P12.33.B ANATOMICAL AND PROGNOSTIC PATTERNS OF BIOMARKER-DEFINED GLIOMAS: A MULTI-COHORT SPATIAL ANALYSIS

Abstract BACKGROUND The spatial distribution of gliomas is closely associated with tumor behavior, treatment response, and outcomes. Gliomas exhibit heterogeneity in both molecular features and anatomical localization, shaped by distinct developmental origins. The 2021 WHO classification incorporates biomarkers such as IDH mutation, 1p/19q co-deletion, MGMT promoter methylation, and CDKN2A/B deletion to define clinically relevant subtypes. While prior studies have reported spatial preferences for certain biomarkers, most lacked systematic, multi-biomarker comparisons or validation. This study analyzed glioma spatial patterns and outcomes across molecular subtypes using public datasets under the 2021 WHO framework, and assessed reproducibility in an internal cohort. MATERIAL AND METHODS We analyzed preoperative MRI scans from over 1,800 glioma patients (~1,400 public, ~400 internal). Each case included four standard MRI sequences (T1, T1-contrast, T2, FLAIR), processed via a unified pipeline. Tumor masks were generated by automated segmentation and manually refined. All data were non-linearly registered to MNI-152 space. Localization was assessed using two approaches: region-based analysis with anatomical atlases (84-region parcellation, AAL template), and voxel-wise frequency mapping. Voxel-level masks were aggregated into regional features. Gliomas were classified per 2021 WHO criteria, and spatial distributions were compared across subtypes. Associations with survival were evaluated. Internal cases served as a validation cohort. RESULTS Distinct spatial patterns emerged across glioma subtypes. IDH-mutant gliomas were predominantly localized in the frontal lobe, while IDH wild-type glioblastomas showed broader distributions, with peaks in the temporal and parietal lobes. Oligodendrogliomas with 1p/19q co-deletion clustered in the frontal pole and superior frontal gyrus, whereas CDKN2A/B homozygous deletions were enriched in the temporal lobe and insula. MGMT-unmethylated tumors favored the temporal lobe and were associated with poorer survival, while methylated tumors showed more diffuse localization and better outcomes. We observed that the temporal pole and insular cortex may be enriched for more aggressive molecular subtypes and poorer outcomes. Several patterns were replicated in the internal validation cohort. CONCLUSION Glioma location correlates with molecular subtype and outcome. Spatial analysis revealed prognostically relevant, subtype-specific patterns that may support personalized surgical and therapeutic strategies. Limited biomarker data in some cohorts, especially for CDKN2A/B, remains a constraint. Future work will integrate spatial and transcriptomic data to explore location-dependent glioma mechanisms.