Development Of Diagnostic Scale For Phlegm-stasis Syndrome And Research And Evaluation Of Its Quantitative Diagnostic Mode

Purpose : Based on the objectification of TCM syndrome diagnosis,the demand of quantifiable research and the bottleneck problems of difficult standardized collection of TCM syndrome information in the research and development of TCM new drugs,the overall goal of this study is to complete the development and evaluation of the diagnosis scale of phlegm and stasis interaction of syndromes.Specific goals include: 1.Taking the syndrome of phlegm and blood stasis interlinking as an example,explore the development ideas and methods of the conceptual framework of TCM syndrome diagnosis scale;2.Taking the syndrome of mutual phlegm and blood stasis as an example,explore the screening methods and steps to construct the most effective diagnostic features of TCM syndromes,and explore the applicability of machine learning algorithm in the construction of quantitative diagnostic model of mutual phlegm and blood stasis and the screening of common key feature items;3.A set of systematic and complete TCM syndrome diagnosis scale development and quantitative model construction method was preliminarily formed.Material and method:1.A research group was established to preliminarily construct the theoretical framework of the diagnostic scale of phlegm and stasis mutual syndrome by means of literature research,and form the item pool of the diagnostic scale of phlegm and stasis mutual syndrome.1.1 with the help of liaoning university of traditional Chinese medicine,the APP named“ZHONG YI DIANHAI” was used as a data access to the database of ancient books,the carding and research of the literature of related estimation phlegm and blood stasis,diagnostics of TCM syndrome differentiation to diagnostics of TCM in the clinical diagnosis and treatment of TCM terms national standard part syndrome "as phlegm and blood stasis syndrome diagnostic criteria of the resources,At the same time,the monograph of TCM Syndrome,Stasis syndrome,Phlegm and stasis syndrome was used to study the diseases of phlegm and stasis.Based on the above information and the opinions of the research group,the theoretical framework of the diagnostic scale of phlegm and stasis syndrome was preliminatively constructed.1.2 select diagnostics of TCM syndrome differentiation diagnostics of TCM in the clinical diagnosis and treatment of TCM terms national standard part syndrome "as phlegm and blood stasis syndrome diagnostic criteria of the resources,at the same time combined with periodical literature about alternating knot phlegm and blood stasis syndrome in clinical research studies,with the help of the data statistics and research group discussion way to entry pool.2.Expert consultation method and clinical cross-sectional investigation method were used to carry out the selection of items in the draft of the diagnosis scale for mutual syndrome of phlegm and stasis:2.1 half open questionnaire design,by combining traditional Chinese and western medicine in liaoning province learn to phlegm and blood stasis treatment branch organization questionnaire,combining the qualitative judge and quantitative judgement method into the design of the questionnaire answers,research to early formation of the phlegm and blood stasis by each entry in the pool entry contribution to evaluate diagnostic scale,for each knot phlegm and blood stasis syndrome diagnosis scale draft entry filter to provide advice.The questionnaire design included the basic information of experts and the information of determining the symptoms and signs of phlegm and blood stasis syndrome as well as supplementary symptoms and signs.Excel is used for data management and locking,and SPSS21.0 and other statistical analysis software is used for data statistics and analysis.2.2 A cross-sectional survey was designed to collect patient data from the outpatient and ward of the Affiliated Hospital of LIAONING University of Traditional Chinese Medicine and the Second Affiliated Hospital of Liaoning University of Traditional Chinese Medicine from December 2019 to December 2021.The design of cross-sectional clinical investigation includes five parts: basic information of patients,daily living habits of patients,basic discrimination of TCM syndromes of patients,information of clinical symptoms and signs related to mutual syndrome of phlegm and blood stasis,and supplementary symptoms.Part I Basic information includes: gender,age,place of birth,nature of work,diagnosis and course of disease in Western medicine,diseases and course of disease,height,weight,blood pressure;The second part includes: smoking,drinking,eating habits,daily physical