To Investigate The Value Of Routine And Contrast-enhanced Ultrasound In Differential Diagnosis Of Salivary Gland Lumps

Objective:To establish property prediction model of salivary gland lumps,and evaluate the value of RUS and CEUS in the identification of salivary gland lumps.Method:Cases with salivary gland lumps met the inclusion criteria,and confirmed by pathology in Sichuan Provincial People’s Hospital from January 2009 to December 2019 were collected as the training set.All patients had complete ultrasound image data,were divided into RUS group and CEUS group according to whether CEUS was performed.Retrospective analysis of clinical data and ultrasound image characteristics were performed,with the pathology as the gold standard,establishing prediction model Ⅰbased on RUS and prediction model Ⅱ based on RUS and CEUS by single factor and dichotomous logistic regression analysis,and evaluating diagnostic value of the two models.Collecting cases confirmed by pathology and underwent CEUS from January 2020 to December 2020 as the test set,and all met the inclusion criteria.To verify the diagnostic efficacy of model Ⅱ by analyzing and extracting the ultrasonic image features,and further evaluate the value of ultrasound in the diagnosis of salivary gland lumps.Result:696 cases(713 lumps)in the RUS group,including 595 benign(about 83.5%)and 118 malignant lumps(about 16.5%).The model Ⅰ was Logit(Y)=0.027*age+capsule(2.056*incomplete/2.943*no capsule)+cervical lymph node(-0.887*reactive hyperplasia/2.793*ipsilateral single metastasis/0.888*ipsilateral multiple or bilateral metastasis)+ 1.326*boundary angulation+3.157*boundary burr-4.331,Y was the predictive value.Statistical analysis showed that the optimal cutoff value of Y was 0.224(Y>0.224 was predicted to be malignant),the AUC was 0.858(0.830-0.883),the predictive sensitivity was 72.0%(63.0%-79.9%),specificity was 92.4%(90.0%-94.4%),and the Youden index was 0.645.CEUS group included 490 cases(all included in the RUS group),totally 498 lumps,with 412 benign(about 82.7%)and 86 malignant lumps(about 17.3%).The model Ⅱ was Logit(Y)=rear echo(0.049*no change/3.181*attenuation)+1.400*chaotic and uneven internal echo+ 1.306*boundary angulation+enhanced intensity(-17.933*no enhanced/-1.564*low enhanced/-1.857*iso-enhanced)+2.917*size increased after enhancement+1.663*no enhancement ring-3.416,the optimal cutoff value of Y was 0.123(Y>0.123 was judged to be malignant),with the AUC was 0.945(0.915-0.966),the predictive sensitivity was 86.9%(75.8%-94.2%),the specificity was 89.6%(85.5%-92.9%),and the Youden index of 0.765.74 cases were included in the validation set,totally 75 masses.The verification test proved that the diagnostic accuracy,sensitivity,specificity,PPV and NPV of the model Ⅱ were 89.3%,81.8%,90.6%,60.0%and 96.7%respectively.Conclusion:1.Multivariate analysis based on ultrasonic image features had high value in differentiating the properties of salivary gland masses,and the diagnostic accuracy was higher than previous single-factor studies.2.Compared with previous studies,the diagnostic sensitivity of RUS in this study was improved,but was relatively low due to the inherent pathological characteristics of salivary gland tumors.3.The combination of CEUS and RUS improved the diagnostic sensitivity of RUS,showed high diagnostic efficacy in assessing the properties of salivary gland lumps,and excelled in differentiating the pathological types of common tumors(e.g.PA and WT).