Characterizing hypoxia in human glioma: A simultaneous multimodal MRI and PET study.

Hypoxia plays an important role for the prognosis and therapy response of cancer. Thus, hypoxia imaging would be a valuable tool for pre-therapeutic assessment of tumor malignancy. However, there is no standard validated technique for clinical application available yet. Therefore, we performed a study in 12 patients with high-grade glioma, where we directly compared the two currently most promising techniques, namely the MR-based relative oxygen extraction fraction (MR-rOEF) and the PET hypoxia marker H-1-(3-[ F]-fluoro-2-hydroxypropyl)-2-nitroimidazole ([ F]-FMISO). MR-rOEF was determined from separate measurements of T , T * and relative cerebral blood volume (rCBV) employing a multi-parametric approach for quantification of the blood-oxygenation-level-dependent (BOLD) effect. With respect to [ F]-FMISO-PET, besides the commonly used late uptake between 120 and 130 min ([ F]-FMISO ), we also analyzed the hypoxia specific uptake rate [ F]-FMISO-k , as obtained by pharmacokinetic modeling of dynamic uptake data. Since pharmacokinetic modeling of partially acquired dynamic [ F]-FMISO data was sensitive to a low signal-to-noise-ratio, analysis was restricted to high-uptake tumor regions. Individual spatial analyses of deoxygenation and hypoxia-related parameter maps revealed that high MR-rOEF values clustered in (edematous) peritumoral tissue, while areas with high [ F]-FMISO concentrated in and around active tumor with disrupted blood-brain barrier, i.e. contrast enhancement in T -weighted MRI. Volume-of-interest-based correlations between MR-rOEF and [ F]-FMISO as well as [ F]-FMISO-k , and voxel-wise analyses in individual patients, yielded limited correlations, supporting the notion that [ F]-FMISO uptake, even after 2 h, might still be influenced by perfusion while [ F]-FMISO-k was severely hampered by noise. According to these results, vascular deoxygenation, as measured by MR-rOEF, and severe tissue hypoxia, as measured by [ F]-FMISO, show a poor spatial correspondence. Overall, the two methods appear to rather provide complementary than redundant information about high-grade glioma biology.