基于隐结构模型和D-S证据理论的乳腺癌中医辨证研究OA

Objective To construct a decision support framework integrating LSM and Dempster-Shafer(D-S)evidence theory to determine clustered syndrome patterns by addressing the issues of significant inter-expert subjectivity and insufficient stability in conclusions when using the latent structure model(LSM)for TCM syndrome differentiation of breast cancer;To enhance the objectivity of syndrome differentiation and the precision of pattern classification during the clustering analysis phase of LSM research.Methods Typical clinical research data containing complete quantitative indicators were screened for the model.Group decision-making theory was applied to integrate the experience of multiple experts.A two-stage syndrome differentiation model was established:in the first stage,LSM was used to identify latent association patterns among syndrome elements;in the second stage,using breast cancer symptoms as the identification framework,the LSM differentiation results were used as the basic probability assignment,and expert experience was incorporated for evidence fusion via the D-S theory.The Dempster combination rule was finally applied to resolve conflicts and fuse the credibility of multi-source TCM evidence,providing a quantitative solution for clustering the results of the LSM.Results Based on symptom data from 927 breast cancer patients,in the analysis process of the latent variable Y1,which was difficult to determine the classification of syndrome types,the decision support framework was used to ultimately determine that the corresponding syndrome type of suspicious and angry syndrome,namely liver depression and phlegm coagulation syndrome,could be used as the clustering criteria for Y1.Conclusion The model effectively reduces the subjective uncertainty of TCM expert evidence through a quantitative evidence fusion mechanism,which can provide an interpretable decision support tool for further clustering analysis and significantly improve the standardization level while preserving holistic syndrome differentiation thinking.