Purpose <p>Targeted therapy with lenvatinib is a preferred option for advanced hepatocellular carcinoma, however, predicting its efficacy remains challenging. This study aimed to build a nomogram integrating clinicoradiological indicators and radiomics features to predict the response to lenvatinib in patients with hepatocellular carcinoma.</p> Methods <p>This study included 211 patients with hepatocellular carcinoma from two centers, who were allocated into the training (107 patients), internal test (46 patients) and external test set(58 patients). Radiomics features were extracted using a Pyradiomics-based system. Biopsy specimens were used for immunohistochemical staining of epidermal growth factor receptor. Risk factors for predicting treatment response were screened out to construct models by logistic regression analysis. The performance of models was evaluated with receiver operating characteristic and decision curve analyses.</p> Results <p>A total of 9370 radiomics features were extracted from five sequences. Subsequently, seven radiomics features were identified for modeling. Intratumoral vessel and expression level of epidermal growth factor receptor were independent risk factors for predicting treatment response to lenvatinib and were used to build a clinical model. No difference was found in predicting performance between clinical model and radiomics model. The combined model, integrating intratumoral vessel, epidermal growth factor receptor and radiomics features, had better predicting performance with areas under the receiver operating characteristic curve of 0.908, 0.877, and 0.870 for the training, internal test and external test sets, respectively.</p> Conclusion <p>This study underscored the significant potential of radiomics features combined with clinicoradiological indicators in the prediction of treatment response to lenvatinib in patients with hepatocellular carcinoma.</p>