Pilot study: T2 MRI radiomics with XGBoost shows promise for noninvasive risk stratification of pediatric neuroblastoma

In a pilot study of 30 children with neuroblastoma, T2-weighted MRI radiomics with XGBoost achieved an AUC of 0.88 for distinguishing high-risk from low/intermediate-risk tumors, suggesting potential as a noninvasive imaging biomarker.

• Features from T2-weighted and diffusion-weighted MRI contributed most to accurate classification.
• The chi-square feature selection method was most frequently used in high-performing model configurations (30.8%).
• The limited sample size (n=30) prevented detection of significant differences between classifiers after correction for multiple comparisons.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.