Development and validation of a CT radiomics nomogram for preoperative prediction of right recurrent laryngeal nerve lymph node metastasis in thoracic esophageal squamous cell carcinoma.

Background: Preoperative prediction of right recurrent laryngeal nerve (RLN) lymph node metastasis (LNM) in esophageal squamous cell carcinoma (ESCC) is essential for balancing oncological clearance and nerve preservation. However, conventional computed tomography (CT) imaging based on size criteria exhibits poor sensitivity, frequently <50%, often failing to detect micrometastases or small metastatic nodes. Unlike general lymph node models, this study focuses specifically on the right RLN station to provide a targeted tool for surgical planning in this anatomically complex region. Methods: This retrospective study analyzed 413 consecutive patients who underwent radical esophagectomy. The gold standard for metastasis was established through site-specific histopathological examination of the right RLN lymph nodes, which were meticulously dissected and separately labeled during surgery to ensure direct correlation with preoperative CT imaging. Patients were divided into training (n=289) and test (n=124) sets. Radiomic features (n=945), comprising first-order statistics, shape, and texture features, were extracted from arterial-phase CT images of right RLN lymph nodes. Six machine learning algorithms [logistic regression (LR), K-nearest neighbors (KNN), decision tree (DT), support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost)] were trained. Clinical risk factors were identified via univariate/multivariate LR. A combined model fused radiomics scores and clinical factors. Performance was evaluated using ROC curves, Hosmer-Lemeshow calibration, and decision curve analysis (DCA). Results: Eleven radiomic features were selected for modeling. The LR-based radiomics model achieved an area under the curve (AUC) of 0.87 [95% confidence interval (CI): 0.83-0.92] in the training set and 0.82 (95% CI: 0.71-0.90) in the test set. Lymph node short axis (P<0.001) and tumor differentiation (P=0.01) were independent clinical predictors. The combined model demonstrated superior performance, with AUCs of 0.90 (95% CI: 0.85-0.94, training) and 0.86 (95% CI: 0.78-0.94, test). Calibration curves confirmed agreement between predicted and observed outcomes (Hosmer-Lemeshow P>0.05). DCA indicated clinical utility across threshold probabilities. Conclusions: An integrated model combining CT radiomics and clinical factors effectively predicts right RLN LNM in ESCC. This tool may guide surgical strategies and adjuvant therapy decisions, enhancing personalized treatment.