Objective  To develop and validate the ability of a model based on intratumoral (TR) and peritumoral (PTR) CT radiomics combined with clinical and hematological inflammatory markers to predict lymph node metastasis (LNM) in esophageal squamous cell carcinoma (ESCC).  

Methods  In this retrospective study, patients with pathologically proven ESCC were enrolled and randomly divided into training and test sets at a 8:2 ratio. Clinical data, hematological inflammatory markers , TR and PTR CT radiomics features were used to construct five models: clinical model, hematological model, TR and PTR radiomics models, Clinical radiomics model, and a hybrid model which combined the clinical, hematological, TR and PTR 1mm radiomics features. The models were evaluated using the area under the receiver operating characteristic curve (AUC), calibration curve and decision curve analysis (DCA).

Results  A total of 304 patients were included in the study (mean age 67.40±6.71 years,182 males). The cN stage was signifcantly correlated with LNM in all cohorts (p<0.05). The hybrid model exhibited the best performance, with AUC values of 0.898 (95% CI: 0.859-0.937) and 0.867 (95% CI: 0.766-0.968) for predicting LNM in the training and test sets, respectively. Compared to the the clinical model ,TR radiomics model，TR+PTR radiomics model and Clinical radiomics model, the hybrid model showed the highest net profit rate in DCA.

Conclusion  The hybrid model based on TR and PTR 1mm CT radiomics combined with clinical and hematological inflammatory markers can effectively LNM in ESCC, offering a promising tool for individualized clinical decision-making.