Objective To construct and validate a nomogram model that integrates computed tomography (CT) radiomics and deep learning (DL) features to predict the risk of lymph node metastasis (LNM) in colorectal cancer (CRC) before surgery. Methods The clinical data of 116 CRC patients who were pathologically diagnosed at Zhengzhou Huazhuo Hospital from January 2022 to June 2025 were retrospectively included. After stratified random sampling based on LNM status and T stage, the data were divided into a training set (n＝81) and a validation set (n＝35) according to 7 vs. 3. Based on the CT images of the venous phase, manually extracted imaging features and DL features from the pre-trained ResNet50 were respectively obtained. After feature screening, the clinical model, single feature model, fusion model and joint nomogram model were constructed and compared. The discrimination of the model was evaluated by the area under the curve (AUC), and the calibration efficacy of the model was assessed using the Hosmer-Lemeshow test and the Brier score. Results The AUC of the joint nomogram model in the validation set was 0.823 (95%CI: 0.694-0.952), with a sensitivity of 85.7% and a negative predictive value of 89.5%, which was significantly better than that of the clinical model (AUC＝0.701), and its prediction probability was also in good agreement with the actual observations (Hosmer-Lemeshow test P＝0.719, Brier score＝0.148). Conclusion The joint prediction model constructed in this study can effectively assess the risk of LNM in CRC, and can provide a quantifiable tool for individualized preoperative decision-making.