, in the range of 1 000～1 440 nm, there were temperature insensitive source-detector separations in human tissue, and the influence of glucose concentration variationon it can be ignored. When the temperature of the sample variates, both the correlation between glucose concentrations and diffuse reflectance and the prediction accuracy of model obtained at the temperature insensitive source-detector separation were significantly better than that of other separations, which was almost close to the situation of constant temperature. Therefore, the measurement method based on the temperature insensitive source-detector separation can effectively reduce the influence of temperature variations on NIR diffuse reflectance, and it is expected to improve the accuracy of NIR diffuse reflectance in non-invasive blood glucose sensing.