近红外光谱技术在热毒宁注射液萃取工艺过程质量控制研究

应用近红外光谱技术建立热毒宁注射液萃取过程绿原酸含量和固含量的分析模型。收集7批金青萃取液样品,扫描离线光谱,测定绿原酸含量和固含量,分别用偏最小二乘法(PLS)和人工神经网络法(ANN)建立定量校正模型,并用此模型对未知样品进行预测。PLS模型中,绿原酸和固含量校正集R2分别为0.987 2,0.981 2;RMSEC分别为0.153 3,0.794 3;预测集R2分别为0.983 7,0.973 3;RMSEP分别为0.246 4,1.259 4;RSEP分别为3.25%,3.31%。ANN模型中,绿原酸和固含量校正集R2分别为0.990 3,0.988 2;RMSEC分别为0.097 4,0.454 3;预测集R2分别为0.986 8,0.969 9;RMSEP分别为0.192 0,0.942 7;RSEP分别为2.61%,2.75%。绿原酸和固含量的PLS模型和ANN模型的RSEP均在6%以内,能够满足中药生产过程中质量分析要求。ANN模型的RSEP低于PLS模型,具有更好的预测准确性。建立的近红外光谱快速检测绿原酸含量和固含量的方法,操作简单,准确可靠,可用于热毒宁注射液金青萃取过程质量控制。

Quality control in liquid-liquid extraction of Reduning injection using near-infrared spectroscopy technology

Quantitative models were established to analyze the content of chlorogenic acid and soluble solid content in the liquid-liquid extraction of Reduning injection by near-infrared(NIR) spectroscopy. Seven batches of extraction solution from the liquid-liquid extraction of Lonicerae Japonicae Flos and Artemisiae Annuae Herba were collected and NIR off-line spectra were acquired. The content of chlorogenic acid and soluble solid content were determined by the reference methods. The partial least square (PLS) and artificial neural networks (ANN) were used to build models to predict the content of chlorogenic acid and soluble solid content in the unknown samples. For PLS models, the R² of calibration set were 0.987 2, 0.981 2, RMSEC were 0.153 3, 0.794 3, the R² of prediction set were 0.983 7, 0.973 3, RMSEP were 0.246 4, 1.259 4, RSEP were 3.25%, 3.31%, for chlorogenic acid and soluble solid content, respectively. For ANN models, the R² of calibration set were 0.990 3, 0.988 2, RMSEC were 0.097 4, 0.454 3, the R² of prediction set were 0.986 8, 0.969 9, RMSEP were 0.192 0, 0.942 7, RSEP were 2.61%, 2.75%, for chlorogenic acid and soluble solid content, respectively. Both the RSEP values of chlorogenic acid and soluble solid content were lower than 6%, which can satisfy the quality control standard in the traditional Chinese medicine production process. The RSEP values of ANN models were lower than PLS models, which indicated the ANN models had better predictive performance for chlorogenic acid and soluble solid content. The established method can rapidly measure the content of chlorogenic acid and soluble solid content. The method is simple, accurate and reliable, thus can be used for quality control of the liquid-liquid extraction of Reduning injection.