Ultrasonographic Features Of Benign And Malignant Thyroid Nodules And Its Logistic Regression Analysis

Objective:Using binary Logistic regression to filter out the statistically significant factors which is to establish the regression model and improve the preoperative diagnostic accuracy.Methods:539 cases of thyroid nodules removed by surgery and diagnosed pathologically were analyzed retrospectively.450 nodules were benign,89 were malignant.The ultrasonic features of thyroid nodules were collected,including boundary,extracapsular invasion,aspect ratio,shape,internal echo,echo unifomity,calcification,acoustic halo,rear acoustic attenuation,cervical lymphadenectasis,flow grade.The factor has statistics significance by χ2 test between two groups were analyzed by binary Logistic regression,after enter regression filtered out the most significant risk factors,then to establish regression model and evaluation of regression coefficients by wald χ2 test,use likelihood-ratio test to estimate the goodness-of-fit of the regression model.Assess the differentiating ability of the regression model by receiver operating characteristic(ROC).P<0.05 was considered to indicate significance.Results:With pathology as the dependent variable and the ultrasonic characteristics of thyroid nodules and the age of patients as independent variables,binary Logistic regression analysis filter out 9 statistically significant independent variables.The regression model was that [Logit(P)=-7.011-0.833×age+2.151×border+2.042×extracapsular invasion+2.631×aspect ratio+2.231×shape+2.140×extremely low echo+2.202×microcalcification+1.949×rear acoustic attenuation+1.978×cervical lymphadenectasis].Use likelihood-ratio test to estimate the goodness-of-fit of the regression model and χ2=271.714,P=0.000.So the model has statistically significance.If P>0.5 was considered malignant and P≤0.5 was considered benign,use the model to evaluate the 539 nodules in this study,the equation was P=1/[1+e-z] where z is the logit,92.8% could be diagnosed correctly. AUC of ROC was 0.948(95% confidence interval:0.927-0.969),Standard error was 0.011,P<0.05.So the binary Logistic regression model has great important mean for d i ff e r e n t i a t i n g m a l i g n a n t t h y r o i d n o d u l e s f r o m b e n i g n o n e s.Conclusion:Using binary logistic regression analysis can filter out the most significant ultrasound characteristics to disdinguish the malignant nodules from benign ones. At the same time it can deeply analysis the interaction of each factor and the importance of main parameters in diagnosis.In this study,AUC of ROC was 0.948 proved the high accuracy and the clinical utility.