Rapid and Quantitative Detection Method for Acteoside during Chromatographic Purification of Adhesive Rehmannia Leaf Extract Using near Infrared Spectroscopy and Chemometrics

The feasibility of rapid and non-destructive analysis of acteoside, the main active component of adhesive rehmannia leaf, in the macroporous resin adsorption and elution process by Fourier transform near infrared spectroscopy together with different chemometrics methods was investigated in this study. The usable spectral region (5500–6200 cm⁻¹) was identified, spectral preprocessing, including Savitzky–Golay smoothing derivative was employed, and spectral dimension was also reduced through principal component analysis for artificial neural networks (ANN) and least square support vector machines (LS-SVM) methods. The multivariate calibration models based on principal component regression, partial least-squares regression, ANN, and particle swarm optimisation-based LS-SVM were developed to correlate the pretreated spectral data and the corresponding acteoside concentrations determined by high performance liquid chromatography. For all regression models, the coefficients of determination for all data sets (calibration, validation and prediction set) were higher than 0.988, indicating that the four methods were effective for building acteoside calibration models using samples taken from either the adsorption or the elution process. However, according to the relative standard error of prediction values for the prediction set samples, the LS-SVM model was slightly more accurate than the models obtained using the other regression techniques.