DCE-MRI radiomics nomogram can predict response to neoadjuvant chemotherapy in esophageal cancer

Abstract

Objectives: To assess volumetric DCE-MRI radiomics nomogram in predicting response to neoadjuvant chemotherapy (nCT) in EC patients. Methods: This retrospective analysis of a prospective study enrolled EC patients with stage cT1N + M0 or cT2-4aN0-3M0 who received DCE-MRI within 7 days before chemotherapy, followed by surgery. Response assessment was graded from 1 to 5 according to the tumor regression grade (TRG). Patients were stratified into responders (TRG1 + 2) and non-responders (TRG3 + 4 + 5). 72 radiomics features and vascular permeability parameters were extracted from DCE-MRI. The discriminating performance was assessed with ROC. Decision curve analysis (DCA) was used for comparing three different models. Results: This cohort included 82 patients, and 72 tumor radiomics features and vascular permeability parameters acquired from DCE-MRI. mRMR and LASSO were performed to choose the optimized subset of radiomics features, and 3 features were selected to create the radiomics signature that were significantly associated with response (P < 0.001). AUC of combining radiomics signature and DCE-MRI performance in the training (n = 41) and validation (n = 41) cohort was 0.84 (95% CI 0.57–1) and 0.86 (95% CI 0.74–0.97), respectively. This combined model showed the best discrimination between responders and non-responders, and showed the highest positive and positive predictive value in both training set and test set. Conclusions: The radiomics features are useful for nCT response prediction in EC patients.