Data Analysis And Empirical Study Of Risk Factors For Aortic Dissection

In clinical symptoms aortic dissection is lack of sensitivity,the rate of missed diagnosis and misdiagnosis is high.In recent years,the incidence rate of aortic dissection in China has been rised,but there is still no definite conclusion on the pathologic mechanism.The risk factors and other unknown pathogenesis of aortic dissection need to be further studied.Therefore,the risk factors of aortic dissection and the prediction methods of aortic dissection are analyzed and studied in this paper.This study collected 157 cases of patients with aortic dissection and 157 cases of patients without aortic dissection in a hospital of Kunming.First,data preprocessing.Research data is divided into two groups,157 cases with aortic dissection as research group,157 cases without aortic dissection patients as control group,discussing with the clinical expert,collecting general information and history clinical data of aortic dissection patient.Missing data is filled by the average method,drawing stacking histogram,box diagram,statistic frequency table,and making descriptive statistical analysis of data.In this study,the number of patients(157 cases)is nearly 10 times of the number of variables(16 cases),the sample size is basically sufficient.Using univariate analysis processing the data of 16 factors,eliminates five factors of no statistical significance,remaining 11 factors(i.e.,hypertension,atherosclerosis,marfan syndrome,white plug's disease,aortic valve abnormalities,smoking,drinking alcohol,coronary heart disease,diabetes,simple renal cyst,aortic aneurysm)as the basis of further research.Second,establishing a Logistic regression model to analyze 11 variables,with statistical significance(P < 0.05)in the six risk factors: high blood pressure(OR = 5.246),atherosclerosis(OR = 4.296),marfan syndrome(OR = 4.284),white plug's disease(OR = 22.271),smoking(OR =1.318),alcohol(OR = 2.291).Third,using correlation analysis model analysis 11 variables,to obein 10 risk factors:hypertension,atherosclerosis,aortic valve abnormalities,simple renal cysts,coronary heart disease,white's disease,smoking,drinking,history of marfan syndrome,aortic aneurysm.Correlation analysis results including six risk factors for logistic regression.Two models have controversial of aortic valve malformation,coronary heart disease,simple renal cysts,the influence of the history.Considering the above two models is mainly in view of the data have the characteristics of linear,but the data of this study has certain characteristics of nonlinear,so this article then nonlinear improvements were made on the model,to differentiate and analyze the dangerous factor of controversial further.Fourth,in view of the aortic dissection of nonlinear correlation between multiple risk factors,the introduction of improved RBF kernel function is to establish a nonlinear relation between logistic regression model,get eight risk factors: hypertension,white plug's disease,marfan syndrome,smoking,drinking,history of atherosclerosis and aortic aneurysm,simple renal cyst.Send get eight risk factors of clinical expert group discussion,the panel think improve logistic regression with eight risk factors and clinical experience doubt of risk factors for high consistency.Refer to the result of literature also confirmed that the improvement of logistic regression to get eight risk factors is relatively accurate.Fifth,using eight risk factors from improve logistic regression to study of 314 samples data,to establish the SVM,LSSVM model to predict the prevalence of aortic dissection.Selecting ten cross validation method to test and accuracy of the SVM model was 70.06%,LSSVM accuracy of85.99%,LSSVM forecasting model can more accurately found that patients with aortic dissection.This study selects eight risk factor for aortic dissection,aortic dissection disease prediction model is established,the early focus on high-risk groups for clinicians,theoretical basis is provided for early found early treatment of aortic dissection.