Abstract
Motivated by the wide use of Scutellariae Radix (SR) in the food and pharmaceutical
industries, a rapid and non-destructive near-infrared spectroscopy (NIRS) method was
developed for the simultaneous analysis of three main active components in raw SR
and SR processed by stir-frying with wine. From seven geographical areas, 58 samples
were collected. The reference contents for the SR components baicalin, baicalein,
and wogonin were determined by high-performance liquid chromatography. Two multivariate
analysis methods, partial least-squares (PLS) regression as a linear regression method
and artificial neural networks (ANN) as a nonlinear regression method, were applied
to the NIR data, and their results were compared. In the PLS model, different model
parameters (i.e., 11 spectral pre-treatment methods), spectral region, and latent
variables were investigated to optimize the calibration model; additionally, the ANN
model was applied with five different spectral pre-treatment methods and six algorithms.
For the optimal model parameters, the correlation coefficients of the calibration
set for baicalin, baicalein, and wogonin were 0.9979, 0.9786, and 0.9773, respectively;
the correlation coefficients of the prediction set were 0.9756, 0.9843, and 0.9592,
respectively; the root mean square error of validation values were 0.215, 0.321, and
0.174, respectively. The optimal NIR models were then employed to analyze the effects
of processing and geographical regions on analyte contents. The established NIR methods
were robust, accurate, and reproducible. NIRS may be a promising approach for the
routine screening and quality control of traditional Chinese medicines.
Key words near-infrared spectroscopy
-
Scutellaria baicalensis
- Lamiaceae - partial least squares - artificial neural networks - quality control