Predictive value of computed tomography radiomics for lymphatic-vascular space infiltration in colon cancer.

This study aimed to construct and validate a preoperative predictive model for lymphatic-vascular space infiltration (LVSI) in colon cancer using clinical features and computed tomography (CT) radiomics, and to evaluate its clinical utility. A total of 244 colon cancer patients treated at Yongkang First People's Hospital from January 2018 to January 2024 were enrolled as the training set (LVSI-positive: n=92, LVSI-negative: n=152), and 58 patients treated between February 2024 and August 2025 served as the validation set. Clinical data were collected, and contrast-enhanced CT images were analyzed. Radiomic features were extracted using PyRadiomics, and features with intraclass correlation coefficient (ICC) >0.8 were retained to ensure reproducibility, and least absolute shrinkage and selection operator (LASSO) regression was applied for dimensionality reduction. A clinical model, a radiomics model (based on Rad-score), and a combined model were established via multivariate logistic regression. Receiver operating characteristic (ROC) curve, calibration curve, Hosmer-Lemeshow test, and decision curve analysis (DCA) were used to assess model performance. The results showed that tumor diameter, differentiation degree, CT-detected extramural vascular invasion (cEMVI), and carcinoembryonic antigen (CEA) were independent risk factors for LVSI (all P<0.05). Four key radiomic features were screened to calculate Rad-score. In the training set, the combined model achieved an area under the curve (AUC) of 0.90 (95% CI: 0.86-0.94), significantly higher than the clinical model (AUC=0.75) and radiomics model (AUC=0.84) (both P<0.001), with accuracy, sensitivity, and specificity of 0.82, 0.80, and 0.86, respectively. In the validation set, the combined model maintained an AUC of 0.92 (95% CI: 0.86-0.99), outperforming the clinical model (AUC=0.71, P=0.004), and showed good calibration (Hosmer-Lemeshow P=0.364) and positive net benefits in DCA. The combined model integrating clinical features and CT radiomics exhibits excellent performance in preoperative prediction of LVSI in colon cancer, providing a reliable tool for individualized treatment decision-making.