MRI-based Radiomics For Pretreatment Prediction Of Efficacy Of Neoadjuvant Chemotherapy And Prognosis In Osteosarcoma

BackgroundWith the introduction of neoadjuvant chemotherapy,the five-year survival rate of patients with localized osteosarcoma has significantly improved.However,there are still some patients with poor histologic responses to NAC,and patients with disease progression still have poor prognosis.If patients with poor response to neoadjuvant and poor survival could be identified before treatment,personalized treatment plans could be helpful for decision support for these patients.ObjectiveTo develop and validate the radiomics models based on MRI for pretreatment prediction of efficacy to neoadjuvant chemotherapy and progression free survival(PFS)in patients with osteosarcoma.Methods1.A total of 102 patients with histologically confirmed osteosarcoma who received NAC before treatment from 4 hospitals were enrolled in this study.Patients were divided into training(n=68)and validation(n=34)cohorts according to the hospital they were recruited from.Quantitative imaging features were extracted from contrast-enhanced fat-suppressed T1-weighted images(CE FS T1WI).Four classification methods,i.e.,the least absolute shrinkage and selection operator logistic regression(LASSO-LR),support vector machine(SVM),Gaussian process(GP)and Naive Bayes(NB)algorithm,were used to select radiomics features and construct the radiomics signature.The predictive performances of the radiomics signatures were assessed with the area under the receiver operating characteristics(ROC)curve(AUC),calibration curve,and decision curve analysis(DCA).2.We retrospectively enrolled 109 patients with histologically confirmed osteosarcoma who underwent follow-up from 4 hospitals(63 in the primary cohort and 46 in the external validation cohort).A total of 1130 radiomics features were extracted from CE FS T1WI.Feature selection was based on Pearson correlation coefficient analysis and the Relief algorithm.Radiomics signatures for predicting PFS were then constructed using the selected features.Univariate and multivariate Cox proportional hazards model and Kaplan-Meier analysis were used to determine the association of the radiomics signature and clinicopathological variables with PFS.Combining clinical risk factors,a radiomics nomogram combining the radiomics signature and clinicopathological variables was developed to predict PFS,and the patients were divided into high-and low-risk groups according to their nomogram score.The predictive performance of the radiomics nomogram was assessed with respect to calibration,discrimination,and clinical usefulness.Survival analysis was performed by Kaplan-Meier curves.Results1.Thirteen radiomics features significantly associated with the pathologic response were selected based on the Least absolute shrinkage and selection operator logistic regression(LASSO-LR)method and were adopted to construct the radiomics signature.The prediction model achieved the best performance between good and poor responders with AUC of 0.882(95%CI,0.837-0.918)in the primary cohort.Calibration curves showed good agreement.Similarly,findings were validated in the external validation cohort with good performance(AUC,0.842[95%Cl,0.793-0.883])and good calibration.DCA analysis confirmed the clinical utility of the selected radiomics signature.2.Seven features significantly associated with PFS were selected to construct the radiomics signature(P<0.001).The radiomics nomogram,incorporating radiomics signature and tumor location,exhibited better prediction performance for PFS than the clinical model or radiomics signature,with the C-index of 0.845(95%CI,0.782-0.907)and 0.759(95%Cl,0.674-0.844)in the training and validation cohorts,respectively.The radiomics nomogram-defined high-risk group had a shorter PFS than those in the low-risk group(P<0.05).DCA analysis confirmed the clinical utility of the radiomics nomogram.ConclusionMRI-based radiomics model could be used as a novel non-invasively novel tool for pretreatment prediction of response to NAC and PFS in patients with osteosarcoma,which is expected to be applied in clinical practice to assist in the formulation of individualized treatment plan.