Prediction of pathological complete response to neoadjuvant chemoradiotherapy in esophageal squamous cell carcinoma patients with CT-based delta-radiomics nomogram

Abstract

Abstract Background The aim of this study was to develop a nomogram model that uses CT-based delta-radiomics features and clinical factors to predict pathological complete response (pCR) in patients with resectable locally advanced esophageal squamous cell carcinoma (ESCC) who received neoadjuvant chemoradiotherapy (nCRT). Methods The study included 232 ESCC patients who underwent computed tomography (CT) scans before and after nCRT between June 2018 and December 2021. The patients were randomly divided into training and validation sets with 174 and 58 patients, respectively. 837 radiomics features were extracted from delineations of the region of interest on pre- and post-treatment CT images, and calculated their deltas. The least absolute shrinkage and selection operator (LASSO) algorithm was used to select delta-radiomics features (DRF) based on classification performance. Logistic regression was used to construct a nomogram with clinical factors, and the performance of the nomogram in predicting pCR was evaluated using the area under the receiver operating characteristics (ROC) curve (AUC) analysis. Results There was no significant difference between the training and validation datasets. The delta-radiomics signatures (DRS), consisting of four features, demonstrated good predictive performance for pCR, with α-binormal-based and empirical AUCs of 0.831 and 0.843. T-stage and differentiation degree were identified as independent predictive factors of pCR in ESCC patients with nCRT. The nomogram combined the DRS and clinical factors improved the classification performance in the training dataset (AUCαbin = 0.963 and AUCemp = 0.964). Furthermore, the validation set showed a similar performance to that of the training set, with AUCs of 0.967 and 0.964. Conclusions A nomogram model based on CT-based delta-radiomics features and clinical factors provided high discriminatory accuracy in predicting pCR status of ESCC patients after nCRT.