Enhanced Accuracy of Near-Infrared Spectroscopy for Traditional Chinese Medicine with Competitive Adaptive Reweighted Sampling

Abstract

The accuracy of near-infrared quantitative models for traditional Chinese medicine is restricted by the matrix, low-concentration markers, and overlapping spectral bands. Competitive adaptive reweighted sampling, a recently developed algorithm, selects an optimal combination of multicomponent spectral data. In this article, the performance of this method was evaluated through the analysis of traditional Chinese medicine. The near-infrared spectra of the pharmaceutics were obtained and the concentration of puerarin was determined. After optimization of spectral pretreatment methods, competitive adaptive reweighted sampling was performed in each dataset to select key wavenumbers. Sixty-eight, thirty, and eight variables were selected for the raw material, intermediate product, and final product, respectively. Partial least squares models were constructed based on the selected variables. Enhanced accuracy was obtained using the competitive adaptive reweighted sampling-coupled models. The results indicated that competitive adaptive reweighted sampling improved the accuracy and simplified calibration while offering a new approach for the rapid and nondestructive analysis of traditional Chinese medicine.