The Value Of Conventional Ultrasound And Ultrasonic Elastography In Differential Diagnosis Of Sotf Tissue Masses Using Logistic Regression Analysis

Background: This study was performed to investigate the value ofhigh-frequency grayscale，color Doppler ultrasound and ultrasonic elastographyof soft tissue masses in differential dianosis of soft tissue masses using Logisticregression model.Methods: The features of gray scale ultrasonography (US), colorDoppler flow imaging (CDFI) and ultrasonic elastography (UE) wereevaluated in113soft tissue masses confirmed by surgical pathology. ALogistic model for predicting malignancy of soft tissue masses on the basis ofultrasonographic features was obtained. A receiver operating characteristiccurve (ROC) was used to assess the performance of the Logistic model.Results: There were a total of59benign and54malignant soft tissuemasses. According to the output results of SPSS17.0,there were fourindependent variables selected on the final step of Logistic regressionanalysis,including:the bone destruction,T/L,the two-dimensional Adler gradingof the blood flow and UE scores(p <0.05).The model was: Logit(P)=-3.704+1.722×T/L(1)+0.769×T/L(2)+2.372×bone destruction-1.578×Adlergrading（1）+0.948×Adler grading（2）+0.799×Adler grading（3）+2.736×UEscores（1）+4.526×UE scores（2）When the regression value P was more than0.5,the prediction wasmalignant tumor;When the regression walue P was less than or equal to0.5,theprediction was benign tumor.A receiver operator characteristic cueve(ROC)was constructed to assess the performance of the Logistic model.The area underROC curve was0.929of a high accuracy.The95%confidence limit was0.883-0.976,standard error wes0.024(P<0.001).It gave a sensitivity and specificity of87.04%and91.53%,positive prediction value was90.38%,negative prediction value was88.52%.Conclusion: The accuracy of the multivariable Logistic regression modelprediction appears to be better than the univarialbe in equation,which had asignificant differences(P<0.001). A receiver operator characteristic cueve(ROC)was constructed to assess the performance of the Logistic model.The area underROC curve was0.929of a high accuracy.It suggested that this Logistic modelcan accuratatly separate malignant from benign soft tissue masses.It hassignificant value of the coalition diagnosis based on conventional ultrasoundand ultrasonic elastography in differentiation between benign and malignantsoft tissue masses.