Recent technological advancements have enabled detailed studies of associations between molecular signatures of cancer and tumor heterogeneity through multi-platform data integration of both genomic and radiomic types. We will present a method to integrate and harness imaging and genomic data in patients with lower grade gliomas (LGG), a type of brain cancer, in order to develop a formal regression framework for modelling association between them. Imaging data is represented through voxel intensity probability density functions of tumor sub-regions obtained from multimodal magnetic resonance imaging (MRI), and genomic data through molecular signatures in the form of enrichment scores corresponding to their gene expression profiles. Employing a Riemannian-geometric framework we construct density-based predictors to include in a Bayesian regression model with pathway enrichment score as the response. Variable selection compatible with the grouping structure amongst the predictors, induced through the tumor sub-regions, is carried out under a group spike-and-slab prior. Our analyses reveal several pathways, relevant to LGG etiology, to have significant associations with imaging data.