无指导的中医证候诊断数据的隐变量分析

Latent Variable Analysis of Unsupervised Diagnostic Data of Traditional Chinese Medicine

本文探讨中医临床诊断得到的"望"、"闻"、"问"、"切"的四诊数据的统计分析方法。在没有中医专家的证候诊断结果(即无指导)的情况下,利用含隐变量的结构方程模型(SEMs),进行中医证候诊断的方法。本文提出证候诊断和病势诊断的两层隐结构模型和两步估计方法.本研究收集了433例脑动脉硬化症患者的临床诊断数据,利用本文提出的方法进行数据分析,得到了与传统中医诊断理论相当吻合的结果.在没有中医专家诊断指导的情况下,本文提出的无指导诊断方法为客观评价中医辨证理论提供了数据分析基础.

In this paper, we applied and expand the method of structural equation models （SEMs） with latent variables for diagnosis in Traditional Chinese Medicine （TCM） to study the relationship among symptoms, syndromes and diseases without supervised information of experts in TCM. We apply the method to a real study which collects clinic diagnostic data from 433 cerebral arteriosclerosis patients, and we obtained the result which matches the theory of TCM very well. The method proposed in this paper provides a statistical foundation of syndrome diagnosis based on unsupervised data in TCM.