中药大蜜丸多源信息融合策略及其整体质量一致性研究

The holistic view of traditional Chinese medicine (TCM) is the basis of its quality consistency. Actually, the key technology to evaluate the overall quality is to establish mathematical models via multi-source information fusion strategy. It is acknowledged that the chemical components of big honeyed pills have little relationship with the pharmaceutical effect. Therefore, it is essential to develop a full-scale model for the overall quality consistency evaluation. In this study, we propose the scientific hypothesis that multi-level information fusion strategy could be implemented to realize the overall quality consistency evaluation of TCM of big honeyed pills..Tongren Niuhuang Qingxin Wan is taken as the representative big honeyed pills in this study. The following steps are carried out: firstly, the three dimensions information including sensory information, chemical information, and pharmacodynamic effect information are acquired for the characterization of overall quality. Secondly, the multi-source information fusion strategy is developed by employing low-level and mid-level fusion strategies, multivariate statistical process control and the variable selection algorithm. Thirdly, the low-level and mid-level fusion strategies combined with the least squares support vector machine are individually employed to establish the three-dimensional models. Moreover, the models provide technical support for the quality control of TCM of big honeyed pills.

中药整体观是其质量一致性的基础。多源信息融合策略是实现中药大蜜丸整体质量一致性评价的核心。针对中药大蜜丸化学指标成分难以关联其药效这一问题，本项目提出基于多层次融合的多源信息融合策略实现中药大蜜丸整体质量一致性评价这一科学假说。.以中药大蜜丸的代表产品同仁牛黄清心丸为载体，开展样品感官信息、化学成分信息和药效信息这三个维度的整体质量表征；进一步通过多元统计过程控制模型和特征变量筛选算法，综合采用数据层融合和特征融合方法，开发基于多层次融合的中药大蜜丸多源信息融合策略；最后基于上述多源信息融合策略，开发同仁牛黄清心丸整体质量一致性最小二乘支持向量机模型，建立其整体质量一致性评价方法，为中药大蜜丸质量控制提供关键技术支撑。

中药大蜜丸由于配方复杂、原料差异大、生产过程主观因素多等方面的制约，导致其质量一致性较难控制。中药整体观是其质量一致性的基础，多源信息融合策略是实现中药大蜜丸整体质量一致性评价的核心。本项目以180个同仁牛黄清心丸为载体，开展样品物理属性和化学属性的共同表征；通过多元统计过程控制模型（MSPC）与主成分分析算法，采用数据融合方法，开发中药大蜜丸多源信息融合策略；通过斑马鱼抗贫血模型，验证上述多源信息融合策略，实现中药大蜜丸整体质量一致性评价。结果表明，基于数据融合策略，Vis-NIR光谱，MIR光谱和质构物理属性三种方法的权重分别为45.10%，42.44%和12.46%，基于可见-近红外（Vis-NIR）光谱，中红外（MIR）光谱和质构物理属性MSPC模型的结果，结合三种方法的权重，实现同仁牛黄清心丸整体质量一致性评价，依据投票法，得到三个异常样品。中药大蜜丸质量一致性评价方法的构建，不仅是大蜜丸质量控制的关键环节，而且是保障中药产品稳定均一的重要内容。本研究为中药大蜜丸的质量控制提供理论依据。