Pretreatment CT Radiomics To Predict Treatment Response And Prognosis In Esophageal Squamous Cell Carcinoma Treated With Chemoradiotherapy

ObjectivesTo investigate the feasibility of using pretreatment CT radiomics to predict treatment response and local progression Free Survival(LPFS)and to develop a nomogram for prediction of CR status and LPFS in esophageal squamous cell carcinoma patients treated with chemoradiotherapy,so as to assist radioation oncologist to develop individual treatment regimen for the patients.MethodsData of patients diagnosed as esophageal squamous cell carcinoma(ESCC)and treated with definitive chemoradiotherapy(dCRT)in our hospital were retrospectively collected.Eligible patients were included in this study after successive screening,and their pretreatment CT files were selected for analysis.In the first chapter,226 patients were included and randomly divided into two groups,160 patients in training set and 66 patients in validation set.VOI(Volume of interest)of tumor was delineated in 3D slicer software.Python package was used to extracted the radiomics features and LASSO regression was used to reduce the features dimension and screen out important features for developing radiomics signature.A nomogam model incorporating Rad-score was developed for predicting CR status based on the result of multivariate analysis and validated in validation set.The performance of the model was determined and its clinical value was evaluated by decision curve.In the second chapter,221 patietns were included and randomly divided into two groups,155 patients in training set and 66 patients in validation set.3D slicer software was used to contour the tumor VOI and python package was used to extract the radiomics features.LASSO regression was performed to select radiomics features for calculating Rad-score.A nomogram model was developed for predicting LPFS of ESCC patients treated with dCRT based on the result of Cox multivariate analysis.Calibration curves and Net reclassification index(NRI)were used to evaluate the performance of the nomogram model.ResultsIn the first chapter,seven radiomics features were screened out to develop a radiomics signature Rad-score,which could predict the CR status of ESCC patients treated with dCRT.Univariate analysis showed that Rad-score was significantly correlated with CR status.The AUC(Area Under Curve)was 0.812(95%CI:0.742-0.869)in the training set and 0.744(95%CI:0.632-0.851)in the validation set.Multivariate analysis showed that Rad-score and clinical stage were independent predictors of CR,with P values of 0.035 and 0.023,respectively.A nomogram model incorporating Rad-socre and clinical stage was developed and validated,with an AUC of 0.844(95%CI:0.779-0.897)in the training set and 0.807(95%CI:0.691-0.894)in the validation set.The sensitivity and specificity were 67.86%and 86.54%in the training set and 77.21%and 77.27%in the validation set.The classification accuracy was 64.77%.Delong test showed that nomogram model was significantly superior to the clinical stage,with P<0.001 in the training set and P=0.026 in the validation set.The decision curve showed that the nomogram model was superior to the clinical stage when the risk threshold was greater than 25%.In the second chapter,17 radiomics features were selected to develop the radiomics signature Rad-score,which divided patients into two groups with high risk of local recurrence and low risk of local recurrence.LPFS survival curves were significantly separated,with P<0.001 in the training set and P=0.026 in the validation set.Multivariate analysis showed that Rad-score,N stage and CR status were independent predictors of LPFS in the patients receiving dCRT for ESCC,with p values all less than 0.05.The nomogram model met the PH test(P=0.21),and the c-index was 0.745(95%CI:0.700-0.790)in training set and 0.723(95%CI:0.654-0.791)validation set,respectively.The calibration curve showed that the LPFS predicted by the model was in good consistence with the actual LPFS.NRI analysis showed that the nomogram model was superior to clinical stage.ConclusionsThe nomogram model based on pretreatment CT radiomics to predict CR status and LPFS of ESCC patients treated with dCRT has good prediction ability,which is better than clinical stage.It can provide individualized prognosis estimation for clinicians and assist clinicians to determine individual treatment regimien for patients receiving dCRT for ESCC.