Preliminary Study Of Multimodal Ultrasound In Predicting The Prognosis Of Salivary Gland Tumors With KI-67

ObjectiveTo establish a KI-67 prediction model for salivary gland tumours based on multimodal ultrasound to assess its diagnostic value.MethodsCases were collected of salivary gland tumors operated at the UESTC Hospital-Sichuan Provincial People’s Hospital from January 2018 to October 2022 with preoperative conventional ultrasound and ultrasonic contrast.And all cases were pathologically confirmed and IHC was performed.Predictive models were established by mono-factorial and multi-factorial logistic regression analyses of conventional ultrasound and ultrasonic contrast characteristics of the tumours according to KI-67values for low risk,medium-to-high risk groups.Model Ⅰ is a two-dimensional grey-scale data prediction model;model Ⅱ is a qualitative ultrasonic contrast index prediction model;model Ⅲ is a multimodal ultrasound prediction model;model Ⅳ is a two-dimensional grey-scale ultrasound combined with pathological benignity and malignancy prediction model,and model Ⅴ is a multimodal ultrasound combined with pathological benignity and malignancy prediction model.ResultsA total of 125 cases are included in the study,75 in the low-risk group and 50 in the medium-to-high risk group.The indicators with statistical significance in the mono-factorial analysis were included in the multi-factorial logistic regression analysis.The model Ⅰ regression equation is:Logistic(Y_KI-67)=2.917*2D boundary（indistinct）+1.460*2D morphology（irregular）,with a sensitivity of 68.0%,specificity of88.0%,Youden’s index of 0.560,and area under the curve（AUC）of 0.805;the model Ⅱ regression equation is:Logistic(Y_KI-67)=-3.645+2.906*CEUS border（indistinct）+1.688*CEUS morphology（irregular）,with a sensitivity of 80.0%,specificity of 90.7%,Youden’s index of 0.707 and area under the curve（AUC）of 0.904;model Ⅲ regression equation is:Logistic(Y_KI-67)=2.557*CEUS boundary（indistinct）,with a sensitivity of80.0%,specificity 90.7%,Youden’s index 0.707,area under the curve（AUC）0.916;model Ⅳ regression equation is:Logistic(Y_KI-67)=3.409*benign and malignant（malignant）+1.968*2D boundary（indistinct）,with a sensitivity of 96.0%,specificity of74.0%,Youden’s index of 0.707,area under the curve（AUC）0.911;model Ⅴ regression equation is:Logistic(Y_KI-67)=3.263*benign malignant（malignant）+2.710*CEUS boundary（indistinct）,with a sensitivity of 96.0%,specificity of 85.3%,Youden index of 0.813 and area under the curve（AUC）of 0.953.Conclusion1.This study illustrates that the ultrasound prediction model can effectively predict the level of KI-67 expression in salivary glands.2.The ultrasonic contrast prediction model has greatly improved the diagnostic capability compared with the two-dimensional grey-scale ultrasound model.3.The inclusion of pathological benignity and malignancy brings a greater increase in the sensitivity of the prediction model,which can reduce postoperative recurrence or metastasis in patients and avoid delays in the disease.