Computed tomography based radiomics signature for predicting the expression of vascular endothelial growth factor in pediatric patients with nephroblastoma

Abstract

To construct a computed tomography (CT) based radiomics signature and assess its performance in predicting vascular endothelial growth factor (VEGF) expression in pediatric patients with nephroblastoma. A total of 73 pediatric nephroblastomaL patients were enrolled (51 in the training cohort and 22 in the test cohort). The region of interest manually marked on the CT images served as the basis for the automatic extraction of radiomics features. A radiomics score was generated utilizing the radiomics signature based formula after retaining a subset of radiomics features to create a radiomics signature. Clinical elements, such as clinicopathological information and CT imaging characteristics, were used to create a clinical model. With the inclusion of a radiomics signature and clinical characteristics, a composite nomogram was created. Decision curve analysis (DCA) was used to evaluate the prediction performance. 5 carefully chosen radiomics features were used to create the radiomics signature. Next, the radiomics score was determined. In the training cohort and the test cohort, the logistic regression model’s area under the curve was 0.761 and 0.791, respectively. Based on the radiomics signature and clinical variables, the clinical radiomics nomogram demonstrated its ability to accurately predict the level of VEGF expression. DCA verified the clinical value of the clinical radiomics nomogram. In pediatric patients with nephroblastoma, the radiomics model based on the CT radiomics signature may accurately predict the level of VEGF expression.