Prediction of High-Risk Neuroblastoma Among Neuroblastic Tumors Using Radiomics Features Derived from Magnetic...

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Neuroblastoma is a common pediatric cancer arising from neuroblastic cells.
• Accurate risk stratification is crucial for optimal neuroblastoma treatment and patient outcomes.
• Distinguishing high-risk neuroblastoma pre-treatment can guide therapeutic decisions.

Purpose of the Study:

• To develop and validate a predictive model for high-risk neuroblastoma using radiomics features from Magnetic Resonance Imaging (MRI).
• To assess the performance of multivariate logistic regression (MLR) and random forest (RF) models in predicting neuroblastoma risk.
• To evaluate the generalizability of the radiomics model on an external test dataset.

Main Methods:

• Retrospective analysis of pre-treatment MRI scans from pediatric patients with neuroblastic tumors (n=46).
• Manual segmentation of tumors to extract 930 radiomics features (first-order and texture).
• Development of MLR and RF models using 10-fold cross-validation, tested on an independent cohort.

Main Results:

• Radiomics features demonstrated moderate to excellent intra- and inter-observer variability (ICCs 0.633-0.985).
• The MLR model achieved an Area Under the Curve (AUC) of 0.94 in the training set and 0.94 in the test set.
• The RF model showed an average AUC of 0.83 in cross-validation and 0.91 in the test set, with high specificity.

Conclusions:

• MRI-based radiomics analysis provides a non-invasive method for predicting high-risk neuroblastoma.
• The developed MLR and RF models show high accuracy and generalizability in identifying aggressive neuroblastic tumors.
• This radiomics approach can potentially improve risk stratification and personalize treatment for pediatric neuroblastoma.