中医证候的偏最小二乘和相似度分析

采用偏最小二乘和相似度分析等，探讨中医临床信息的提取和证候的数学模型。以肝炎后肝硬化为研究对象，利用相似度分析方法对临床表征信息(症状)、生物学指标和证候进行相似性分析；将逐步回归与偏最小二乘法相结合，在逐步回归方法筛选各证侯相关指标的基础上，采用偏最小二乘法建立的肝炎后肝硬化中医证候判识模型，通过Leave-One-Out法完成建模和预测。相似度分析方法所得证候与临床信息间的相似关系与临床实际基本相符，该技术可简化临床指标数量，为寻找中医证候特征提供简捷、可靠的方法；偏最小二乘法提高中医临床生物信息的利用度，并有较高的预测准确率，在中医证候建模的研究中有一定的应用价值。

Study of Traditional Chinese Medicine Syndrome by Partial Least Squares and Similarity Analysis

Partial least squares (PLS) and similarity analysis were developed to study the mathematic model of traditional Chinese medicine (TCM) syndrome with post-hepatitis cirrhosis (PHC). The similarity of clinical characteristic information (symptoms) and biological parameters with TCM syndromes were calculated. The resemblance between TCM syndrome and clinical information of disease obtained from similarity analysis is basically in accordance with the clinical practice, so, the technique could be taken as a method for finding the characteristics of TCM syndrome with simplified clinical parameters. Stepwise regression was used to analyze the clinical information, including clinical symptoms and biological parameters. The types of syndromes and the related clinical information which was screened by stepwise regression were set up as dependent variable matrix and independent variable matrix respectively. Then, the models in TCM were obtained for the classification and diagnosis of TCM symptom by Leave-One-Out in PLS program. PLS was definite useful in classification and modeling of TCM complicated syndromes.