Quality Grading of A. mongholicus Based on Effective Component
Content and Hyperspectral
WU Qiang1, ZHAO Peng2, WANG Meng2, ZHAO Cai-quan2, BAI Li-ge2, GUO Jia-hua2, GAO Xue-feng2, HOU Ding-yi3, GENG Zhi-gang4, LU Ling2*, LIU Jie2*
1. College of Agronomy，Henan Agricultural University，Zhengzhou 450046，China
2. School of Ecology and Environment，Baotou Teacher's College，Baotou 014030，China
3. Hohhot Agricultural and Animal Husbandry Technology Promotion Center，Hohhot 010020，China
4. Key Laboratory of Seaweed Fertilizers，Ministry of Agriculture and Rural Affairs，Qingdao 266400，China
Abstract：Huang qi is the dried root of the legume Astragalus membranaceus (Fisch. ) Bge. var. Mongholicus (Bge. ) Hsiao (A. mongholicus) or Astragalus membranaceus (Fisch. ) Bge., which has the functions of bu qi gubiao. However, its traditional quality evaluation methods are time-consuming, destructive, and subjective. The purpose of this study is to establish a rapid and non-destructive quality classification model of A. mongholicus by using ground feature hyperspectral technology and the key effective component content. Two hundred A. mongholicus root samples were collected from Guyang County, Baotou City, Inner Mongolia Autonomous Region; Astragaloside (AS) and calycosin-7-glucoside (C7G) content was determined by HPLC. Based on the effective component content, the samples were divided into four quality grades: ultra high AS, high AS, high C7G and ordinary by K-means clustering analysis; The diffuse reflectance spectrum data of each sample powder in the range of 350～2 500 nm were obtained using ASD FieldSpec 4 surface spectrometer, and SG smoothing pretreatment was performed; The competitive adaptive reweighted sampling (CARS) algorithm was used to select the characteristic wavelength from the full band spectrum, and the partial least squares discriminant analysis (PLS-DA), support vector machine (SVM) and random forest (RF) classification models were constructed based on the characteristic wavelength. The results showed that: (1) the average content of AS in ultra high AS (28), high AS (44), high C7G (36) and ordinary (92) were 0.130%, 0.112%, 0.096% and 0.089%, respectively, and the average content of C7G was 0.039%, 0.034%, 0.046% and 0.029%, respectively; (2) The spectral curves of A. membranaceus samples with different quality grades were significantly different in shape and absorption intensity, and the effective component content showed a significant correlation with the spectral reflectance in a specific wavelength region. The AS content had the highest correlation with the 1 890～1 900 nm band (r=0.621), while the C7G content had the highest correlation with the 1 356～1 365 nm band (r=0.636); (3) Among the three classification models, RF model performed best, and the overall accuracy of its correction set and validation set reached 94.8% and 92.3%, respectively, and the kappa coefficient reached 0.893. PLS-DA and SVM models also showed good classification performance. This study proved that the ground feature hyperspectral technology combined with CARS feature selection and RF classification model can realize the rapid and non-destructive grading of A. mongholicus quality, which can provide a new way for the evaluation of Huang qi quality.
[1] Pharmacopoeia Commission of the People's Republic of China（国家药典委员会）. Pharmacopoeia of the People's Republic of China（中华人民共和国药典）. Beijing：China Medical Science and Technology Press（北京：中国医药科技出版社），2020.
[2] Lin Z H，Lu L L，Guo W L，et al. Food Bioscience，2025，68：106560.
[3] Zhao Y，Teng H，Yu Z X，et al. Food Chemistry，2025，480：143631.
[4] Liu D D，Qin H W，Liu G H，et al. Journal of Functional Foods，2025，126：106695.
[5] Shu L X，Zhang S M，Qian J，et al. Rapid Communications in Mass Spectrometry，2025，39（8）：e9977.
[6] Hua S Y，Liu J Y，Pu S J，et al. Microchemical Journal，2025，210：113017.
[7] Xue F F，Xie L T，Zhang X，et al. Biomedical Chromatography，2025，39（5）：e70073.
[8] WEI Hua-hua，CHANG Li-jing，LI Ming-yue，et al（韦花花，常丽静，李明月，等）. Journal of Chinese Medicinal Materials（中药材），2022，45（4）：890.
[9] SUN Xiu-rui，WEI Gui-jie，WANG Hong-ya，et al（孙秀蕊，魏桂杰，王宏雅，等）. China Journal of Chinese Materia Medica（中国中药杂志），2021，46（4）：966.
[10] Xu X Y, Xie W G, Xiang C，et al. Journal of Food Measurement and Characterization，2023，17（4）：3218.
[11] WANG He，LIU Shi-yu，WANG Hui，et al（王 翯，刘诗雨，王 慧，等）. China Journal of Chinese Materia Medica（中国中药杂志），2023，48（16）：4320.
[12] Xin X，Sun J，Shi L，et al. Journal of Food Composition and Analysis，2025，143：107625.
[13] Xiao Q L，Bai X L，Gao P，et al. Sensors，2020，20（17）：4940.
[14] Xu Y N，Wu W F，Chen Y，et al. Frontiers in Plant Science，2022，13：1031849.
[15] Yang J，Yin C L，Miao X，et al. Spectrochimica Acta Part A：Molecular and Biomolecular Spectroscopy，2021，248：119251.
[16] Chen X F，Du H J，Liu Y，et al. Scientific Reports，2024，14（1）：7209.
[17] Zhu J Q，Bao J Q，Tao Y. Sensors，2023，23（23）：9345.
[18] Zhang H，Pan Y X，Chen Y，et al. Analyst，2024，149（6）：1837.
[19] Tsai C C，Wu H H，Chang C P，et al. Journal of the Formosan Medical Association，2019，118（3）：730.
[20] Wang X M，Wang Y L，Huang D M，et al. International Immunopharmacology，2022，112：109186.
