Evaluation of clinical and imaging features for differentiating rhabdomyosarcoma from neuroblastoma in pediatric soft...

In this study, we developed a nomogram predictive model based on clinical, CT, and MRI parameters to differentiate soft tissue rhabdomyosarcoma (RMS) from neuroblastoma (NB) in children preoperatively.

A total of 103 children with RMS (n=37) and NB (n=66) were enrolled in the study from December 2012 to July 2023. The clinical and imaging data (assessed by two experienced radiologists) were analyzed using univariate analysis, and significant factors were further analyzed by multivariable logistic regression using the forward LR method to develop the clinical model, radiological model, and integrated nomogram model, respectively. The diagnostic performances, goodness of fit, and clinical utility of the integrated nomogram model were assessed using the area under the curve (AUC) of the receiver operator characteristics curve (ROC) with a 95% confidence interval (95% CI), calibration curve, and decision curve analysis (DCA) curves, respectively. Diagnostic efficacy between the model and radiologists' interpretations was examined.

The median age at diagnosis in the RMS group was significantly older than the NB group (36.0 months vs. 14.5 months; P=0.003); the fever rates in RMS patients were significantly lower than in patients with NB (0.0% vs.16.7%; P=0.022), and the incidence of palpable mass was higher in patients with RMS compared with the NB patients (89.2% vs. 34.8%; P<0.001). Compare NB on image features: RMS occurred more frequently in the head and neck and displayed homogeneous density on non-enhanced CT than NB (48.6% vs. 9.1%; 35.3% vs. 13.8%, respectively; all P<0.05), and the occurrence of characteristics such as calcification, encasing vessels, and intraspinal tumor extension was significantly less frequent in RMS children compared to children with NB (18.9% vs. 84.8%; 13.5% vs. 34.8%; 2.7% vs. 50.0%, respectively; all P <0.05). Two, three, and four features were identified as independent parameters by multivariate logistic regression analysis to develop the clinical, radiological, and integrated nomogram models, respectively. The AUC value (0.962), calibration curve, and DCA showed that the integrated nomogram model may provide better diagnostic performance, good agreement, and greater clinical net benefits than the clinical model, radiological model, and radiologists' subjective diagnosis.

The clinical and imaging features-based nomogram has potential for helping radiologists distinguish between pediatric soft tissue RMS and NB patients preoperatively, and reduce unnecessary interventions.