IMG-06. RADIOMIC-BASED RISK STRATIFICATION OF PEDIATRIC LOW-GRADE GLIOMA REVEALS DIFFERENTIALLY EXPRESSED MOLECULAR PROCESSES CONTRIBUTING TO VARIABLE PATIENT PROGNOSIS

• Published in: Neuro-oncology (Charlottesville, Va.) Vol. 25; no. Supplement_1; pp. i47 - i48
• Main Authors: Kazerooni, Anahita Fathi, Kim, Meen Chul, Kraya, Adam, Mahtabfar, Aria, Haldar, Debanjan, Rathi, Komal, Khalili, Nastaran, Familiar, Ariana, Ware, Jeffrey B, Bagheri, Sina, Anderson, Hannah, Storm, Phillip B, Resnick, Adam C, Vossough, Arastoo, Davatzikos, Christos, Nabavizadeh, Ali
• Format: Journal Article
• Language: English
• Published: US Oxford University Press 12.06.2023
• Subjects: Final Category: Imaging - IMG
• ISSN: 1522-8517; 1523-5866
• DOI: 10.1093/neuonc/noad073.183

Abstract In pediatric low-grade gliomas (pLGG), prognosis and responses to treatments are heterogeneous. This heterogeneity may be explained by the differences in molecular composition of the tumors of the same histology and the likely upstream alterations following administered radiation or systemic therapy. Integration of radiomics and clinical variables could generate a non-invasive biomarker that provides upfront prediction about patient’s risk of progression. We show that our proposed radiomic-based risk-stratification signature for pLGGs is associated with alterations in transcriptomic pathways. Standard multiparametric MRI sequences of 134 pLGG patients from Children’s Brain Tumor Network (CBTN) were retrospectively collected and 881 quantitative radiomic features were extracted. A multivariate Cox proportional hazard’s (Cox-PH) regression model was fitted on clinical (age, sex, tumor location, and extent of tumor resection) along with radiomic variables using 5-fold cross-validation, to predict patient’s risk of progression. The Cox-PH model showed excellent performance in prediction of PFS and patient’s risk scores, supported by the concordance index of 0.78. Radiogenomic analysis was performed to determine the transcriptomic pathways (1594 pathways, c2 MsigDb Reactome v2022.1) that contribute to the pLGG risk, predicted by the radiomic signature. ElasticNet regression was applied on the scores obtained by gene set enrichment analysis (GSEA) (in 70/134 subjects) to predict radiomic-based risk scores. Increased risk, corresponded to upregulation of DNA repair pathways, dysregulation of lipophagy, fatty acid beta oxidation, and vitamin D pathways, which are tumor-promoting. BRAFfusion signaling inversely correlated with risk, consistent with known favorable prognosis of KIAA1549-BRAF fused pLGGs. Upregulation in immune related pathways and Toll-Like Receptor (TLR) signaling was associated with lower risk. This study elucidates the synergistic dynamics between the biological processes that promote the risk of progression and radiomic-based risk-stratification signature. The proposed biomarker may be used to encourage targeted therapies in patients with increased predicted risk.