NIMG-74. RADIOIMMUNOMIC SIGNATURES IN PEDIATRIC LOW-GRADE GLIOMA BASED ON MULTIPARAMETRIC MRI SCANS

Abstract Understanding the immune microenvironment in pediatric low-grade glioma (pLGG) patients may help identification of who benefit from anti-tumor immunotherapies. However, surgical resection is not feasible for many pLGG tumors certain anatomical locations. Therefore, developing non-invasive tools that characterize tumor prior to therapeutic interventions could contribute stratification and enrollment into relevant clinical trials. In this work, we derived radiomic signatures profiles (radioimmunomics) based on machine learning (ML) analysis readily available conventional MRI scans. Transcriptomic data a cohort 197 subjects was retrospectively collected Open Pediatric Brain Tumor Atlas (OpenPBTA). The were categorized three groups (Group1-3) their immunological using consensus clustering algorithm. This revealed greater cell infiltration non-BRAF mutated pLGGs. Group1 showed more enrichment M1 macrophages, scores compared Group2 Group3. Elevated inflammation score (TIS), as predictor response anti-PD-1 blockade, observed (p= 1.4e-7) Group3 0.0054). Radiomic features, including volumetric, morphologic, histogram, texture descriptors, extracted segmented regions multiparametric (mpMRI) scans 71 (of 197) patients. Multivariate ML models trained predict features cross-validated random forest classifier along with recursive feature elimination, which yielded AUC 0.72 multi-class classification problem. Our findings indicate presence distinct tumors, possibly favorable immunotherapies tumors. Furthermore, developed radioimmunomic pre-operative mpMRI can potentially stratify microenvironment. Based these initial promising results, are exploring additional increase accuracy radioimmunomics model.