基于近邻传播聚类方法的慢性胃炎症状群分布特征研究

目的：本研究旨在挖掘慢性胃炎（Chronic Gastritis，GC）常见中医证候（体征）的特征组合，分析发病特征，以此探讨中医证候的客观化研究。方法：对临床调查获取的919例慢性胃炎数据，运用互信息构建各症状（体征）的相似矩阵，采用近邻传播聚类（Affinity Propagation Clustering，APC）算法对88个慢性胃炎症状（体征）进行最优聚类。结果：88个慢性胃炎症状（体征）经APC算法分析后，最终被聚成22类症状群，且大部分症状聚类可以解释为某个相关的中医证素，能较完整的体现中医临床慢性胃炎证素分布规律及特点。结论：APC算法获取的慢性胃炎中医症状（体征）聚类可为疾病的发病特征分析提供依据，与中医理论有较高的一致性，这提示APC算法可用于中医证候的客观化研究。

Symptoms Grouping Distribution of Chronic Gastritis Based on Affinity Propagation Clustering

This study was aimed to explore the general and pathognomonic rules of symptoms grouping distribution of chronic gastritis (CG) in traditional Chinese medicine (TCM,) and to further define and discuss the standardization of syndrome differentiation of TCM. A total of 919 CG cases were collected. Mutual information was the first method to be used in the establishment of the similarity matrix among symptoms (signs). And then, the optimal clustered symptom groups were identified for the involved 88 CG symptoms (signs) based on affinity propagation clustering (APC) algorithm. The results showed that 88 CG symptoms (signs) finally formed into 22 groups. Almost every group can be interpreted with an associated TCM element. It can better reflect the general rules of TCM clinical syndrome elements distribution of CG. It was concluded that the symptoms grouping distribution was consistent with clinical experiences and TCM theories. It suggested that the APC algorithm can be used in the objectivization study of TCM syndromes.