Application Of Near Infrared Spectroscopy To Process Quality Analysis Of Traditional Chinese Medicine

The quality evaluation of Chinese herb, quantitative analysis of Chinese herb extract and fast analysis of the extraction process using near infrared (NIR) spectroscopy were studied in this paper. The main works can be summarized as follows.1. NIR spectroscopy was used for quality analysis of Chinese herb with Salvia miltiorrhiza herb as an example. NIR diffuse reflectance spectra of 62 samples were obtained. 10 out of 19 samples from production base were randomly selected to form the standard set. The similarity match model was built based on this set. Using this model, the similarity match values of the other 9 samples from production base and 43 samples from other production areas were computed. The results showed that the proposed method could reflect the quality differences of Salvia miltiorrhiza herbs and distinguish the herbs from production base from those of other production areas.2. The determination of 4 active components of Coptis chinensisextracts was carried out using NIR diffuse reflectance spectroscopy. Partial least square regression (PLSR) was utilized to build the calibration models using second derivative spectra with HPLC as a reference method for each component. RMSEP for berberine, palmatine, jatrorrhizine and total alkaloid were 0.184, 0.109, 0.054 and 0.325, respectively. The recoveries were determined to be 97.67 - 99.59%, 96.63 -100.70%, 95.15 - 101.15% and 97.41 - 99.89% for berberine, palmatine, jatrorrhizine and total alkaloid, respectively. The relative standard deviations were 0.32% for berberine, 0.63% for palmatine, 1.76% for jatrorrhizine, and 0.33% for total alkaloid, respectively.3. The real-time detection of the percolation extracting process of panax notoginseng herb using NIR transmittance spectroscopy together with an optic fiber was further studied. Reference measurements of ginsenoside Rgl, Rbl, and Rd were performed by HPLC, and panax notoginseng saponins (PNS) were determinedaccording to the colorimetric assays. Radial basis function neural networks(RBFNN) and PL5R were studied comparatively to establish the calibration model, resulting in that, for Rgl, Rbl, Rd, PNS, using RBFNN, RMSECV were 1.120, 1.230, 0.267, 4.749 and RMSEP were 0.677, 0.969,0.155, 8.065, respectively.In addition, the transfer of calibration model is a key problem for NIR analysis because it has important effects on the share of data base and application spread. The Ceftazidime for injection was used as a simple example in this paper to investigate the transfer performance of calibration model between two instruments of the same type, while the correction methods being also investigated.