Study On Methods Of Conception Synonymy Analysis And Connotation Classification Of Symptoms In Traditional Chinese Medicine

Symptom phenotypes are essential in syndrome differentiation and treatment of traditional Chinese medicine(TCM)and are main input features for the computer-assisted diagnosis and treatment system in the field of artificial intelligence in Chinese medicine as well.Therefore,unified expressions and unambiguous connotations of symptoms would play an important basic role in the construction of symptom ontology and clinical data mining in the field of TCM.However,there are three main difficulties about symptom phenotypes in the applications of artificial intelligence in TCM.First,the same symptom concept can be expressed with different symptom terms:how to make the computer recognize and integrate symptom synonyms from different sources in order to provide a convenient and unified interface for multi-source symptom data mining and analysis?Second,symptoms have a widely variety of connotations:how to make the computer understand and quantify the diversity of symptom connotations in order to provide a reference for the research on the TCM symptom connotation diversity with modern technology?Third,different patients may have different symptom connotations because of individual differences in the clinical environment:how to make the computer automatically identify the individualized symptom connotation of one patient in order to provide assistance for the development of precision medicine and intelligent clinical decision-making technology in TCM?To solve the above problems,this paper made a systematic and profound study on the terms,concepts and connotations of TCM symptoms based on a large number of symptom data in TCM monographs and clinical medical records with machine learning algorithms,complex network and statistical techniques.The main work of this paper included the following aspects:(1)Aiming at solving the problem of the identification and integration of symptom synonyms from different sources,this paper collected symptom data from three TCM monographs and constructed a symptom concept name and source information table and a standard data set of symptom synonymous relations.This paper also proposed a symptom synonyms prediction method based on similarities of symptoms' features and a concept-based clinical symptom terms integration method,which could identify the symptom synonyms automatically and solve the integration problem of symptom terms from multiple sources.(2)Aiming at solving the problem of quantifying the diversity of symptom connotations,this paper constructed a TCM clinical symptom network based on co-occurrences between two symptoms,and proposed a quantitative metric of symptom connotation diversity based on the TCM clinical symptom network with correlation analysis and multiple regression analysis,which could fit the standard symptom connotation diversity and realize the explicit expression of symptom connotation diversity(3)Aiming at solving the problem of individual identification of the symptom connotation,this paper designed two encoder-decoder models to identify the specific symptom connotation in clinical environments and made comparative studies and case analysis on two different clinical data sets,which verified the validity and interpretability of these two models.