| Objective:To evaluate the efficacy of two-dimensional(2D),three-dimensional(3D),3D+2D prediction models for platinum-resistance of advanced serous ovarian cancer(HGSOC)based on plain computed toography(CT)radiomics,and to compare them horizontally.Method:105 cases of advanced serous ovarian cancer in three Grade A hospitals were retrospectively analyzed.According to platinum-resistant criteria,patients with recurrent ovarian cancer were divided into platinum-resistant group(n= 35)and nonplatinum-resistant group(n=70).2D and 3Dwere delineated on plain CT images,and texture features were extracted.Cases were randomly divided into training group(n=74)and testing group(n= 31)in a ratio of 7:3.The Least absolute shrinkage and selection operator(LASSO)was used to construct the radiomics nomograms model and draw receiver operating characteristic(ROC)curve.Calibration curve and decision curve analysis(DCA)evaluated and verified the results of radiomics nomograms,and compared the difference of 2D,3D and 2D+3D diagnostic performance.Results:1339 quantitative radiomics feature parameters were extracted from 2D group and 3D group respectively.In 2D,3D,3D+2D training and testing groups,the AUC of clinical model was significantly different from that of Normograph model.The AUC of clinical model was 0.83(95%CI:0.74-0.93)in 2D training group and 0.79(95%CI:0.68-0.90)in 3D training group.The AUC of the 3D+2D training group was0.79(95%CI:0.70-0.89).The AUC of Normograph imaging model was 0.92%(95%CI: 0.86-0.98)in 2D training group,0.92%(95%CI: 0.83-1.00)in 3D predictive training group,and 0.93%(95%CI: 0.87-0.99)in 3D+2D predictive training group.Delong’s tests showed statistically significant differences between the clinical model and the Normograph model in the 2D and 3D training groups.Area under the curve(AUC)of 2D training group and testing group were 0.92%and 0.84%,respectively.The AUC of platinum resistance predicted by 3D radiomics normogram was 0.92% in training group and 0.82% in test group,respectively.The AUC of platinum resistance predicted by 3D+2D radiomics normogram was 0.93% in the training group and 0.90% in the test group,respectively.Delong’s tests showed that the 3D+2D radiomics group showed better predictive performance than others model(P=0.001 in the training group and P=0.008 in the test group,both < 0.05).Conclusion:In establishing a model for predicting platinum resistance of advanced HGSOC,the radiomics nomomodel is more valuable than the clinical model.In the horizontal comparison of 2D,3D and 3D+2D radiomics normogram models,3D+2D radiomics normogram models further improved the accuracy of prognosis prediction of platinum resistance in patients with advanced HGSOC. |
