目的 应用近红外光谱（near-infrared reflectance spectroscopy，NIRS）与中红外光谱（mid-infrared reflectance spectroscopy，MIRS）技术，对热毒宁注射液（Reduning Injection，RI）制剂过程的投料和二次热处理工序中6种质控指标进行快速检测，提高制剂过程的质量控制水平。方法 利用NIRS透射技术与MIRS衰减全反射技术，结合偏最小二乘法（partial least squares，PLS），经过光谱预处理方法的优选以及波段筛选，分别建立绿原酸、新绿原酸、隐绿原酸、栀子苷、断氧化马钱子苷和固含量的快速预测模型，以校正集相关系数（r_cal）、验证集相关系数（r_pre）、校正均方根误差（root mean square error of calibration，RMSEC）、交叉验证均方根误差（root mean square error of cross validation，RMSECV）、预测相对误差（relative standard error of prediction，RSEP）为评价指标，评价模型性能。结果 NIRS预测模型的6个质控指标的RMSEC和RMSECV均小于0.3，RSEP小于4.0%；MIRS预测模型的6个质控指标的RMSEC和RMSECV均小于0.4，RSEP均小于5.0%，建立的PLS模型具有模型性能好、预测精度高的优点。结论 NIRS及MIRS分析技术，均可用于RI投料和二次热处理工序中6种质控指标的快速检测，模型RSEP在5%以内，方法操作简单，结果可靠。

Objective The near-infrared reflectance spectroscopy (NIRS) and mid-infrared reflectance spectroscopy (MIRS) techniques were used to detect six quality control indexes in the feeding and secondary heat treatment process of Reduning Injection (热毒宁注射液, RI) preparation process, so as to improve the quality control level of preparation process. Methods The quantitative correction models of chlorogenic acid, neochlorogenic acid, cryptochlorogenic acid, gardenoside, secoxyloganin and solid content were established by using NIRS technology and attenuated total MIRS technology, combined with partial least squares (PLS) method, after spectral pretreatment method selection and band selection respectively. Calibration set correlation coefficient (r_cal), validation set correlation coefficient (r_pre), root mean square error of calibration (RMSEC), root mean square error of cross validation (RMSECV) and relative standard error of prediction (RSEP) were used as evaluation indexes to evaluate the model performance. Results The RMSEC and RMSECV of the six quality control indexes of the NIRS prediction model were both less than 0.3 and close to each other, and the RSEP value were all less than 4.0%. The RMSEC and RMSECV of the six quality control indexes of the MIRS prediction model were all less than 0.4 and close to each other, and the RSEP value were all less than 5.0%. These PLS models had the advantages of good model performance and high prediction accuracy. Conclusion Both NIRS and MIRS technology can be used for rapid detection of six quality control indexes in RI feeding-secondary heat treatment process, and the RSEP value is within 5%, and the method is simple and reliable.

R283.6

2022年中央财政转移支付地方项目：基于重点研究室研究领域的中医药多学科研究能力提升项目——中药提取精制新技术