Nomogram model based on ultrasonic radiomics features for preoperative differentiation of early and mid-late stage cervical squamous cell carcinoma
Objective To explore the value of nomogram model based on transvaginal ultrasonic radiomics features for preoperative differentiation of early and mid-late stage cervical squamous cell carcinoma.Methods A total of 227 patients with cervical squamous cell carcinoma confirmed by postoperative pathology were retrospectively collected.3D-Slicer software was used to delineate lesion ROI on preoperative transvaginal ultrasound images.Imaging features within ROI were extracted,redundancy analysis,least absolute shrinkage and selection operator(LASSO)as well as ten-fold cross-validation were performed to screen important radiomics features.Then radiomics model was constructed,and Radscore was calculated.Finally a nomogram model was established combining with clinical data and Radscore using multivariate logistic regression.The area under the receiver operating characteristic curve(AUC)for preoperative differentiation of early and mid-late stage cervical squamous cell carcinoma of radiomics model and nomogram model were compared.Calibration and decision curve analysis(DCA)of nomogram model were assessed.Results Eighteen radiomics features were finally enrolled,and a radiomics model for preoperative differentiation of early and mid-late stage cervical squamous cell carcinoma was constructed.AUC of this model in training set and validation set was 0.839 and 0.744,respectively.A nomogram model was constructed combining with age,number of abortions and Radscore,with AUC of 0.882 and 0.773 in raining set and validation set,respectively.DeLong test showed that AUC of these 2 models were significantly different in training set(P＜0.05).Hosmer-Lemeshow test indicated that the calibration degree of the nomogram model was good in both training set and validation set(x2=5.053,7.063,both P＞0.05).Decision curve analysis(DCA)showed that the net benefit of this nomogram model was relatively high in threshold of 0.01-1.00.Conclusion Nomogram model established based on transvaginal ultrasonic radiomics features could effectively distinguish early and mid-late stage cervical squamous cell carcinoma preoperation.
万方数据
维普
