目的 旨在通过近红外光谱（NIRS）信息与腰痹通胶囊（Yaobitong Capsules，YC）中间体中值粒径（D₅₀）的关系分析，探讨对该品种生产过程中4种中间体建立D₅₀近红外通用定量模型的可行性。方法 采集YC生产过程中的原料细粉、干燥颗粒、整粒颗粒和总混颗粒4种中间体的NIRS，考察不同预处理方法对模型的影响，并采用间隔偏最小二乘法（iPLS）、组合间隔偏最小二乘法（siPLS）和移动窗口偏最小二乘法（mwPLS）优选NIRS波段，采用偏最小二乘法（PLS）对4种中间体建立1个D₅₀通用定量模型。结果 通用模型的交叉验证均方根误差（RMSECV）为3.918 μm，预测均方根误差（RMSEP）为2.832 μm，预测相对偏差（RSEP）为2.26%，小于5%，性能偏差比（RPD）为4.60，大于3，该模型可以用于定量测定，且与独立模型比，预测准确性相差不大。结论 NIRS通用定量模型可用于YC 4种中间体D₅₀的测定。

Objective To explore the feasibility of establishing a D₅₀ near-infrared universal quantitative model for the four intermediates in the productive process of Yaobitong Capsules (YC) by analyzing the relationship between near infrared spectrum (NIRS) information and the median particle size (D₅₀). Methods The NIRS of the four intermediates in the production of YC, including raw material fine powder, dry particles, sieving particles and total mixed particles, were collected. Different pretreatment methods were investigated. The interval partial least squares (iPLS), the synergy interval partial least squares (siPLS), and moving window partial least squares (mwPLS) were used to screen the optimal NIRS interval. The partial least squares (PLS) method was employed to build a D₅₀ universal quantitative model for the four intermediates. Results In the universal model, the root mean square error of cross-validation (RMSECV) was 3.918 μm, the root mean square error of prediction (RMSEP) was 2.832 μm; The relative standard error of prediction (RSEP) was 2.26%, which was less than 5%; The ratio of predicton to deviation (RPD) was 4.60, which was greater than 3. The universal model could be used for quantitative determination, and its prediction accuracy was close to that of the independent model. Conclusion The NIRS universal quantitative model can be used to determine the D₅₀ of four intermediates in YC.

R283.6

国家科技重大专项：基于功效成分群的中药口服固体制剂先进制药与信息化技术融合示范应用（2018ZX09201010-004）；康缘中药智能化固体制剂工厂项目（KYYY190820）