Voxel-based analysis of dual-time-point 18F-FDG PET images for brain tumor identification and delineation

Abstract

We have investigated dual-time-point (18)F-FDG PET for the detection and delineation of high-grade brain tumors using quantitative criteria applied on a voxel basis.

Methods: Twenty-five patients with suspected high-grade brain tumors and inconclusive MRI findings underwent (11)C-methionine PET and dual-time-point (18)F-FDG PET. Images from each subject were registered and spatially normalized. Parametric maps of standardized uptake value (SUV) and tumor-to-normal gray matter (TN) ratio for each PET image were obtained. Tumor diagnosis was evaluated according to 4 criteria comparing standard and delayed (18)F-FDG PET images: any SUV increase, SUV increase greater than 10%, any TN increase, and TN increase greater than 10%. Voxel-based analysis sensitivity was assessed using (11)C-methionine as a reference and compared with visual and volume-of-interest analysis for dual-time-point PET images. Additionally, volumetric assessment of the tumor extent that fulfills each criterion was compared with the volume defined for (11)C-methionine PET.

Results: The greatest sensitivity for tumor identification was obtained with any increase of TN ratio (100%), followed by a TN increase greater than 10% (96%), any SUV increase (80%), and an SUV increase greater than 10% (60%). These values were superior to visual analysis of standard (18)F-FDG (sensitivity, 40%) and delayed (18)F-FDG PET (sensitivity, 52%). Volume-of-interest analysis of dual-time-point PET reached a sensitivity of only 64% using the TN increase criterion. Regarding volumetry, voxel-based analysis with the TN ratio increase as a criterion, compared with (11)C-methionine PET, detected 55.4% of the tumor volume, with the other criteria detecting volumes lower than 20%. Nevertheless, volume detection presented great variability, being better for metastasis (78%) and glioblastomas (56%) than for anaplastic tumors (12%). A positive correlation was observed between the volume detected and the time of acquisition of the delayed PET image (r = 0.66, P < 0.001), showing volumes greater than 75% when the delayed image was obtained at least 6 h after (18)F-FDG injection.

Conclusion: Compared with standard (18)F-FDG PET studies, quantitative dual-time-point (18)F-FDG PET can improve sensitivity for the identification and volume delineation of high-grade brain tumors.