Intratumoral and peritumoral radiomics for preoperative prediction of pathological complete response to neoadjuvant immunochemotherapy in patients with esophageal squamous cell carcinoma

Background Pathological complete response (pCR) after neoadjuvant immunochemotherapy (nICT) confers a significant survival advantage in locally advanced esophageal squamous cell carcinoma (ESCC), yet remains difficult to anticipate preoperatively. This study aimed to develop and validate a noninvasive, multimodal model integrating intratumoral and peritumoral radiomic features with clinical variables to predict pCR after nICT. Materials and Methods We retrospectively analyzed 122 consecutive patients with ESCC who underwent contrast-enhanced computed tomography (CT) before nICT followed by surgical resection. Radiomic features were extracted from manually delineated intratumoral and peritumoral regions, and corresponding radiomic scores were constructed. Independent clinical predictors were combined into a clinical score. These three scores were integrated using multivariable logistic regression to develop a multimodal nomogram. Model performance was evaluated using receiver operating characteristic analysis, calibration curves, and decision curve analysis. Feature contributions were assessed using SHapley Additive exPlanations (SHAP). Results Clinical T stage and tumor location constituted the clinical score. The peritumoral radiomic score showed numerically higher AUCs than the intratumoral score in the training (0.818 vs. 0.774) and validation sets (0.808 vs. 0.656). The multimodal nomogram showed good discrimination, with AUCs of 0.901 and 0.862 in the training and validation sets, respectively. SHAP analysis indicated that the peritumoral radiomic score had the largest relative contribution among model components. Conclusions A CT-based multimodal nomogram incorporating intratumoral and peritumoral radiomic features with clinical variables showed good performance for preoperative prediction of pCR after nICT in ESCC. Peritumoral radiomic features may provide complementary imaging information, although their biological interpretation remains exploratory.