Background: High-risk neuroblastoma is associated with a poor prognosis, making it crucial to identify patients within this group who face an even higher risk of adverse outcomes.
Objective: To determine if integrating clinical indicators and venous-phase enhanced computed tomography radiomics features could improve the prediction of overall survival in high-risk neuroblastoma.
Materials and methods: We retrospectively included high-risk neuroblastoma patients treated at a primary institution, randomly stratifying them into a training set (70%) and a test set (30%). Univariate and multivariate Cox regression analyses were used to identify independent clinical risk factors. We then extracted radiomics features from venous-phase enhanced computed tomography images. Clinical risk factors, radiomics score, and combined model were evaluated in the training, test, and external validation sets.
Results: The training, test, and validation sets included 70, 30, and 40 patients, respectively. Neuron-specific enolase was identified as the independent clinical risk factor, with concordance indices of 0.616, 0.627, and 0.595 in the training, test, and validation sets, respectively. The radiomics score achieved concordance indices of 0.699, 0.690, and 0.684 in the training, test, and validation sets, respectively. The combined model showed concordance indices of 0.730, 0.707, and 0.690 in the training, test, and validation sets, respectively. The combined model predicted 5-year overall survival with an area under the receiver operating characteristic curve of 0.780 in the training set, 0.742 in the test set, and 0.710 in the validation set.
Conclusion: Combining neuron-specific enolase and venous-phase enhanced computed tomography radiomics improves survival prediction in high-risk neuroblastoma.
Keywords: Child; Neuroblastoma; Prognosis; Radiomics; Survival analysis; Tomography; X-ray computed.
© 2025. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.