Research And Application Of Intelligent Diagnosis Of Disease And Syndrome Based On Data Fusion Of Traditional Chinese Medicine And Western Medicine

PurposesFacing the clinical diagnosis of Traditional Chinese Medicine（TCM）,with the data of the four diagnosis of TCM as the core content,based on the integrated data of TCM and Western medicine,to explore the methods of intelligent diagnosis of disease and syndrome and health status identification.（1）Based on qualitative data of TCM and Western medicine,the correlation between symptom and Western medicine index was analyzed quantitatively,and explore the connotation of the relationship between symptom and index of Western medicine from the data level.（2）Based on the quantitative data of TCM and Western medicine,explore the correlation between tongue data,pulse data and Western medicine index data in different health states,so as to provide basis for intelligent diagnosis of disease and syndromes.（3）Based on the qualitative and quantitative fusion data of TCM and Western medicine,the diagnosis models of disease and syndrome were explored to provide technical support for the modern accurate diagnosis.Methods:This study was divided into four parts.（1）Based on the qualitative and quantitative data of TCM and Western medicine,My SQL and Mongo DB were used to build the data platform.The main data included 20 items of questions in the Health Status Assessment Questionnaire Scale,254 items of symptoms in the Information Record Form of Clinical Four Diagnostics of TCM,240 items of indexes of health examination items in Western medicine,as well as the original information of the tongue and sphygmogram,and their corresponding analysis data.According to the health assessment methods and diagnostic criteria of diseases,using Python to realize automatic health status identification and diagnosis of disease.（2）Based on the macroscopic qualitative indexes,an improved node contraction method was used to screened out the core symptom and Western medicine index of sub-health fatigue people（n=361）,patients with fatigue disease（n=1529）and Non-small Cell Lung Cancer（NSCLC）patients（n=533）,and the core symptom network,Western medicine index network and symptom-index correlation network were constructed to analyze the correlation between symptom and Western medicine index in different populations.（3）Based on the objective and quantitative data of tongue and pulse and Western medicine,multivariate statistical method was used to analyze the characteristics of tongue and pulse data of healthy people（n=250）,sub-healthy people with fatigue（n=361）,patients with fatigue（n=1529）and patients with NSCLC（n=533）,the correlation among tongue and pulse and Western medicine indexes was analyzed to explore the correlation strength between the data of TCM and Western medicine.（4）For macroscopic qualitative data combined with standardized quantitative data,using machine learning method,decision tree,logistic regression,support vector machine,random forest and neural network methods to establish diagnosis models of disease and syndrome based on the tongue,pulse,symptoms and index of Western medicine,the diagnostic models of sub-healthy people with fatigue（n=242）and patients with fatigue（n=244）,NSCLC of Qi deficiency syndrome（n=163）and Yin deficiency syndrome（n=174）were established.Results（1）The complex network method was used to screen out the core symptoms and indexes of Western medicine of fatigue people and NSCLC patients,and the core symptom network,the core index network and the symptom-index correlation network were constructed.（2）The multivariate statistical analysis based on the data of tongue and pulse of the fatigue population showed that:the tongue and pulse of the group of sub-health fatigue had no canonical correlation,the data of tongue and pulse of the group of disease fatigue had a canonical correlation,and the correlation coefficient of the tongue and pulse was0.627（P<0.05）.In the group of disease fatigue,the important parameters in tongue diagnosis were TC-Cr,TC-Cb,TB-B,TB-Cb,per Part,TC-B and TC-H,and the canonical load coefficients were 0.353,-0.349,0.347,-0.344,0.344,0.340 and 0.332,respectively,P<0.05.The important parameters in pulse diagnosis were t₅,w₂/t and t₄,and the canonical load coefficients were 0.387,-0.271 and 0.268,respectively,P<0.05.（3）Multicultural statistical analysis showed that tongue and pulse data of NSCLC patients were correlated with blood routine indexes,biochemical indexes and tumor markers to varying degrees.There were many tongue features that were strongly correlated with tumor marker indexes.The tongue features of NSCLC patients（TB-Cr,TC-CON,per All,TC-ENT,TC-MEAN）were more closely correlated with carbohydrate antigen 242（CA242）and cytokeratin 19 fragment.The parameters of tongue body were closely related to the fragment of multicellular keratin 19,and the parameters of tongue coating were closely related to the CA242.（4）Classification model of disease fatigue and non-disease fatigue based on machine learning method showed,except random forest,the classification efficiency order of models based on different data sets was as follows:tongue<pulse<tongue&pulse<tongue&pulse&BMI.In the“tongue&pulse”data set,the classification efficiencies of neural network and logistic regression were better than other models.（5）The classification model of Qi deficiency syndrome and Yin deficiency syndrome of NSCLC based on machine learning method showed that classification efficiency order of models based on different data sets was as follows:tongue&pulse<symptom<tongue&pulse&symptom.Among them,the Support Vector Machine,（SVM）model had a better classification performance for the“symptom”data set,its area under curve（AUC）was 0.9321,the logistic regression model had a good classification performance for the data set of“tongue&pulse”,its AUC was 0.8022,the neural network model had a good classification performance for the data set of“tongue&pulse&symptom”,and its AUC was 0.9401.Conclusions（1）An intelligent health status identification method was constructed based on the data of TCM and Western medicine.（2）Based on the qualitative data of macro-symptom indexes,using complex network technology to screen out the core symptoms and core Western medicine indexes of fatigue people and patients with NSCLC,and construct a core symptom/index network,core symptoms-index correlation network,which provided a new methodological reference for identifying diseases from the perspective of TCM integrated with Western medicine.（3）Based on the quantitative data of tongue and pulse and Western medicine indexes,the multivariate statistical method could be used to explore the correlation strength between the data of tongue and pulse and between the data of tongue and pulse and Western medicine indexes of different health states.（4）Based on the qualitative together with quantitative data,the fatigue diagnosis model based on symptom,data of tongue and pulse,Western medicine index,and the TCM syndrome diagnosis model of NSCLC were established by using machine learning methods,providing a new method for intelligent diagnosis of disease and syndrome based on the data of TCM and Western medicine.