Abstract
The biologic processes that contribute to contrast enhancement on magnetic resonance imaging of glioblastoma patients remain poorly understood. Glioblastoma tumors from The Cancer Imaging Archive were segmented using iterative probabilistic voxel labeling. Three parameters of contrast enhancement (CE) were calculated: intensity (CE(i)), heterogeneity (CE(h)), and volumetric ratio to necrosis (CE(r)). Associations between these parameters and gene expression from The Cancer Genome Atlas were examined to gain insights into their underlying biologic process. Glioma CpG island methylator phenotype glioblastomas (G-CIMP) were poorly enhancing. No differences in CE parameters were found between proneural, neural, mesenchymal, and classical glioblastomas. Increased CE(i) was associated with expression of genes that mediate inflammatory immune responses. High CE(h) was associated with increased genes required for tumor migration and invasion, including those that modulate extracellular matrix (ECM) and endothelial vascularity. High CE(r) was associated increased gene expression associated with stressful metabolic states, including hypoxia and starvation. Our results indicate aspects of CE are associated with distinct underlying biology. Integrative analysis of these CE parameters may yield meaningful information pertaining to the biologic state of glioblastomas and guide future therapeutic paradigms.