Multiparameter MRI Radiomics Model Predicts Progression-Free Survival In Epithelial Ovarian Cancer

Objective: This study investigated the predictive value of multiparameter magnetic resonance imaging(mp-MRI),combined clinical data and mp-MRI image parameters for progressionfree survival(PFS)in epithelial ovarian cancer(EOC).Materials and methods: Patients with epithelial ovarian cancer undergoing mp-MRI in our hospital from January 2015 to January 2020 were collected for a retrospective study and were randomly divided into training and validation groups in a 7:3 ratio.All patients underwent total double adnexectomy plus omenectomy and pelvic lymph node dissection after mp-MRI.The advanced artery fat-suppressed T2-weighted imaging(fs T2WI),diffusion weighted imaging(DWI)and dynamic contrast enhanced-magnetic resonance imaging(DCE-MRI)images of each patient before surgical operation were manually delineated as region of interest(ROI)using ITK-snap software.Then,1316 quantitative features are obtained form ROI using Py Radiomics.The 10-fold cross-validation of Least absolute shrinkage and selection operator(LASSO)methods and COX regression mode were used for feature screening.Clinical independent predictors were screened using univariate and multivariate COX risk regression.Fainlly,The fs T2 WI model,the DWI model,the DCE model,the combined fs T2WI-DWI-DCE model and the hybrid imaging histology-clinical model were developed.Radiomics scores were calculated for each patient,patients were categorised into high and low risk groups,and a nomogram was created.The predicivet efficacy of the models was assessed comparatively using the receiver operating charactsteristic(ROC)curve,decision curve,and calibration curve.Kaplan-Meier curves were applied to compare PFS across risk groups.All statistics were analysed using R language and P<0.05 was considered statistically significant.Results: A total of 139 patients were included in this study,and 84 patients had disease progression.Among the radiomics models,the combined fs T2WI-DWI-DCE model had higher efficacy than the T2 WI model,DWI model and DCE model alone,with an AUC of0.932 in the training group and 0.917 in the validation group.The mixed model consisting of histological features and clinical data(histologic subtypes,pathological grade,FIGO stage,age,CA125,HE4,and ADC values)performed best in predicting PFS,with AUCs of 0.947 and 0.933 respectively.Both the decision curve and the calibration curves showed a good diagnostic efficacy of the combined model.The survival curves showed that the high-risk group had PFS was lower than that of the low-risk group.According to the cut-off value of the comble model,the high-risk group and the low-risk group of patients were trained.And the survival curve showed that the PFS of the high-risk group was lower than that of the low-risk group,and the difference was statistically significant.Moreover,patients with higher Rad score,type II,high grade,advanced stage,advanced age,high levels of CA125 and HE4,and smaller ADC values had shorter PFS.The rate of disease progression increased over time,from 15.8% at year 1 to half at 3 years,and only 11.5% at 5-year progression-free survival.Conclusion: The hybrid model constructed from radiomics and clinical data has shown excellent performance in predicting PFS in EOC patients.The nomogram provides a non-invasive diagnostic tool for clinical risk stratification of patients.