Species discrimination and VIP-stacking quantitative models for Curcumae Rhizoma utilizing multi-modal spectra...

Area of Science:

• Herbal Medicine Analysis
• Chemometrics and Spectroscopy
• Pharmaceutical Quality Control

Background:

• Curcumae Rhizoma (Ezhu) faces challenges in species differentiation and accurate minor component analysis.
• Traditional methods for Ezhu analysis are time-consuming and cumbersome.
• Spectroscopic techniques combined with chemometrics offer a promising alternative for Ezhu quality control.

Purpose of the Study:

• To develop accurate qualitative and quantitative models for Ezhu using multi-modal spectroscopy and chemometrics.
• To establish robust species discrimination models for Ezhu.
• To create advanced models for predicting the content of key minor constituents in Ezhu.

Main Methods:

• Multi-modal spectroscopy including Fourier transform infrared (FT-IR), Fourier transform near-infrared (FT-NIR), and ultraviolet (UV) was employed.
• Data fusion of UV+FT-NIR+FT-IR spectral data was used for qualitative analysis with LDA, KNN, and DT models.
• A novel variable importance in projection (VIP)-guided stacking ensemble strategy was developed for quantitative analysis.

Main Results:

• Qualitative models achieved 100% classification accuracy for species discrimination.
• The VIP-stacking ensemble strategy demonstrated superior predictive accuracy and robustness for minor constituent content prediction compared to conventional methods.
• Accurate prediction models were successfully constructed for seven specific minor constituents in Ezhu.

Conclusions:

• Spectral data fusion is highly effective for both qualitative and quantitative analysis of Ezhu.
• The VIP-stacking ensemble strategy significantly enhances the performance of content prediction models.
• This approach provides an efficient method for Ezhu identification and quality control, applicable to pharmaceutical, agricultural, and food science.