Purpose: Positive patient outcomes with the use of aggressive concurrent chemoradiotherapy have been shown to be effective for classical and mesenchymal glioblastoma multiforme (GBM) subtypes, while not showing the same positive outcomes for patients with neural and proneural subtypes. Further, the classical subtype of GBM has been shown to trend toward improved survival from the combination of Bevacizumab and CCNU, while the other GBM subtypes did not improve significantly with this treatment. Quantitative MR imaging features from multiparametric MRI have been recently explored for GBM subtype classification with some promising results. This thesis extends previous work by integrating 3D textures and moment invariant features to determine if the association between glioblastoma MR image features and genomic markers may be enhanced. Methods: 147 subjects with a complete set of T1, T1C (post-contrast T1), T2, and FLAIR imaging modalities without significant imaging artifacts acquired prior to surgery and with a known genomic subtype (classical: 30, mesenchymal:47, neural:25, and proneural:45) were chosen from multiparametric MRI data from the TCIA GBM database. The subject data were preprocessed and tumor and edema volumes were contoured. Features were extracted from the four modalities and difference fields of T1C-T1 and FLAIR-T2 from the segmented data. These 3D features include volume, surface area, surface roughness, moment invariants, intensity statistics, and texture features. Feature selection and classification was performed using random forest. Results: Random forest out-of-bag classification accuracy with one-vs-rest methodology in classifying the four genomic subtypes is 85.7%. Features varied depending on the classes being compared. The selected features show that tumor heterogeneity, intensity shape, as well the tumor and edema border information distinguish the molecular subtypes. Textures based on independent component analysis (ICA) tumor and edema derived 3D kernels are shown to be strong features for GBM subtype discrimination. This is the first study to use textures based on ICA tumor derived 3D kernels for molecular subtype classification. Conclusion: 3D texture features are useful for classification of GBM genomic subtypes, including ICA based textures introduced in this thesis. Imaging features may lead to correct classification of molecular subtypes to enable stratification for targeted treatment planning.