CT-based multi-station lymph-node radiomics for preoperative prediction of lymph node metastasis in colorectal cancer: a proof-of-concept comparison with primary-tumor radiomics

Abstract Objective Accurate preoperative prediction of lymph node metastasis (LNM) in colorectal cancer remains challenging using conventional CT assessment. This proof-of-concept study compared CT-based multi-station lymph-node radiomics with primary-tumor radiomics for patient-level preoperative prediction of LNM. Materials and methods This single-center retrospective study included patients with pathologically confirmed colorectal cancer who underwent preoperative contrast-enhanced CT and surgical resection. Patients were divided into training and validation cohorts. Radiomics features were extracted from the primary tumor and visible regional lymph nodes across multiple nodal stations. Three radiomics strategies were developed and compared: primary-tumor radiomics, multi-station lymph-node radiomics, and combined radiomics. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration analysis, decision curve analysis, and five-fold cross-validation. Pairwise AUCs were compared using the DeLong test. A clinical-radiomic model was further constructed by integrating the optimal radiomics strategy with preoperative clinical variables. Results In the validation cohort, the multi-station lymph-node radiomics model outperformed the primary-tumor radiomics model for predicting LNM (AUC, 0.864 versus 0.637; p = 0.0055). The combined radiomics model achieved an AUC of 0.840, without significant improvement over the lymph-node model ( p = 0.2211). After integration with clinical variables, the clinical-radiomic model achieved an AUC of 0.879. Conclusion CT-based multi-station lymph-node radiomics showed significantly higher validation performance than primary-tumor radiomics for patient-level preoperative LNM prediction in colorectal cancer. However, this single-center proof-of-concept model is not ready for direct clinical implementation and requires multicenter external validation before clinical translation. Key Points Question Can CT-based multi-station lymph-node radiomics improve preoperative nodal metastasis prediction beyond primary-tumor radiomics in patients with surgically treated colorectal cancer? Findings Multi-station lymph-node radiomics achieved higher validation performance than primary-tumor radiomics for patient-level prediction of pathological lymph-node metastasis in colorectal cancer. Relevance statement CT-based multi-station lymph-node radiomics may support individualized preoperative nodal risk stratification in colorectal cancer, but multicenter validation and prospective evaluation are required before clinical translation.