Zhi-Liang Fan

Ziziphi Spinosae Semen (ZSS) is a traditional Chinese medicinal material used for both medicinal and dietary purposes, but substitution and adulteration can compromise its authenticity and quality. Ten batches of authentic ZSS and one source batch of each of five adulterant materials were used to prepare 325 laboratory samples. This study aimed to develop a rapid method for ZSS identification and adulteration quantification using Fourier transform near-infrared spectroscopy (FT-NIR). Spectral distributions of six sample classes were analyzed using PCA-X, PLS-DA, and OPLS-DA, and the classification performance of SVM, DA, ANN, and KNN was compared. DD-SIMCA was used for one-class authentication, while PLSR was applied to quantify five adulterants using 15 spectral processing approaches. In the internal test set, SVM achieved 98.0% accuracy and 98.4% balanced accuracy. Within the laboratory samples studied, DD-SIMCA achieved 100% accuracy, sensitivity, and specificity. Across the five selected PLSR models, prediction-set R² ranged from 0.9272 to 0.9940 and RMSEP from 1.36 to 4.33 percentage points of adulterant mass fraction. These results demonstrate the feasibility of combining FT-NIR with chemometric methods for rapid ZSS classification, authenticity assessment, and adulteration-level prediction. Independent validation with additional source batches is required before practical application.

Dendrobium officinale is a valuable medicinal and edible homologous health food. It has immunomodulatory, antioxidant, and metabolism-regulating properties. However, its adulteration is widespread, seriously compromising product quality and safety. Traditional adulteration detection methods are complex, costly, and time-consuming, making it urgent to establish a rapid and non-destructive detection approach. This study developed a rapid identification and quantification method for adulterated D. officinale. The method combined near-infrared (NIR) spectroscopy with data-driven soft independent modeling of class analogy (DD-SIMCA) and partial least squares regression (PLSR) models. PCA, PLS-DA, and OPLS-DA were first used to visualize sample clustering and group differences. DT, SVM, ANN, and NB were used for classification. DD-SIMCA and PLSR were used for one-class modeling and quantitative analysis. Raw spectral data were preprocessed using multiplicative scatter correction (MSC), the standard normal variate (SNV), the first derivative, and Savitzky–Golay smoothing. In the identification analysis, the DD-SIMCA model achieved 100% sensitivity and 100% specificity in the validation set. Its overall accuracy in the independent test set was 99.2%, demonstrating excellent discrimination performance. In addition, SVM combined with NIR also achieved good accuracy. In the quantitative analysis of adulteration, the PLSR model predicted different adulteration levels. Most calibration and validation sets showed R² values above 0.99 and RMSE values below 0.05, indicating excellent predictive performance. The results indicate that NIR combined with DD-SIMCA and PLSR can achieve rapid identification and accurate quantification of adulterated D. officinale samples. This approach provides strong support for quality control and regulatory supervision of high-value health foods.