Glioblastoma is the most aggressive primary brain tumor, and despite being relatively rare, it has a devastatingimpact on patients and their families. The tumor cells invade and migrate to otherwise healthy parts of thebrain, making them difficult to track and even harder to overcome. The treatment is aggressive and includessurgery, chemo- and radiotherapy. However, the tumor almost inevitably recurs - often close to where it wasoriginally located. This indicates that the current treatment regimen is inadequate. Anatomical imaging usingcomputed tomography and magnetic resonance imaging (MRI), play a central role in all phases of treatmentmanagement, including planning and subsequent evaluation of radiotherapy. However, the full infiltrative natureof glioblastomas is usually not depicted by conventional imaging.The aim of this PhD project was to evaluate the pattern of recurrence after introduction of positron emissiontomography (PET) for radiotherapy planning and to investigate the feasibility of advanced, multi-modalityimaging for glioblastoma management.The results showed that patterns of failure remained unchanged after the clinical introduction of PET forradiotherapy planning. Simulations indicated that treatment margins could potentially be reduced, which wouldlead to lower radiation doses to the brain, and possibly fewer side-effects. These results were found to be robustagainst local tissue deformations that are not usually taken into account. In addition, the alignment uncertaintyof MRI at the time of recurrence to pre-radiotherapy MRI varied from 1 mm to more than 5 mm depending onregistration method and distance to the tumor.A prospective diagnostic study revealed that combined PET and MRI imaging is feasible, despite long scantimes, but recruitment of patients was impeded by a complex set-up that required two visits on two separate days.Multi-parametric analysis revealed significant differences in imaging parameters, dependent on the outcome atthe time of tumor recurrence. These differences were exploited to establish a model of recurrence probabilitiesfrom images obtained at the time of radiotherapy planning. This novel method could potentially be used toinvestigate alternative treatment strategies.