目的 基于药材粉末颜色数字化分析建立一种不同生长方式黄芩Scutellaria baicalensis的快速鉴别方法，同时建立一种黄芩主要药效成分含量的准确预测模型。方法 采用分光测色仪测定黄芩样品粉末颜色亮度值（L^*）、红绿色值（a^*）、黄蓝色值（b^*），并计算总色值（E^*）；采用HPLC测定黄芩样品中4个主要药效成分（黄芩苷、汉黄芩苷、黄芩素、汉黄芩素）的含量，基于色度值并结合多元统计分析及机器学习算法，建立可区分不同生长方式黄芩的定性判别模型、主要药效成分的定量预测模型。结果 不同生长方式黄芩的黄芩素和汉黄芩素含量具有显著差异（P＜0.05）；黄芩药材粉末色度值与有效成分含量具有相关性（除a^*与黄芩苷、汉黄芩苷含量无显著相关性外，其他因素之间P＜0.05）；基于色度值并结合机器学习算法构建的黄芩生长方式判别模型中，随机森林分类器结合十倍交叉验证法的分类准确率最高，为100%；并在基于色度值构建的主要药效成分含量预测模型中，4种成分的预测值与实测值的相关系数均大于0.97。结论 通过药材粉末的色度值建立的定性判别及定量预测分析方法具有快速、准确、可及的优点，为揭示黄芩“辨色论质”的科学内涵提供理论依据。

Objective In order to establish a rapid method for distinguishing different growth modes of Huangqin (Scutellariae Radix) based on digital analysis of the color of the herbal powder, and to establish an accurate prediction model for the main active components of Scutellariae Radix based on the color of the herbal powder. Methods The L^*(luminance), a^*(red-green), b^*(yellow-blue) colorimetric values and E^*(total color) of the powder color of Scutellariae Radix were measured using a spectrophotometer, and the contents of flavonoid components (baicalin, wogonoside, baicalein, and wogonin) were determined by HPLC. Based on the colorimetric values, combined with multivariate statistical analysis and machine learning algorithms, a qualitative discriminant model for different growth modes of Scutellariae Radix and a quantitative prediction model for four main active components were established. Results There were significant differences in the contents of flavonoid components baicalein and wogonin between different cultivation methods of Scutellariae Radix (P < 0.05). The colorimetric values of the powdered herb were correlated with the content of effective components (except a^* was not significantly correlated with baicalin and wogonin contents, P < 0.05). In the model for discriminating the growth patterns of Scutellariae Radix based on colorimetric values and integrated with machine learning algorithms, the Random Forest classifier combined with ten-fold cross-validation achieved the highest classification accuracy of 100%. Furthermore, the main active component content prediction model based on chromaticity values showed that the correlation coefficients between the predicted values and the measured values of the four effective components were all greater than 0.97. Conclusion The qualitative and quantitative prediction analysis method established by using the chromaticity value of the herbal powder has the advantages of rapidity, accuracy, and accessibility, providing theoretical basis for revealing the scientific connotation of "distinguishing colors to judge quality" of Scutellariae Radix.

R286.2

山西省2022-2023年度中医药科技创新工程项目[2100601中医（民族医）药专项];2023年北中医基本科研业务费（揭榜挂帅）重点项目立项（2023-JYB-JBQN-058）