activity,staying up late,sitting for a long time,etc.The third part of TCM syndrome discrimination of patients includes: the most painful symptoms,tongue,pulse,syndrome diagnosis and so on;The fourth part included 40 items of clinical symptoms and signs related to mutual syndrome of phlegm and blood stasis,and the severity of each item was judged according to three grades of "none","light" and "severe".The fifth part mainly combined with clinical investigation to supplement the main symptoms not included in the previous study.Descriptive analysis,correlation analysis,principal component analysis,factor analysis and other methods were used to classify and analyze the items of the diagnostic scale for mutual syndrome of phlegm and stasis,and further combined with expert opinions,discrimination analysis,correlation coefficient analysis,literature comparison and other methods to conduct comprehensive screening of the items.2.3 Based on the clinical data collected in the early clinical cross-sectional survey,the reliability,validity and reactivity of the scale were evaluated.Internal consistency reliability and split-half reliability were used to evaluate the initial draft of the scale.The validity of the first draft of the scale was evaluated from two aspects: content validity and structure validity.The ability to distinguish phlegm-stasis syndrome or other syndrome types was evaluated by the degree of reactivity.Four algorithms including logistic regression analysis(LR),support Vector machine(SVM),random forest(RF)and limit gradient enhancement(XGboost)were used to carry out the preliminary evaluation of the diagnostic performance of the draft scale.The specificity,sensitivity and area under ROC curve were mainly investigated.3.The combination of traditional machine learning algorithm and deep learning algorithm was adopted to further optimize the diagnostic model items of phlegm and stasis mutual syndrome,and the optimized results of machine learning algorithm model were further used to obtain the final selected items and their weights of the diagnostic scale of phlegm and stasis mutual syndrome by logistic stepwise regression equation.ROC curve diagnostic test was used to evaluate the diagnostic performance of the scale.3.1 Whether the research object is in the relationship between phlegm and blood stasis is a dependent variable,24 core symptoms and signs are selected as characteristic independent variables,and 12 locally optimal features are selected as independent variables,and the data are fitted and analyzed by machine learning algorithm.The feature selection method is sequential traversal and jump selection.The possibility of selecting 12 features from 24 features is equal to 2.7 million feature combinations.One feature combination is selected for analysis for every 10,000 feature combinations,and the best and worst results are selected for270 times in total.Logistic regression analysis(LR),Random Forest(RF),Support Vector Machine(SVM),SVM,The Extreme Gradient Boosting(XGboost)and Convolutional Neural Network(CNN)algorithms are used to construct The diagnosis model on The training set.A cross validation method was used to optimize the parameters in the diagnostic model.Draw diagnostic grid tables of machine learning models on test sets.Five indexes,accuracy,sensitivity,missed diagnosis rate,specificity and misdiagnosis rate,were used to evaluate the performance of the model,and the optimal parameter combinations were selected according to accuracy to extract key feature items.3.2 Based on the common key feature items extracted by the 5 machine learning algorithms in3.1,SPSS21.0 software was used for statistical analysis,and the logistic stepstep regression equation was used to further optimize the items and determine the weight of the items,so as to form the final draft of the diagnostic scale for mutual syndrome of phlegm and stasis.This research has built the phlegm and blood stasis of alternating knot final version back to generation to clinical syndrome diagnosis scale cross-sectional survey to collect 866 cases of qualified,get a scale to judge by the diagnosis of patient data,the result of the scale,the discriminant result doctor made four tables,preliminary evaluation scale of diagnosis,specific degree of accuracy and sensitivity.Results:1.Based on literature research and group discussion,the theoretical framework of the diagnostic scale of phlegm and stasis mutual syndrome was initially constructed,and the original item pool of the diagnostic scale of phlegm and stasis mutual syndrome was formed.1.1 The theoretical framework of the scale for diagnosis of phlegm and blood stasis mutual syndrome is based on the overall