The Value Of Clinical-Radiomics Nomogram In Predicting Risk Stratification Of Endometrial Cancer

Objective: To develop a radiomics nomogram based on clinical and MRI features for predicting risk stratification of endometrial cancer(EC)and evaluate the diagnostic performance of nomogram for high-risk EC.Methods: A total of 143 patients with EC who underwent total hysterectomy and bilateral appendectomy were collected in our hospital from September 2018 to December 2022.The patients were divided into low-risk EC and high-risk EC by the gold standard of surgical pathological results.All patients underwent pelvic MRI before surgery.Clinical data of each patient included age,cancer antigen 125(CA125),human epididymis protein 4(HE4),preoperative diagnostic dilatation and curettage(D&C)results,MRI reported depth of myometrium invasion,cervical stromal invasion(CSI),and maximum tumor diameter were collected.Univariate and multivariate regression analyses were used to identify clinically independent risk factors.All patients were randomly divided into a training set(n=100)and a test set(n=43)in a ratio of 7:3.3D-Slicer software was used to delineate areas of interest along the tumor margins on axial T2-weighted imaging,diffusion-weighted imaging and dynamic enhanced MRI,and extract radiomics features from each sequence.Minimum redundancy maximum correlation algorithm and minimum absolute contraction and selection operator algorithm were used to screen the best radiomics features of each single sequence and combination sequence to calculate radiomics score(rad-score).Finally,the radiomics nomogram was constructed by combined with independent clinical risk factors and rad-score.A receiver operating characteristic(ROC)curve was used to evaluate the performance of each model.The clinical net benefit of the nomogram was assessed using decision curve analysis(DCA),net reclassification index(NRI),and integrated discrimination index(IDI).Results: Multivariate Logistic regression analysis found that age,MRI reported CSI and D&C were independent risk factors for distinguishing high and low risk EC(P<0.05),and the above independent risk factors were selected to construct the clinical model.The radiomics nomogram was constructed by combined with independent clinical risk factors and rad-score.The areas under the ROC curves of the clinical model,rad-score,and radiomics nomogram were 0.817(95% confidence interval [CI]: 0.737–0.896),0.823(95%CI:0.7392–0.9061)and 0.926(95%CI: 0.876–0.975)for the training set;0.765(95%CI: 0.634–0.895),0.815(95%CI: 0.5913–0.9109)and 0.892(95%CI: 0.796–0.988)for the test set,respectively.The calibration curve showed good agreement between the prediction of high-risk EC by nomogram and the actual pathological results.DCA showed that when the probability threshold was in the range of 2%–94%,the net clinical benefit from using nomogram to predict high-risk EC was greater than that obtained by the clinical model.In addition,NRIs were 0.959(95%CI: 0.608–1.310)and 0.499(95%CI: 0.121–0.619),and IDIs were 0.262(95%CI:0.160–0.365)and 0.215(95%CI: 0.168–0.415)in the training set and teat set,respectively.The differences were statistically significant(P<0.05).Conclusions: Patient age,MRI reported CSI,D&C and rad-score based on multi-sequence radiomics features can be used as independent risk factors for distinguishing high-and low-risk EC.The radiomics nomogram constructed based on independent risk factors exhibited good performance in the prediction of high-risk EC.