Imaging patterns predict patient survival and molecular subtype in glioblastoma via machine learning techniques