Determination of glucose in a biological matrix by multivariate analysis of multiple band-pass-filtered Fourier transform near-infrared interferograms.
Determination of glucose in a biological matrix by multivariate analysis of multiple band-pass-filtered Fourier transform near-infrared interferograms.
复制标题
DOI：
10.1021/ac9705529
复制
发表时间：
1997-11
影响因子：
7.4
通讯作者：
M. Mattu;Gary W. Small;Mark A. Arnold
中科院分区：
文献类型：
--
作者：
M. Mattu;Gary W. Small;Mark A. Arnold
描述了一种多变量校准方法，其中使用傅立叶变换近红外干涉图数据来确定牛血清白蛋白（BSA）和三醋精的水性基质中的葡萄糖的临床相关水平。BSA和三醋精分别用于模拟血液中的蛋白质和甘油三酯，并且以跨越正常人体生理范围的水平存在。构建全析因实验设计用于数据收集，葡萄糖为10个水平，BSA为4个水平，三醋精为4个水平。将高斯形带通数字滤波器应用于干涉图数据以提取与感兴趣的吸收带相关联的频率。将不同宽度的单独滤光片置于4400 cm-1处的葡萄糖带、4606 cm-1处的BSA带和4446 cm-1处的三醋精带上。每个滤波器应用于原始干涉图，根据所使用的滤波器的数量，产生一个、两个或三个滤波干涉图。这些滤波后的干涉图的片段一起用于偏最小二乘回归分析，以建立葡萄糖校准模型。最佳的校准模型是通过使用三个过滤器集中在葡萄糖，BSA和三醋精带过滤的干涉图的单独的片段来实现的。在1 - 20 mM葡萄糖的生理范围内，该17项模型的R2值、校准标准误差和预测标准误差分别为98.85%、0.631 mM和0.677 mM。这些结果是可比的光谱数据的常规分析中获得的那些。基于干涉图的方法无需使用单独的背景测量，并且仅采用干涉图的一小部分。
A multivariate calibration method is described in which Fourier transform near-infrared interferogram data are used to determine clinically relevant levels of glucose in an aqueous matrix of bovine serum albumin (BSA) and triacetin. BSA and triacetin are used to model the protein and triglycerides in blood, respectively, and are present in levels spanning the normal human physiological range. A full factorial experimental design is constructed for the data collection, with glucose at 10 levels, BSA at 4 levels, and triacetin at 4 levels. Gaussian-shaped band-pass digital filters are applied to the interferogram data to extract frequencies associated with an absorption band of interest. Separate filters of various widths are positioned on the glucose band at 4400 cm-1, the BSA band at 4606 cm-1, and the triacetin band at 4446 cm-1. Each filter is applied to the raw interferogram, producing one, two, or three filtered interferograms, depending on the number of filters used. Segments of these filtered interferograms are used together in a partial least-squares regression analysis to build glucose calibration models. The optimal calibration model is realized by use of separate segments of interferograms filtered with three filters centered on the glucose, BSA, and triacetin bands. Over the physiological range of 1-20 mM glucose, this 17-term model exhibits values of R2, standard error of calibration, and standard error of prediction of 98.85%, 0.631 mM, and 0.677 mM, respectively. These results are comparable to those obtained in a conventional analysis of spectral data. The interferogram-based method operates without the use of a separate background measurement and employs only a short section of the interferogram.