Related Experiment Video
Updated: Jan 10, 2026
Author Spotlight: Integrating 2D-HPLC-MS and Molecular Networking in Natural Medicine Analysis
Published on: December 8, 2023
Identification of Pueraria lobata origin using terahertz precision spectroscopy and CNN-transformer hybrid network
Yongni Shao1, Dan Chen2, Binggan Wang3
1Terahertz Technology Innovation Research Institute, Terahertz Spectrum and Imaging Technology Cooperative Innovation Center, Shanghai Key Lab of Modern Optical System, University of Shanghai for Science and Technology, Shanghai 200093, China; Shanghai Institute of Intelligent Science and Technology, Tongji University, Shanghai 200092, China.
Terahertz (THz) spectroscopy combined with a deep learning hybrid model accurately authenticates the geographical origin of Pueraria lobata. This non-destructive method enhances traditional medicine quality control and traceability.
Area of Science:
- Analytical Chemistry
- Spectroscopy
- Chemometrics
Background:
- Pueraria lobata is a traditional Chinese medicine with varying bioactive component profiles based on geographical origin.
- Accurate geographical authentication is crucial for quality control and ensuring the efficacy of Pueraria lobata.
- Existing authentication methods may be destructive or lack the required precision.
Purpose of the Study:
- To explore the potential of terahertz (THz) spectroscopy for non-destructive geographical origin authentication of Pueraria lobata.
- To develop and evaluate a novel deep learning model integrating THz spectroscopy for Pueraria lobata classification.
- To compare the performance of THz spectroscopy with other analytical techniques and machine learning algorithms.
Main Methods:
- Samples of Pueraria lobata from eight Chinese regions were analyzed using THz spectroscopy.
- A CNN-Transformer hybrid deep learning network was developed for spectral data classification.
- High-Performance Liquid Chromatography (HPLC) was used to correlate spectral features with key bioactive compounds.
- Comparative analysis included Raman spectroscopy and five traditional machine learning algorithms.
Main Results:
- The THz spectroscopy approach with the CNN-Transformer hybrid model achieved a high accuracy of 91.67% for geographical origin classification.
- This accuracy significantly outperformed traditional methods (60.42%-64.58%) and a standard CNN architecture (85.42%).
- Perfect classification (F1-score = 1.000) was achieved for samples from Jiangxi and Shaanxi regions.
- HPLC analysis confirmed the link between THz spectral features and bioactive components like puerarin.
Conclusions:
- Terahertz spectroscopy, coupled with advanced deep learning, offers a robust and accurate solution for Pueraria lobata origin traceability.
- This non-destructive technique provides a superior alternative for quality control in traditional medicine.
- The developed hybrid model demonstrates significant potential for application in food and pharmaceutical authentication.
