Study On The Method Of Distribution Uniformity Evaluation Of Solid Components Based On Near Infrared Chemical Imaging

The solid dosage forms are the most common pharmaceutical preparations. Homogeneity is a critical quality attribute to ensure the quality of final product. Homogeneity not only includes content uniformity but also includes the spatial distributional homogeneity. The homogeneity of components, especially for active pharmaceutical ingredients (APIs), has an important effect on the effectiveness and safety of solid dosage products. Near infrared chemical imaging (NIR-CI) is a newly developed technique which combines conventional near-infrared (NIR) spectroscopy with chemical imaging to provide spectral and spatial information simultaneously. With the help of chemometric methods, the spectral information could be transformed to physical or chemical information. This article is aimed on the evaluation of component distributional homogeneity of solid dosage forms using NIR-CI as technical means combined with chemometric methods and imaging analysis methods, main contents are as follows:Firstly, the research of NIR-CI parameters optimization.The optimization method was established to investigate the effect of near NIR-CI detection parameters on hyperspectral data quality. In order to optimize the detection parameters, chlorpheniramine maleate (CPM) tablets were chosen as examples and the L9 (34) orthogonal-test design was adopted to research the effects of spectral resolution, spatial resolution, scan times and scan height. Binary image coupled with statistical measurement were proposed to quantitatively analyze hyperspectral data and determine the content of CPM on the tablet surface. High-performance liquid chromatography (HPLC) was used as reference method for accurate CPM determination. The absolute value of the difference between CPM contents obtained from NTR-CI and HPLC was chosen as index. The result demonstrated that the optimum parameters for acquiring hyperspectral data were:25μm×25μm (spatial resolution),5340 (scan height, the value of Z, precise focus),16cm-1 (spectral resolution) and 16 (scan times). The influence of scan height on hyperspectral data was firstly investigated. The optimized parameters could be applied to CPM tablets and other drugs for NIR-CI data acquisition and methodology establishment.Secondly, NIR-CI quantitative analysis of CPM tablets and Yinhuang tablets. The hyperspectral data were unfolded to two-dimensional data. Classical least squares (CLS) was used to obtain APIs quantitative information of samples. Different pretreat methods, including first derivative, S-G smoothing, MSC (multiplicative scatter correction) and SNV (standard normal variation), were used to improve the accuracy of prediction. The result showed that the RMSEP was the minimum after the MSC preprocessing. However, the high similarity between pure spectra of components (correlation coefficient was 0.95 or above) affected the accuracy of spectral discrimination by CLS, which lead to larger deviation between prediction value and theoretical value (RSD>10%).PLS quantitative models were built based on CPM tablets and Yinhuang tablets after comparing pretreat methods including first derivative, S-G smoothing, MSC and SNV. synergy interval partial least squares (SiPLS)and backward interval partial least squares (BiPLS) were used to select the best infrared band. The results show in the study of CPM analysis, the optimal pretreatment method is S-G 11 point smoothing plus second derivative.The quantitative method had a good prediction performances while RMSEC and RMSEP values were 0.484% and 0.631%; in the study of Jinyinhua extract (chlorogenic acid) quantitative analysis, the optimal pretreatment method was SNV, quantitative band of the SiPLS model were 7416 cm-1-7040 cm-1,6648 cm-1-6272 cm-1 and 5880 cm-1 -5504 cm;The quantitative method has a good prediction performances while RMSEC and RMSEP values are 0.286% and 0.175%; in the study of Huangqin extract (baicalin) quantitative analysis, the optimal pretreatment method was SNV, quantitative band of the SiPLS model were 7032cm-1-6656 cm-1,6648 cm-1 -5888 cm-1 and 5880 cm-1 -5504 cm-1. The quantitative method had a good prediction performance while RMSEC and RMSEP values were 4.67% and 1.70%.Thirdly, the evaluation of distributional homogeneity of solid dosage forms. Different methods were applied to evaluate the APIs distributional homogeneity based on the reconstructions of distributional images. Histogram analysis was used to evaluate the homogeneity through calculating parameters of mean, standard deviation, kurtosis and skewness. The result demonstrated that the sequence of CPM distributional homogeneity was: sample 1、sample3、sample2. However, the parameters to evaluate chlorogenic acid distributional homogeneity showed different evaluation results, illustrating the limitations of histogram analysis.Another criterion called the "percentage of homogeneity" (H%) was calculated to assess image homogeneity. The distribution of components was determined by macropixel analysis, which splited an image into non-overlapping blocks of a size and computed several parameters for the resulting divisional structure. Such parameters were used to further calculate mixing indices and H% was obtained. The closest H% was to 100, the more homogeneous the distribution would be. However, this method worked with non-overlapping macropixels, it was limited for the analysis of small distributional images. The criterion called distributional homogeneity index (DHI) was used to evaluate homogeneity. This approach was based on continuous-level moving block (CLMB), which could be performed on small images. For each macropixel size, the standard deviation of the macropixel was calculated and so called "homogeneity curve" was plotted. DHI was based on the ration of the area under the homogeneity curve of the real distributional maps and the area under the homogeneity curve of the randomized distributional maps. The closest DHI was to 1, the more homogeneous the distribution would be. DHI could avoid the optimization of marcopixel size, which made it a practical method to evaluate distributional homogeneity.At last but not the least, the evalution of API distributional homogeneity of commercial tablets.The CPM distributions of commercial CPM tablets from 6 different brands were successfully visualized using NIR-CI coupled with characteristic wavenumber method and binary images. Distributional Homogeneity Index (DHI) was utilized to assess CPM distributional homogeneity of different brands. According to the DHI values, the CPM spatial distribution of brand 1 and brand 6 were less homogeneous. To further evaluate the CPM homogeneity of different brands,1 was set as the expectation value of DHI and the divergence values of DHI values from 1 were calculated. The divergence values indicated that the distributional homogeneity of brand 4 was the best among 6 brands, following by brand2, brand 3, brand 5, brand 6 and brand 1.This paper focused on the evaluation of component distributional homogeneity of solid dosage forms. Hyperspectral data were unfolded to obtain two-dimensional data and conventional two-way chemometric methods were performed to obtain quantitative information. Then, the spatial distribution of components was realized through the reconstruction of ingredient distributional maps. Different methods were performed to evaluate the homogeneity of distributional maps. The result demonstrated that NIR-CI would be a powerful tool to assess the distributional homogeneity of components in pharmaceutical field.