Related Experiment Video
Updated: Jan 17, 2026
Author Spotlight: Harnessing DNA Barcode Technology to Enhance the Efficiency of Medicinal Plant Identification
Published on: November 1, 2024
Efficient artificial intelligence algorithms for on-site discrimination of Angelica sinensis
Jianan Yue1, Tingting Feng1, Liang Zhong1
1NMPA Key Laboratory for Technology Research and Evaluation of Drug Products, School of Pharmaceutical Sciences, Cheeloo College of Medicine, Shandong University, Jinan 250012, China; Key Laboratory of Chemical Biology (Ministry of Education), Shandong University, Jinan 250012, China.
Near-infrared (NIR) spectroscopy effectively identifies medicinal parts of Angelica sinensis. A convolutional neural network (CNN) model demonstrated versatility across different spectrometers for accurate classification of traditional Chinese medicine.
Area of Science:
- Analytical Chemistry
- Chemometrics
- Pharmacognosy
Background:
- Angelica sinensis is a valuable traditional Chinese medicine (TCM) with distinct medicinal parts (body, head, tail).
- Significant variations in chemical composition and pharmacological effects exist between different Angelica sinensis parts, necessitating accurate identification.
- Current classification methods for TCM parts lack clear characteristics, impacting quality control and efficacy.
Purpose of the Study:
- To develop and evaluate rapid classification methods for identifying the medicinal parts of Angelica sinensis.
- To compare the performance of benchtop and miniature near-infrared (NIR) spectrometers for this classification task.
- To assess the efficacy of classical machine learning algorithms and a convolutional neural network (CNN) for spectral analysis.
Main Methods:
- Utilized benchtop and four miniature near-infrared (NIR) spectrometers for spectral data acquisition.
- Applied classical machine learning algorithms, including partial least squares discrimination analysis (PLSDA) and support vector machine (SVM).
- Developed a convolutional neural network (CNN) classification model for identifying three medicinal parts of Angelica sinensis.
Main Results:
- Performance varied among NIR spectrometers, with some miniature devices showing limitations in accurately classifying TCM materials.
- The proposed CNN framework demonstrated versatility across different spectrometer configurations.
- Both Fourier-transform and miniature NIR spectrometers, coupled with chemometric classifiers, proved suitable for evaluating Angelica sinensis parts.
- The CNN model effectively reduced computational burden by eliminating complex parameter tuning.
Conclusions:
- Near-infrared (NIR) spectroscopy, particularly with advanced chemometric models like CNNs, is a viable tool for rapid and accurate identification of Angelica sinensis medicinal parts.
- Miniature NIR spectrometers show promise but require careful selection and validation for specific TCM classification tasks.
- The developed CNN approach offers a robust and adaptable solution for spectral analysis in TCM quality control, applicable across various spectroscopic instruments.
