基于多源信息融合方法的中医诊疗指南临床应用多维评价模型研究

With the development of many clinical practice guidelines (CPGs)，more attention has been paid to the evaluation methods of how to understand the various elements of a high quality CPG and identify the quality of the CPG. There were multiple CPGs evaluation tools in the world. These guidelines paid more important to the evidence-based evidence than the clinical effect. The evaluation project of Traditional Chinese Medicine (TCM) application guidelines lack systematicness and objectivity, although it had explored the guidelines' framework and the data base. Multi-source information fusion method, a computer technology, can simulate the principle of human's acknowledgment, using the multi-source information between the related and complementary characteristics, and the appropriate fusion algorithm, we can obtain more comprehensive and accurate detection, identification, estimation and decision making about the study object, by this process the technology provides the possibility of establishing guidelines' objective evaluation method. In this study, we explore the effect of multi-source information fusion methods on improving the reliability of guide evaluation, by using the experience of traditional Chinese medicine experts and the rich information resources of ancient and modern literatures, combining the evaluation data of TCM CPGs.It can improve the applicability, objectivity and accuracy of TCM CPGs evaluation decision By constructing a multi-source information fusion model for CPG application evaluation of Chinese medicine, screening and optimizing the appropriate multi-source information fusion method to improve the fusion algorithm and strategy of different types of evaluation information, The study shows a great theoretical and practical value for establishing objective evaluation method of TCM CPGs.

随着大量临床实践指南的制定，能够帮助使用者理解高质量指南的各个要素并且分辨出其质量高低的评估方法日益受到重视。国际上制定了多个指南评价工具，但多以循证证据为评价要点，忽视临床应用效果。中医临床诊疗指南应用评价项目虽然探索了应用性评价的框架和数据基础，但仍缺乏系统性和客观性。多源信息融合方法是通过计算机模拟人类认识事物的原理，利用多源信息之间相关和互补，获得关于研究对象更全面、更准确的检测、识别、估计和决策，为建立指南客观评价方法提供了可能性。本课题针对中医专家诊疗经验、古今文献等丰富的信息资源，结合中医诊疗指南应用评价数据，探索多源信息融合方法对提高指南评价可靠性的作用。通过构建指南应用评价的多源信息融合模型，筛选和优化适宜的多源信息融合方法，改进不同类型评价信息融合的算法和策略，提高中医诊疗指南评价决策的适用性、客观性和准确性，对于建立中医自主的指南客观评价方法具有重要的理论和应用价值。

本研究依据多源信息融合方法，利用多源信息之间相关和互补，获得关于研究对象更全面、更准确的检测、识别、估计和决策，为建立指南客观评价方法提供了可能性。本课题针对中医专家诊疗经验、古今文献等丰富的信息资源，结合中医诊疗指南应用评价数据，探索多源信息融合方法对提高指南评价可靠性的作用。研究内容：.（一）基于信息融合技术的中医诊疗指南应用可靠性评价模型研究。通过研究临床应用评价数据库、专家经验、指南相关的古籍和现代文献等不同数据源信息的融合方法，建立基于信息融合的可靠性评估功能模型，比较优化多源信息的融合层次、模式的适用性，确定不同信息融合结构的适用范围，和相应结构层次完成的信息融合的决策可靠性程度。.（二）可靠性数据融合算法的适用性研究。通过试用不同的融合算法，比较评估结果，确定不同聚合方法对指南评价数据融合的适用性。两种不同策略的应用结果表明：若先进行多源信息聚合，会增加最终融合结果的特征复杂度，但能实现更深的信息探究。若先进行多源信息融合，能增加最终融合结果的可解释性。.（三）基于Bayes方法的可靠性信息融合策略研究.不同指南评价相关信息特点，重点研究共扼方法、经验Bayes法、最大熵法在构建中医诊疗指南应用评价信息先验分布的适用性。.（四）基于证据理论的可靠性信息融合策略研究。选择较权威的、应用较多、利于中医诊疗指南评价数据确定权重的方法，根据不同权重确定的规则，进行不同策略下的证据组合权重比较。同时比较不同权重确定的规则对评价信息证据组合的影响；利用DS证据理论的组合规则来综合各专家对此项概括性指标（标准的疗效评价终点指标合理性）的意见，最终得到各专家对此项指标评分的综合结果。.通过构建指南应用评价的多源信息融合模型，筛选和优化适宜的多源信息融合方法，改进不同类型评价信息融合的算法和策略，提高中医诊疗指南评价决策的适用性、客观性和准确性，对于建立中医特色的指南客观评价方法具有重要的理论和应用价值。