Determining the geographical origin of Fritillaria by terahertz spectroscopy and machine learning algorithms

Yuhao Feng, Chengqian You, Chunyi Zhang, Shuo Zhao, Qiuhong Qu, Pengfei Wang, Mingxia He,
Determining the geographical origin of Fritillaria by terahertz spectroscopy and machine learning algorithms,
Chemical Physics Letters,
Volume 878,
2025,
142350,
ISSN 0009-2614,
https://doi.org/10.1016/j.cplett.2025.142350.
(https://www.sciencedirect.com/science/article/pii/S0009261425004920)
Abstract: Fritillaria species have distinct medicinal values with respect to their geographical origins, but it remains a challenge to directly achieve accurate classification due to their similar appearance. Considering the high time consumption and operation complexity of conventional techniques, the combination of terahertz (THz) spectroscopy and machine learning algorithms offers a fast and non-destructive approach. In this work, Fritillaria cirrhosa D. Don, Fritillaria ussuriensis Maxim, Fritillaria pallidiflora Schrenk, and Fritillaria thunbergii Miq, are selected and characterized by THz spectroscopy to obtain their corresponding absorption coefficient. After data preprocessing, principal component analysis (PCA) is applied to reduce the dimensionality of THz data. Machine learning algorithms, including least squares support vector machine (LSSVM), particle swarm optimization support vector machine (PSO-SVM), random forest (RF), and convolutional neural networks (CNN), are used for origin classification, where PSO-SVM model without preprocessing exhibits the advantageous classification performance with overall accuracy of 95.8 %. It also displays favorable reliability by conducting iteration test, achieving highest mean overall accuracy of 95.04 % and smallest standard deviation of 0.74 % respectively. This work reports an effective method for precisely categorizing the geographical origins of Fritillaria, which provides a new perspective for ensuring the pharmaceutical efficacy of traditional Chinese medicine.
Keywords: Fritillaria; Geographical origin; Terahertz spectroscopy; Machine learning algorithms