The hypothesis of this study was that distinct experimental glioblastoma phenotypes resembling human disease can be noninvasively distinguished at various disease stages by imaging in vivo. Methods: Cultured spheroids from 2 human glioblastomas were implanted into the brains of nude rats. Glioblastoma growth dynamics were followed by PET using 18F-FDG, 11C-methyl-L-methionine (11C-MET), and 3′-deoxy-3′-18F-fluorothymidine (18F-FLT) and by MRI at 3-6 wk after implantation. For image validation, parameters were coregistered with immunohistochemical analysis. Results: Two tumor phenotypes (angiogenic and infiltrative) were obtained. The angiogenic phenotype showed high uptake of 11C-MET and 18F-FLT and relatively low uptake of 18F-FDG. 11C-MET was an early indicator of vessel remodeling and tumor proliferation. 18F-FLT uptake correlated to positive Ki67 staining at 6 wk. T1- and T2-weighted MR images displayed clear tumor delineation with strong gadolinium enhancement at 6 wk. The infiltrative phenotype did not accumulate 11C-MET and 18F-FLT and impaired the 18F-FDG uptake. In contrast, the Ki67 index showed a high proliferation rate. The extent of the infiltrative tumors could be observed by MRI but with low contrast. Conclusion: For angiogenic glioblastomas, noninvasive assessment of tumor activity corresponds well to immunohistochemical markers, and 11C-MET was more sensitive than 18F-FLT at detecting early tumor development. In contrast, infiltrative glioblastoma growth in the absence of blood-brain barrier breakdown is difficult to noninvasively follow by existing imaging techniques, and a negative 18F-FLT PET result does not exclude the presence of proliferating glioma tissue. The angiogenic model may serve as an advanced system to study imaging-guided antiangiogenic and antiproliferative therapies.