| Clear cell Renal Cell Carcinoma(ccRCC)is the malignant tumor with the highest incidence of kidney,accounting for about 70% to 80% of Renal Cell Carcinoma(RCC);its early clinical symptoms are atypical,and it has usually progressed to the middle and late stages when diagnosed.The main treatment for ccRCC is radical resection,but one-third of cases after surgery will have tumor recurrence or progression.Studies have shown that tumor necrosis,stage and grade are important prognostic indicators of ccRCC and can reflect the potential aggressiveness of ccRCC.Radiomics is an emerging technology that organically integrates big data technology and medical imaging-assisted diagnosis in recent years.It mines massive image features that can quantify lesions from image big data,thereby constructing radiomics signature,in order to analyze images and tumor genes,the potential relationship between pathology and prognosis to fully reflect tumor heterogeneity.At present,there is no research on the preoperative prediction of ccRCC tumor necrosis,stage and grade by CT-based radiomics.Therefore,this project aims to develop and validate a CT radiomics signature for the preoperative prediction of tumor necrosis,stage and grade in patients with ccRCC in multicenters,providing evidence for clinicians' preoperative decision-making,and assisting clinicians in targeting ccRCC patients to develop personalized treatment plans.The specific methods and results obtained include the following two parts:In part 1 of this study,132 patients with pathologically confirmed ccRCC in a center were enrolled as a training cohort,while 123 ccRCC patients from another center served as the independent validation cohort.Radiomics signatures were extracted from corticomedullary and nephrographic phase contrast-enhanced computed tomography(CT)images.A radiomics signature based on optimal features selected by consistency analysis and the least absolute shrinkage and selection operator(LASSO)was developed.An image features model was constructed based on independent image features according to visual assessment.By integrating the radiomics signature and independent image features,a radiomics nomograph was constructed.The predictive performance of the above models were evaluated using receiver operating characteristic(ROC)curve analysis.Furthermore,the nomogram was assessed using calibration curve and decision curve analysis.Thirty-seven features were used to establish a radiomics signature,which demonstrated better predictive performance than did the image features model constructed using tumor size and intratumoral vessels in the training and validation cohorts(p<0.05).The radiomics nomogram demonstrated satisfactory discrimination in the training(area under the ROC curve(AUC)0.93 [95% CI 0.87–0.96])and validation(AUC :0.87,95%CI :0.79–0.93)cohorts and good calibration(Hosmer-Lemeshow p<0.05).Decision curve analysis verified that the radiomics nomogram had the best clinical utility compared with the other models.In part 2 of this study,330 patients with ccRCC from three centers were classified into the training,external validation 1,and external validation 2 cohorts.Through consistent analysis and LASSO,a radiomics signature was developed to predict the risk groups of stage and grade including low-risk group(score 0-3)and intermediate to high-risk group(score ?4).An image feature model was developed according to the independent image features,and a fusion model was constructed integrating the radiomics signature and the independent image features.Furthermore,the predictive performance of the above models for the stage and grade risk groups were evaluated with regard to their discrimination,calibration,and clinical usefulness.An radiomics signature consisting of sixteen relevant features from the nephrographic phase CT images achieved a good calibration(all p<0.05)and favorable prediction efficacy in the training cohort(AUC: 0.940,95% CI: 0.884–0.973)and in the external validation cohorts(AUC: 0.876,95% CI: 0.805–0.929;AUC: 0.928,95% CI: 0.844–0.975;respectively).The radiomics signature performed better than the image feature model constructed by the intratumoral vessels(all p < 0.05)and showed a similar performance with the fusion model integrating radiomics signature and intratumoral vessels(all p>0.05)in terms of the discrimination in all cohorts.Moreover,the decision curve analysis verified the clinical utility of the radiomic signature in both external cohorts.Radiomics signature can effectively predict tumor necrosis,stage and grade of ccRCC.Studies have shown that the predictive power of radiomics nomogram and radiomics signature models is significantly higher than traditional radiological image feature model.The radiomics signature and radiomics nomogram constructed by the study can effectively predict tumor necrosis,stage and grade of ccRCC,thereby improving the quality of preoperative clinical decision-making for ccRCC cases and developing personalized for clinicians.The treatment plan provides supplementary information. |
PDF Full Text Request