Research On TCM Intelligent Auxiliary Diagnosis Method Of Chest Impediment Based On Model Of Pathogenesis

Objective:This project creates a diagnostic database for auxiliary diagnosis of chest impediment.It studies traditional Chinese medicine(TCM)syndrome differentiation methods and pathogenesis theory by using natural language processing technology suitable for the characteristics of heterogeneous TCM data under the guidance of TCM theory and clinical thinking.The study also design models based on differentiation of syndrome elements as well as knowledge graph of pathogenesis based on model of pathogenesis,which can evaluate the obstruction of disease like the application and verification on chest impediment.All these research provides new thoughts of introducing intelligent auxiliary diagnosis method in TCM.Method:This research work includes two parts:theoretical analysis and application verification to achieve the research purpose.(1)Construct a relevant diagnostic database suitable for intelligent auxiliary diagnosis of TCM,by using natural language processing technology to collect diagnostic and treatment standard data and symptom data related to chest impediment,and then manually proofread it.(2)Through the study of the theory of syndrome differentiation,a multi-level representation method of syndrome differentiation elements is designed,and the deep learning technology is applied to transform the process of TCM syndrome differentiation into a multi-label text classification algorithm to build a syndrome differentiation model.(3)Based on the study of the theory of the relationship between the pathogenesis and syndromes,a pathogenesis factor hierarchy framework is designed by referring"Clinical Diagnosis and Treatment of Traditional Chinese Medicine Terminology Part 2:Syndrome".Such framework uses syndromes and pathogenesis of data mining,analyzes symptoms and signs and pathogenesis factor,the relationship between pathogenesis factor combination pattern mining and the logical relationship between the pathogenesis unit,it also designs the mechanism model and constructs the knowledge graph for pathogenesis.(4)To evaluate the accuracy of the three schemes,111 cases of well-known TCM medical records of chest impediment have tested by using the syndrome differentiation model based on syndrome differentiation elements,the pathogeneticknowledge graph based on model of pathogenesis and the fusion of the two methods.Results:(1)In the construction of diagnostic database related to auxiliary diagnosis,a total of 22 standard syndromes related to chest impediment from different sources are extracted,380 syndrome-symptom data items are formed,125 syndromes related to diagnosis knowledge data are extracted,and 2619 syndrome-symptom data items are formed.The symptom database containing 441 concept words and 22,758 symptom words is constructed.The database normalizes treatment of symptoms in more TCMdiagnosis and treatment data and adds new symptom words to the symptom database.(2)Based on the TCM syndrome differentiation process,the TCM syndromes are classified into compound syndromes,minimal syndromes,and basic syndromes.By restoring the reasoning process of syndrome diagnosis in the process of syndrome differentiation,and analyzing the constituent elements of syndrome name,a multi-level representation framework of syndrome differentiation elements is designed.The syndrome differentiation elements are divided into five categories:disease location elements,disease nature elements(basic material,etiology,pathological state)and connectives.A pattern extraction method based on the idea of pattern matching is adopted to realize the automatic extraction technology of syndrome differentiation elements.The accuracy of the method was 71.4%,26.4%of the students needed manual training,total errors accounted for 2.2%.This method is suitable for mining syndrome names and syndrome differentiation elements in a large number of TCM diagnosis and treatment data to provide data support for auxiliary diagnosis.(3)Based on the extraction of syndrome differentiation elements from national standards,teaching materials and industry standards for chest impediment,a multi-level classification data set of syndrome differentiation elements has been constructed.By taking the differentiation elements of chest impediment as the classification label,and applying deep learning technology,three multi-label classification differentiation models of symptom-syndrome element,symptom-minimum syndrome,and symptom-syndrome have been designed and constructed.The model was trained and tested by the selected famous doctors cases of chest impediment.The model's prediction accuracy F1 value of syndrome differentiation elements can reach 82.90%,and the syndrome diagnosis accuracy F1 value accounts for 72.08%.(4)Through the theoretical research of TCM pathogenesis and syndrome,the relationship among disease,symptom and sign,pathogenesis and syndrome is sorted out,and further analysis is made according to the Semantic Connotation of pathogenesis and syndrome name,and the theoretical basis of the model construction of pathogenesis is found.Then take the syndrome pathogenesis data as the research object,through the research on the pathogenesis elements extracted from the syndrome pathogenesis data,design the hierarchical framework of the pathogenesis elements,and the pathogenesis elements are divided into 32 categories with 4 levels.We extracte 5675 data of pathogenesis unit from "Chinese Medicine Clinical Diagnosis and Treatment Terminology Part 2:Syndrome",and extracted 851 elements of pathogenesis.Among them,there are 197 disease elements,619 disease elements,and 44 connectives.Among the disease elements,there are 30 basic substances,50 physiological functions,207 etiological elements,and 332 pathological conditions.Then,the relationship between symptoms and signs and pathogenesis elements,the combination mode of pathogenesis elements,and the logical relationship between pathogenesis units are analyzed,and the model of pathogenesis of syndrome-pathogenesis unit-pathogenesis elements-symptoms and signs is constructed.Based on this model,the knowledge graph of pathogenesis is constructed.Knowledge retrieval and knowledge query are realized in the knowledge graph of pathogenesis,and the functions of pathogenesis reasoning and syndrome prediction are realized.