This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Neuroblastoma is one of the most common pediatric cancers. This study used machine learning (ML) to predict the mortality and a few other investigated intermediate outcomes of neuroblastoma patients non-invasively from CT images. Performances of multiple ML algorithms over retrospective CT images of 65 neuroblastoma patients are analyzed. An artificial neural network (ANN) is used on tumor radiomic features extracted from 3D CT images. A pre-trained 2D convolutional neural network (CNN) is used on slices of the same images. ML models are trained for various pathologically investigated outcomes of these patients. A subspecialty-trained pediatric radiologist independently reviewed the manually segmented primary tumors. Pyradiomics library is used to extract 105 radiomic features. Six ML algorithms are compared to predict the following outcomes: mortality, presence or absence of metastases, neuroblastoma differentiation, mitosis-karyorrhexis index (MKI), presence or absence of MYCN gene amplification, and presence of image-defined risk factors (IDRF). The prediction ranges over multiple experiments are measured using the area under the receiver operating characteristic (ROC-AUC) for comparison. Our results show that the radiomics-based ANN method slightly outperforms the other algorithms in predicting all outcomes except classification of the grade of neuroblastic differentiation, for which the elastic regression model performed the best. Contributions of the article are twofold: (1) noninvasive models for the prognosis from CT images of neuroblastoma, and (2) comparison of relevant ML models on this medical imaging problem.

Purpose Vesicular monoamine transporters 1 and 2 (VMAT1 and VMAT2) are thought to mediate MIBG uptake in adult neuroendocrine tumors. In neuroblastoma, the norepinephrine transporter (NET) has been investigated as the principal MIBG uptake protein, though some tumors without NET expression concentrate MIBG. We investigated VMAT expression in neuroblastoma and correlated expression with MIBG uptake and clinical features. Methods We evaluated VMAT1 and VMAT2 expression by immunohistochemistry (IHC) in neuroblastoma tumors from 76 patients with high-risk metastatic disease treated on a uniform cooperative group trial (COG A3973). All patients had baseline MIBG diagnostic scans centrally reviewed. IHC results were scored as the product of intensity grading (0-3+) and percent of tumor cells expressing the protein of interest. Association of VMAT1 and VMAT2 scores with clinical and biological features was tested using Wilcoxon rank-sum tests. Results Patient characteristics were typical of high-risk neuroblastoma, though the cohort was intentionally enriched for patients with MIBG non-avid tumors (n=20). VMAT1 and VMAT2 were expressed in 62% and 75% of neuroblastoma tumors, respectively. VMAT1 and VMAT2 scores were both significantly lower in MYCN amplified tumors and in tumors with high mitotic karyorrhectic index. MIBG avid tumors had significantly higher VMAT2 scores compared to MIBG non-avid tumors (median 216 vs. 45; p = 0.04). VMAT1 expression did not correlate with MIBG avidity. Conclusions VMAT1 and VMAT2 are expressed in the majority of neuroblastomas. Expression correlates with other biological features. Expression level of VMAT2 but not VMAT1 correlates with avidity for MIBG.

No abstract