UNLABELLED: Brain tumor volume assessment is a major challenge. Molecular imaging using PET may be a promising option because it reflects the biologically active cells. We compared the agreement between PET- and histology-derived tumor volumes in an orthotopic glioblastoma rat model with a noninfiltrating (U87MG) and an infiltrating (T87) tumor phenotype using 2 different radiotracers, 2 different image reconstruction algorithms, parametric imaging, and 2 different image segmentation techniques.

METHODS: Rats with U87MG- and T87-derived glioblastomas were continuously scanned with PET for 1 h starting immediately after the injection of 11C-methylaminoisobutyric acid (11C-MeAIB). One hour later, 18F-FDG was injected, followed by a 3-h dynamic PET scan. Images were reconstructed using 2-dimensional ordered-subsets expectation maximization and 3-dimensional maximum a posteriori probability (MAP3D) algorithms. In addition, a parametric image, encompassing the entire tumor kinetics in a single image, was calculated on the basis of the 11C-MeAIB images. All reconstructed images were segmented by fixed thresholding of maximum voxel intensity (VImax) and mean background intensity. The agreement between PET- and histology-derived tumor volumes and intra- and interobserver agreement of the PET-derived volumes were evaluated using Bland-Altman plots.

RESULTS: By PET, the mean U87MG tumor volume was 35.0 mm3 using 18F-FDG and 34.1 mm3 with 11C-MeAIB, compared with 33.7 mm3 by histology. Corresponding T87 tumor volumes were 122.1 mm3 using 18F-FDG, 118.3 mm3 with 11C-MeAIB, and 125.4 mm3 by histology. None of these volumes were significantly different. The best agreement between PET- and histology-derived U87MG tumor volumes was achieved with 11C-MeAIB, MAP3D reconstruction, and fixed thresholding of VImax. The intra- and interobserver agreement was high using this method. For T87 tumors, the best agreement between PET- and histology-derived volumes was obtained using 18F-FDG, MAP3D reconstruction, and fixed thresholding of mean background intensity. The agreement using 11C-MeAIB, parametric imaging, and fixed thresholding of VImax was slightly inferior, but the intra- and interobserver agreement was clearly superior.

CONCLUSION: Estimation of tumor volume by PET of noninfiltrating brain tumors was accurate and reproducible. In contrast, tumor volume estimation by PET of infiltrating brain tumors was difficult and hard to reproduce. On the basis of our results, PET evaluation of highly infiltrating brain tumors should be further developed.