Objective: To explore the predictive value of nomogram model based on ultrasonic characteristics for Ki-67 expression in phyllodes tumor of breast (PTB).Methods: The ultrasonic imaging data of 119 cases of PTB confirmed by pathology were analyzed retrospectively

82 cases were divided into low expression group (Ki-67≤10%) and 37 cases were high expression group (Ki-67>10%) according to the expression level of Ki-67.The features of the two groups were compared and analyzed

including maximum diameter

aspect ratio

shape

edge

parenchymal echo

cystic change

calcification

rear echo

blood flow signal

resistance index and peak systolic flow rate.Independent predictors of Ki-67 high expression were selected by univariate analysis and multiple Logistic regression analysis

and a joint nomogram model was constructed based on the independent predictors. Bootstrap method was used for internal verification and calibration curve was used to judge the goodness of fit of the model. Receiver operating characteristic curve (ROC) was drawn to evaluate the predictive efficacy of the model

and the clinical benefit of the model was analyzed by decision curve.Results: The maximal diameter

cystic change

blood flow signal and systolic peak velocity were correlated with the expression level of Ki-67 in PTB patients (P<0.05)

among which the maximal diameter (OR=0.642

95% CI: 0.295 ~ 0.993)

cystic change (OR=0.078

95% CI: 0.008～0.808)and systolic peak flow rate (OR=1.404

95%CI: 1.106～1.782) were independent predictors of Ki-67 high expression level (P<0.05).The independent predictors quantified by the nomogram model showed a good agreement between the prediction effect and the actual results

and the agreement index (C-index value) was 0.814 (95%CI: 0.710 ~ 0.913).The ROC curve analysis showed that the combined prediction probability model efficiency of independent predictors [area under the curve (AUC) =0.812] was better than that of the maximum diameter (AUC=0.696)

sac change (AUC=0.705) and peak systolic flow rate (AUC=0.736)

and the clinical benefit was higher when the threshold was 0.10 ~ 0.48.Conclusion: The nomogram model based on ultrasound features can be used to predict the level of Ki-67 expression in PTB patients

and the model has high accuracy and reliability.