CT-based Radiomics Model To Predict The Disease-free Survival For Patients With Esophageal Squamous Cell Carcinoma After Esophagectomy

Purpose:Most patients may suffer progression after esophagectomy for esophageal carcinoma(EC).Thus,to predict high risk of progression among patients who underwent esophagectomy and make individualized treatment administration are crucial in clinical practice,which could improve patient’s prognosis.Our study aims to evaluate the role of computed tomography(CT)-based radiomics in prediction of the disease-free survival(DFS)for patients with esophageal squamous cell carcinoma(ESCC)after esophagectomy.Materials and methods: Retrospective study was performed on patients diagnosed with ESCC confirmed by postoperative pathology between February 2013 and August2019.In total of 184 patients with ESCC were included,training set(n=184)and internal test set(n=56).Pre-treatment CT images were loaded into ITK-SNAP software to outline the regions of interest.And then radiomic features were extracted in Python 3.6 based on the pyradiomics.The least absolute shrinkage and selection operator(LASSO)–cox analysis was used to select the most powerful prognostic features and built the radiomic model.Univariate and multivariate cox regression analysis methods were used to select the clinical factors that were significantly related to the DFS,and then the clinical model was established.Finally,the combined model,radiomic-clinical model was established.The concordance index(C-index)was used to evaluate the models’ discriminative performance.The calibration curves were used to validate the consistency between the observed and predicted values.Kaplan–Meier analysis and the log-rank test were applied to assess the models’ risk stratification ability of progression.Results: In total,ten key radiomic features contributed to the radiomic model.Pathologic lymph node stage(p N)was selected as significant clinical factor to build the clinical model.The radiomic-clinical model showed good discriminatory ability in the training and internal test sets with C-index 0.744,95% CI(0.689,0.799)and0.774,95% CI(0.676,0.872)respectively,which performed well in stratifying patients in high and low risk of progression after esophagectomy.The median DFS of high-risk group was obviously shorter than that of low-risk group,16 months vs 39 months in the training set(log-rank test,P =0.006),11 months vs 24 months in the test set(log-rank test,P <0.001).Conclusion: Our study has developed and validated that the radiomic-clinical model based on CT radiomics in prediction of the DFS for patients with ESCC after esophagectomy,which also has achieved good ability in stratifying patients in high risk of progression.It could help guide individualized treatment strategy.