(5)According to the inclusion criteria,111 well-known TCM medical cases of chest impediment were screened.After standardizing the information of symptoms and signs,the application test is carried out on the syndrome differentiation model based on syndrome differentiation elements and the pathogenesis knowledge grahp based on the model of pathogenesis.In the end,the syndrome differentiation model based on the elements of syndrome differentiation is accurate in 34 cases,basically accurate in 45 cases,incorrect in 32 cases,and the accuracy rate of the model is 71.17%;the pathogenesis knowledge graph is accurate in 21 cases,basically accurate in 52 cases,incorrect in 38 cases,and.the accuracy rate of pathogenesis knowledge graph is 65.77%.The two methods have their own advantages.The dialectical model based on the dialectical elements has a higher accuracy rate,and the pathogenesis knowledge graph has a slightly lower accuracy,but the pathogenesis element reasoning process is added in the middle,which has a higher reducibility for the TCM dialectical thinking,which makes the reasoning the interpretability of the results is increased.(6)The fusion scheme is verified according to the weight ratio of the syndrome differentiation model based on syndrome differentiation elements and the pathogenesis knowledge graph of 6:4 and 4:6 respectively.In the fusion result with a weight ratio of 6:4,51 cases are dialectically accurate,35 cases are basically accurate,and 25 cases are wrong,with a total accuracy rate of 77.48%.In the fusion results with a weight ratio of 4:6,56 cases are dialectically accurate and 36 cases are basically accurate.There are 19 errors,and the total accuracy reached 82.89%,which is the scheme with the highest accuracy among several methods.Conclusion:(1)By analyzing the data structure and semantic characteristics of TCM heterogeneous data and introducing appropriate natural language processing technology,we can automatically process TCM heterogeneous data,improve data processing efficiency,reduce labor costs,and design TCM diagnosis database with reasonable structure and multi-dimensional attributes of symptoms,which can achieve data and knowledge expansion.Finally,it can provide better data support for the research of intelligent auxiliary diagnosis of TCM.(2)Through the research on the structure and semantic expression law of TCM syndrome name data,a multi-level representation framework for syndrome differentiation elements is proposed for the first time,which is especially suitable for natural language processing technology to process TCM syndrome data,and applies a pattern based on pattern matching ideas.The extraction method can process syndrome data from different sources in batches.Through the extraction of the syndrome differentiation elements of chest impediment,the multi-level representation framework of the syndrome elements is used as the label system,and the deep learning technology is used to construct a symptom-syndrome element,symptom-minimal syndrome,and symptom-syndrome multi-label classification dialectical model.The model has a good accuracy after testing,can be used for the diagnosis of chest arthralgia syndrome,and simulate the path of TCM syndrome differentiation based on the special knowledge.(3)Through the theoretical research on the relationship between TCM pathogenesis and syndrome,taking syndrome and pathogenesis data as the research object,the hierarchical framework of pathogenesis elements is designed,and under this framework,the data of pathogenesis unit in the newly revised "Chinese Medicine Clinical Diagnosis and Treatment Terminology Part 2:Syndrome" promoted by National Administration of Traditional Chinese Medicine is extracted,and the syndrome and pathogenesis database is constructed.By analyzing and mining the relationship between symptoms and signs and pathogenesis elements,the combination pattern of pathogenesis elements and the logical relationship between pathogenesis units,following the basic thinking method of TCM syndrome differentiation,applying the hierarchical framework of pathogenesis elements to design the model of pathogenesis,taking pathogenesis as the core,unifying the symptoms and syndrome information into the same framework,simulating the thinking path of general practice syndrome differentiation,The knowledge graph of pathogenesis has been constructed to realize the functions of pathogenesis reasoning and syndrome prediction.(4)The syndrome differentiation model based on syndrome differentiation elements and the pathogenesis knowledge graph designed in this study have their own advantages and disadvantages in the aspect of syndrome prediction,and both of them have a certain accuracy.The diagnosis accuracy of syndrome differentiation model based on syndrome differentiation elements is higher,but the diagnosis accuracy of some complex syndromes is not as good as that of pathogenesis knowledge graph.The knowledge graph based on pathogenesis can predict the path of pathogenesis elements,which further increases the interpretability of auxiliary diagnosis results.The two auxiliary diagnosis methods have certain complementary advantages.Furthermore,the fusion scheme based on a certain weight ratio is used for verification and testing,and the results are better than the accuracy of using a single method.Therefore,the scheme based on the integration of the two methods should be the best method in this study.