Value of multiparametric MRI-based radiomics in predicting HER2 expression in bladder cancer
10.3969/j.issn.1009-8291.2025.08.005
- VernacularTitle:基于多参数核磁共振成像的影像组学在预测膀胱癌人表皮生长因子受体2表达状态中的应用价值
- Author:
Tonglei ZHAO
1
;
Weipu MAO
;
Zihui ZHAO
;
Zejun WANG
;
Dakun ZHANG
;
Ming CHEN
;
Jianping WU
Author Information
1. 东南大学医学院,江苏 南京 210003
- Publication Type:Journal Article
- Keywords:
bladder cancer;
urothelial carcinoma;
multiparametric magnetic resonance imaging;
radiomics;
human epidermal growth factor receptor 2
- From:
Journal of Modern Urology
2025;30(8):662-670,688
- CountryChina
- Language:Chinese
-
Abstract:
Objective To evaluate the value of radiomics models and comprehensive models based on multiparametric magnetic resonance imaging(mpMRI)in predicting the expression of human epidermal growth factor receptor 2(HER2)in bladder cancer(BCa).Methods A total of 76 pathologically confirmed BCa patients undergoing pelvic mpMRI during Jan.2022 and Nov.2024 at the Affiliated Zhongda Hospital of Southeast University were retrospectively included.After the volume of interest(VOI)was sketched,7 modal features were obtained,including T2WI,DWI,DCE,T2WI+DWI,T2WI+DCE,DWI+DCE,and T2WI+DWI+DCE,which were analyzed with logistic regression to obtain the predictive values.After that,the best sequences were screened with receiver operating characteristic(ROC)curves,and then combined with support vector machine,logistic regression,K-nearest neighbor,plain Bayes and adaptive enhancement,to construct the radiomics prediction models.Logistic regression analysis was used to screen the predictors of high HER 2 expression and to construct a comprehensive prediction model and a nomogram.Finally,decision curve analysis(DCA)was used to quantify the clinical benefits.Results Among the radiomics models based on the T2WI+DWI+DCE sequence,the AdaBoost model demonstrated the best predictive performance,with area under the ROC curve(AUC)being 0.863(95％CI:0.807-0.920)in the training set and 0.716(95％CI:0.601-0.830)in the validation set.Based on radiomics features and clinical imaging characteristics,logistic regression analysis identified tumor pedicle and risk group as the predictors of high HER2 expression.The comprehensive prediction model based on the two factors achieved the AUC of 0.869(95％CI:0.772-0.965)in the training set and 0.875(95％CI:0.712-0.986)in the validation set.Conclusion The radiomics model based on the T2WI+DWI+DCE sequence has high accuracy in predicting HER2 expression,outperforming single-sequence models.The nomogram based on the comprehensive prediction model has high clinical decision-making efficacy and is useful for non-invasive identification of HER2 expression.
