Reagent-free determination of multiple analytes is an active and promising field of research in clinical analysis. In this work, the determination of glucose and cholesterol in whole blood using Fourier transform infrared (FTIR) spectroscopy equipped with an attenuated total reflectance (ATR) accessory was performed. A comprehensive sample selection rule in multi space based on SPXY was proposed, termed C-SPXY. The core idea is to make full use of different derivative spectra space to construct the calibration set which preserves the more effective information. On this basis, a partial least squares (PLS) regression fusion modeling method was also presented aiming at improving prediction accuracy of glucose and cholesterol concentration in whole blood samples. Compared with other methods based on single spectra space, the proposed fusion model based on multi spectra space C-SPXY method provides smaller RMSEP values. Experimental results demonstrate that the proposed method and model provides superior predictive power and holds a good application prospect in the field of clinical analysis.
Keywords: Comprehensive sample selection; Derivative spectra; Fourier transform infrared (FTIR) spectroscopy; Fusion modeling; Whole blood.
Copyright © 2017 Elsevier B.V. All rights reserved.