Diagnostic performance of multiparametric model integrating ¹⁸F-MFBG PET/CT imaging parameters and clinical factors for differentiating neuroblastoma from ganglioneuroblastoma: a preliminary study.
In a retrospective cohort of 42 pediatric patients, a model combining ¹⁸F-MFBG PET/CT SUVmax, age, and bone metastasis differentiated neuroblastoma from ganglioneuroblastoma with an AUC of 0.856, but lacked external validation.
Open original publication →What the AI sees
In a retrospective cohort of 42 pediatric patients, a model combining ¹⁸F-MFBG PET/CT SUVmax, age, and bone metastasis differentiated neuroblastoma from ganglioneuroblastoma with an AUC of 0.856, but lacked external validation.
Research significance
The study provides preliminary evidence that integrating ¹⁸F-MFBG uptake with clinical factors may improve noninvasive diagnostic classification; it is an untested inference that better classification could guide treatment selection, reduce diagnostic uncertainty, or improve outcomes.
Source abstract
PURPOSE: To develop and preliminarily validate a multiparametric diagnostic model integrating ¹⁸F-meta-fluorobenzylguanidine (¹⁸F-MFBG) PET/CT imaging parameters with clinical factors for noninvasive differentiation of neuroblastoma (NB) from ganglioneuroblastoma (GNB) in pediatric patients. METHODS: This retrospective study enrolled 42 pediatric patients with histologically confirmed NB (n = 23) or GNB (n = 19) who underwent ¹⁸F-MFBG PET/CT prior to treatment. Quantitative imaging parameters and clinical data were extracted. Correlations were assessed using Spearman's rank correlation coefficient. Multicollinearity among predictors was evaluated using variance inflation factors (VIF). Univariate and multivariable Firth logistic regression analyses were performed to identify independent predictors. The discriminative performance of the model was evaluated via receiver operating characteristic (ROC) analysis, the DeLong test, bootstrap validation, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). RESULTS: SUVmax, SUVmean, TBRmean, TBRmax, AUC-CSH and TL-hNET differed significantly between NB and GNB(P < 0.05). Severe multicollinearity (VIF > 10) was observed among imaging parameters, SUVmax was selected for multivariable modeling, along with age, bone metastasis and 11q23 deletion as the strongest clinical predictors. In the multivariable Firth model, SUVmax (OR = 8.665; P = 0.029) and age (OR = 0.676; P = 0.043) remained independently associated with NB. Bone metastasis showed a trend toward independent association with NB (OR = 11.753; P = 0.050), with the association approaching statistical significance. Three diagnostic models were constructed and compared: (1) imaging-only model (SUVmax); (2) clinical-factor model (age + bone metastasis); and (3) integrated model (SUVmax + age + bone metastasis). The combined model achieved an AUC of 0.856(95%CI:0.737-0.975), showing a trend toward improved discrimination compared with both the SUVmax-only model (AUC = 0.810, 95%CI:0.668-0.952) and the clinical-factor model (AUC = 0.787, 95%CI:0.659-0.916). Bootstrap validation (2000 resamples) and IDI analysis provided supportive evidence of incremental value of the integrated model (Combined vs. SUVmax-only: IDI = 0.122, P = 0.014; Combined vs. clinical: IDI = 0.083, P = 0.045). CONCLUSION: The multiparametric model integrating ¹⁸F-MFBG PET/CT imaging parameters with clinical factors generated a promising preliminary hypothesis for diagnostic performance for differentiating NB from GNB, pending rigorous external validation in larger, multicenter cohorts.