近红外漫反射光谱在人体血糖无创检测中的应用

Application of NIR Diffusion Reflectance Spectrum Technology in the Noninvasive Measurement for Human Blood Glucose

采用近红外漫反射光谱技术对人体血糖进行了无创检测。实验使用Nexus-870傅立叶红外光谱仪及其光纤附件,采集了 6名不同年龄健康志愿受试者手腕处的近红外漫反射光谱。对光谱进行了平滑、基线校正和二次求导预处理,采用偏最小二乘 (PLS)方法在含有葡萄糖吸收峰的 7500~8500cm-1波段建立同一个体、相同年龄段的不同个体、以及不同年龄段的不同个体的校正模型。采集漫反射光谱的同时抽适量的血样在 752型紫外光栅分光光度计上标定血糖的实际值,并对校正模型计算值和实际标定值进行了比较,结果表明个体建模的相关性很好,相关系数达到 0. 99980,均方差在≤0. 346,误差分布在±0. 8mmol/l之间。对部分不参与建模的样品进行了预测,结果表明个体建模的自我预测结果好于该模型对其它个体样品的预测结果,预测误差≤0. 89544mmol/l。

The noninvasive measurement of human blood glucose is achieved with NIR diffusion reflectance spectrum technology.The wrist NIR diffusion reflectance spectrum of six healthy volunteers is collected using Nexus-870 and its NIR Fiberport Smart Accessory.The test is implemented with the changing of the blood glucose concentration.The calibration model is set up using PLS method with the smoothing,baseline correct and second derivatives pretreatment spectrum in 7500~8500cm^(-1) region for single volunteer and the combination of them.When the spectrum is got,the actual blood glucose value of every spectrum sample is demarcated using ultraviolet spectrophotometer.The relativity between the calibration value and the actual one for single volunteer is better than that of the combination,and the correlative coefficient is up to 0.99980.RMSEC are less than 0.364,and the error is between-0.8~0.8mmol/l.Some samples that are chosen optionally don’t participate in setting up calibration model,and they are predicted by themselves model and other’s model respectively.The result of self-model prediction is better than that of other (volunteer’s) model,and the error of the prediction is less than≤0.89544mmol/l.