concept and the basic principle of syndrome differentiation and treatment,including two levels of syndrome differentiation information collection approach and syndrome differentiation thinking method.It is mainly based on eight categories of syndrome differentiation and syndrome differentiation of qi,blood and body fluid,and collects syndrome differentiation diagnosis information from observation,inquiry and pulse diagnosis.1.2 reference diagnostics of TCM syndrome differentiation diagnostics of TCM in the clinical diagnosis and treatment of TCM terms national standard part syndrome "and mutual phlegm and blood stasis syndrome diagnostic criteria in the extraction of 13 signs and symptoms entries,with the help of a qualified research paper 114 extraction 38 items,finally the above51 items standardization split,merge processing,Finally,38 items of symptoms and signs were extracted as the original item pool for the development of the diagnostic scale of phlegm-stasis syndrome.2.Items were screened by combining expert questionnaire consultation and clinical cross-sectional investigation.2.1 44 qualified expert questionnaires were completed,and all the experts had professional and technical titles above deputy senior level.Descriptive analysis,discrete trend method and multidimensional scale analysis were used to comprehensively analyze the expert questionnaire consulting data,and 26 items such as chest tightness and phlegm,heavy limbs and full of blobs(chest and abdomen)were screened out.2.2 A total of 866 qualified cases were collected in the cross-sectional clinical investigation.A variety of statistical analysis methods were used to select and analyze the items.Finally,28 items including pain(tingling pain,stuffy pain,swelling pain),fixed location of pain,heavy limbs,stuffy chest,phlegm,swelling(chest,stomach,abdomen),dizziness and fatigue were retained.In the end,a total of 24 items were included in the preliminary draft of the diagnostic scale for mutual syndrome of phlegm and stasis based on the selection results of items from expert questionnaire consultation and clinical cross-sectional survey.2.3 The initial reliability,validity,reactivity and diagnostic performance of the 24-item draft of the sputum and stasis syndrome diagnosis scale were evaluated.1.Reliability evaluation:The results indicated that Cronbach’s Alpha coefficient was 0.766 and Spearman-Brown coefficient was 0.763 in the internal consistency reliability of the scale.2.Validity evaluation:Spearman correlation coefficient was used to investigate the content validity of the first draft of the scale.The correlation coefficients between the 24 items and the general table of the scale ranged from 0.086 to 0.731,with significant statistical differences(P <0.01,P < 0.05).The exploratory factor analysis showed that the KMO statistic was 0.731 and the Bartlett sphericity test P value was less than 0.01,suggesting that the factor analysis method was suitable for analysis.A total of 8 common factors were extracted and the cumulative variance contribution rate was 61.778%,which was basically consistent with the preset theoretical framework.3.Reactivity evaluation: Independent sample T test was used,and the results showed that there was no significant statistical difference between the two groups in the distinction of "pulse astringency"(P=0.135)and "pulse sink"(P=0.087),and the other 22 items and the general scale showed good reactivity between the two groups,with significant statistical difference(P<0.01).Iv.Diagnostic performance evaluation: XGboost model algorithm was used to evaluate the diagnostic performance of the scale draft,with specificity of 87.0%,sensitivity of 80.2% and area under ROC curve of 0.870.The diagnostic performance of the draft scale was evaluated by SVM linear function model algorithm.The specificity was 85.2%,the sensitivity was 76.3%,and the area under ROC curve was 0.807.The diagnostic performance of the draft scale was evaluated by SVM linear function model algorithm.The specificity was 85.2%,the sensitivity was 76.3%,and the area under ROC curve was 0.807.The RBF kernel function model was used to evaluate the diagnostic performance of the draft scale.The specificity was 88.5%,the sensitivity was 83.0%,and the area under ROC curve was 0.858.The diagnostic performance of the draft scale was evaluated by SVM polynomial kernel model algorithm,with specificity of 85.4%,sensitivity of 89.7%,and area under ROC curve of 0.875.Sigmoid function model algorithm was used to evaluate the diagnostic performance of the scale.The specificity was 88.4%,the sensitivity was 74.6%,and the area under ROC curve was 0.815.LR model algorithm was used to evaluate the