Prediction of Lymph Node Metastasis in Colorectal Cancer Based on Clinical Data, Body Composition, and Radiomics

Objective To develop and validate a positron emission tomography/computed tomography (PET/CT)–based model integrating clinical variables, body composition indices, and radiomics features for predicting lymph node metastasis (LNM) in colorectal cancer (CRC). Materials and Methods This retrospective study included 233 CRC patients who underwent preoperative ¹⁸ F-FDG PET/CT. Skeletal muscle and adipose tissue parameters were quantified from low-dose CT. Radiomics features were extracted from primary tumors on PET and CT images. Multilayer perceptron models based on clinical, body composition, and radiomics features were constructed in a training cohort and tested in an independent cohort. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis. Results LNM was present in 100/233 (42.9%) patients. Multivariable logistic regression identified carcinoembryonic antigen (CEA; odds ratio [OR] = 1.055, 95% confidence interval [CI]: 1.010–1.102, P  = 0.016), non-ulcerative macroscopic tumor type (OR = 2.142, 95% CI: 1.048–4.379, P  = 0.037), vascular invasion (OR = 3.015, 95% CI: 1.517–5.993, P  = 0.002), and a lower SMA/SFA ratio (OR = 0.367, 95% CI: 0.138–0.974, P  = 0.044) as independent predictors of LNM. The integrated Body_Clinical_Radiomics model achieved an AUC of 0.836 (95% CI: 0.775–0.891) in the training set and 0.811 (95% CI: 0.665–0.931) in the test set, outperforming the single-domain models. Conclusion A PET/CT-based model that integrates radiomics with clinical and body composition information enables more accurate preoperative prediction of LNM in CRC than models using any single feature domain.