diagnostic performance of the scale draft,with specificity of 83.8%,sensitivity of83.9%,and area under ROC curve of 0.839.The diagnostic performance of the draft scale was evaluated by RF model algorithm,with specificity of 87.0%,sensitivity of 81.6%,and area under ROC curve of 0.843.It is suggested that the first draft of sputum and stasis syndrome diagnosis scale has good diagnostic performance.3.The quantitative diagnostic model of mutual syndrome of phlegm and blood stasis was preliminarily constructed,and the quantitative score assignment and diagnostic threshold study of items in the diagnostic scale of mutual syndrome of phlegm and blood stasis were completed.3.1Five machine learning algorithms including LR,RF,SVM,XGboost and CNN were used to construct the quantitative diagnosis model of phlegm-stasis mutual syndrome,and five indicators including accuracy,sensitivity,missed diagnosis rate,specificity and misdiagnosis rate were evaluated for model evaluation and feature extraction.The optimal parameters of each model are as follows: the accuracy of linear SVM model is 0.85,sensitivity is 0.85,specificity is 0.86,missed diagnosis rate is 0.15,misdiagnosis rate is 0.14;The accuracy,sensitivity,specificity,misdiagnosis rate of POLYNOMIAL kernel SVM model were 0.86,0.91,0.80,0.09 and 0.20 respectively.The accuracy,sensitivity and specificity of RBF SVM model were 0.87,0.87,0.13 and 0.13 respectively.The accuracy of LR model was 0.85,sensitivity was 0.88,specificity was 0.81,missed diagnosis rate was 0.12,misdiagnosis rate was 0.19.The accuracy,sensitivity and specificity of RF model were 0.86,0.85,0.86,0.15 and 0.14 respectively.The accuracy,sensitivity,specificity,misdiagnosis rate of XGBoost model were 0.84,0.81,0.86,0.19 and 0.14 respectively.The accuracy,sensitivity,specificity,missed diagnosis rate and misdiagnosis rate of CNN model were 0.838,0.876,0.795,0.124 and 0.205 respectively.By integrating the data results of seven models of the five algorithms,20 common key features of the diagnosis of phlegm-stasis syndrome were extracted.3.2model based on machine learning to extract 20 common key characteristics,using SPSS software,the scale of quantitative diagnosis model is established by means of stepwise regression analysis to 12 items into the final model,and according to the original weight coefficient of every purpose built phlegm and blood stasis by quantitative diagnosis model,the model about an index value is 0.684,The corresponding sensitivity and specificity were81.1%,87.3% and 3.6625 respectively.The diagnostic threshold of 3.6625 was set as 20 points,which was equivalent to 9.872 times of expansion.The 12 items were expanded in equal proportion,and the integral values were retained in accordance with the rounding principle.A diagnostic scale with weight scoring function was initially constructed,with the highest diagnostic score of 76 points and the lowest diagnostic threshold of 20 points.Form the diagnostic scale of phlegm and blood stasis mutual syndrome with quantitative scoring function.3.3 The above constructed scale for the diagnosis of mutual syndrome of phlegm and stasis was subsumed into the clinical information of 866 qualified cases collected clinically,and its diagnostic performance was evaluated.The sensitivity,specificity and diagnostic accuracy of the scale were 80.33%,88.10% and 83.72%,respectively.The diagnostic performance of the scale basically reached the expected goal.Conclusion:1.The syndrome of mutual phlegm and stasis is a common modern clinical syndrome,and its main common clinical characteristics are pain,heavy limbs,dark complexion,dark or spotted mouth,lips and tongue,thick and greasy tongue coating,smooth and heavy pulse,etc.The quantitative diagnosis of the syndrome of mutual phlegm and stasis is helpful to promote the quantitative identification of the syndrome in clinical studies.2.The diagnostic scale for mutual syndrome of phlegm and blood stasis has been preliminarily developed,which has certain universality in the clinical research of mutual syndrome of phlegm and blood stasis,and provides an applicable quantitative diagnostic information collection tool for the clinical research of such TCM new drugs.3.Taking the mutual syndrome of phlegm and blood stasis as an example,the methodological thinking of quantitative diagnosis of syndrome itself under the mode of syndrome integration was established,and the quantitative diagnosis model of mutual syndrome of phlegm and blood stasis was preliminarily constructed.The artificial intelligence machine learning algorithm introduced in the study has applicability in the extraction of the commonness and key features of TCM syndrome macro diagnosis.