Exploratory Evaluation of MR Permeability with ¹⁸F-FDG PET Mapping in Pediatric Brain Tumors: A Report from the Pediatric Brain Tumor Consortium

DISCUSSION

A comprehensive assessment of tumor anatomy and physiology often involves spatial correlation of imaging acquired on separate MR and PET scanners. MR imaging with contrast has high soft-tissue resolution, multiplanar capability, and lack of ionizing radiation, making it the modality of choice for determining tumor size, location, and characterization. MR permeability and ¹⁸F-FDG PET are advanced imaging techniques that assess different physiologic parameters and can be combined to provide insight into tumor physiology and complement anatomic information given by standard MR (T1, T2, fluid-attenuated inversion recovery, and postgadolinium T1) images. Further, MR and PET can both be used to assess tumor response or progression and to monitor treatment effects.

In clinical practice, the visual comparison of MR and brain ¹⁸F-FDG PET images placed side by side is often used to subjectively coregister images. Previous work in adults has evaluated the registration of MR diffusion images with PET images; however, no common ROI was applied to either study. For example, Palumbo et al. coregistered MR diffusion and PET images with ROIs independently drawn on MR and PET using MRIcro (http://www.sph.sc.edu/comd/rorden/mri-cro.html), a statistical parametric mapping tool for viewing medical images (16). Holodny et al. coregistered MR diffusion and PET images through the application of a 6-parameter rigid-body transformation matrix, followed by trilinear interpolation (17). Each transformation matrix was determined by a nonlinear gradient search for a maximum normalized mutual information cost function applied to the voxel intensities of each registered 3D image pair. ROIs were manually drawn on both MR and PET images using software from the McConnell Brain Imaging Centre of the Montreal Neurologic Institute. To effectively investigate the correlation between MR and PET in pediatric brain tumors, the registration of MR with PET images using a common ROI that can be applied to both studies is desirable.

In our study, we evaluated the feasibility of registering MR permeability and ¹⁸F-FDG PET images using a common ROI. MR and PET images were acquired on separate scanners, and fused PET/MR images were obtained. Although either T1 perfusion images or T2* gradient images could have been used, the PET images were registered with the last time point of the T1 perfusion time series instead of T2* gradient images to minimize susceptibility artifacts that can be present with T2* acquisitions (18). Our results suggest that this automated registration method can be used successfully to transfer an ROI manually drawn on MR permeability images to ¹⁸F-FDG PET images.

MR permeability can be used to assess vascular integrity. High-grade malignancy is often associated with increased macrovasculature and microvasculature and higher capillary leakage. Roberts et al. reported that in 22 patients aged 14–79 y with gliomas, MR permeability was correlated with pathologic grade (19). Law et al. reported that vascular permeability was correlated with tumor grade in 73 patients aged 4–85 y with primary gliomas, although the correlation was not as strong as expected (20). Several MR parameters can be used to evaluate tumor vascularity. Indeed, in the study by Law et al. (20), there was a stronger correlation between relative CBV (rCBV) and tumor grade than vascular permeability and tumor grade, suggesting that these parameters characterize different aspects of tumor physiology. The use of complementary MR parameters has been proposed for the evaluation of pediatric brain tumors (21), and it is possible that a combination of imaging parameters could be superior to any single parameter for tumor characterization. Indeed, Law et al. concluded that the combination of rCBV and vascular transfer constant was the best set of metrics to predict glioma grade but did not provide additional predictive value over rCBV alone (22).

Results in the literature on ¹⁸F-FDG PET in adult brain tumors report that more intense uptake suggests increased tumor malignancy (2), aggressiveness (3), grade (4,5), and cell density (6). Delbeke et al. retrospectively reviewed 58 patients with histologically proven brain tumors aged 2–81 y (mean age ± SD, 38 ± 25 y) referred for baseline ¹⁸F-FDG PET. A tumor–to–white matter ratio greater than 1.5 suggested high-grade disease, with a sensitivity of 94% and specificity of 77% (7). Barker et al. reported that more intense uptake was a significant predictor of poor survival in 55 patients with malignant glioma aged 11–65 y (median age, 45 y) (8). Patronas et al. (9) and Padma et al. (10) suggested PET could be used as a prognostic tool for survival.

Initial studies of ¹⁸F-FDG PET in children with brain tumors have suggested that intensely ¹⁸F-FDG–avid disease and extensive ¹⁸F-FDG uptake is associated with a higher histologic grade and worse outcome. Kwon et al. found that in 12 children (age range, 3–14 y), intense ¹⁸F-FDG uptake was more common in glioblastoma, whereas mild or no ¹⁸F-FDG uptake was more common in anaplastic astrocytomas or low-grade astrocytomas (23). Williams et al. found that in 12 children (age range, 4.6–17.4 y), 3D maximum and mean tumor ¹⁸F-FDG uptake were associated with PFS in supratentorial anaplastic astrocytomas when using 3D PET analysis techniques (11). Kruer et al. reported that in 46 children (age range, 4 mo to 18 y) with low-grade astrocytomas, ¹⁸F-FDG PET and proliferation index were independent useful measures in the risk stratification of disease (24). Pirotte et al. found that in 20 children (age range, 1–13 y), ¹⁸F-FDG PET guidance improved the diagnostic yield of stereotactic biopsy in newly diagnosed BSG and could have prognostic value (25). Gururangan et al. found that in 22 children (age range, 2.8–25 y) with either recurrent or newly diagnosed medulloblastoma there was a significant negative correlation between increased ¹⁸F-FDG uptake and survival (26). Holthoff et al. found that ¹⁸F-FDG uptake was found to be increased in patients with medulloblastoma, compared with other posterior fossa tumors (27). Borgwardt et al. found that in 38 children (age range, 0–16 y) with primary central nervous system tumors, increased ¹⁸F-FDG uptake reflected histologic grade and that ¹⁸F-FDG PET and MR coregistration improved tumor characterization (1). Of note, Borgwardt et al. performed both imaging with ¹⁸F-FDG PET and imaging with ¹⁵O-labeled water in 16 patients and found no correlation between cerebral blood flow and malignancy or cerebral blood flow and glucose metabolism.

Through the PBTC, we recently evaluated associations between MR imaging and ¹⁸F-FDG PET in pediatric diffuse intrinsic BSG (12). If ¹⁸F-FDG uptake was hyperintense, compared with gray matter, and was found in 50% or more of the tumor, survival was poor. We found an association of tumor diffusion values with uniformity of ¹⁸F-FDG uptake, suggesting that increased tumor cellularity represents a greater number of viable tumor cells and higher ¹⁸F-FDG uptake throughout the tumor. This finding agreed with reports by Palumbo et al. and Holodny et al. showing a correlation between lower apparent diffusion coefficient and increased ¹⁸F-FDG uptake (16,17). Further, we suggested that higher ¹⁸F-FDG uptake was associated with tumor enhancement, possibly reflecting more aggressive disease because most BSGs do not enhance (28).

The 24 children with brain tumors included in this study underwent both MR and ¹⁸F-FDG PET as part of a multiinstitutional phase II study of bevacizumab and irinotecan therapy. When we compared MR and PET across the spectrum of pediatric brain tumors, we found that there was no significant correlation of baseline Kps_max with T/W_max, although our sample size was small. This lack of correlation could reflect the fact that T1 permeability on MR and ¹⁸F-FDG uptake on PET measure different aspects of tumor physiology and may be complementary rather than correlative. Specifically, T1 permeability characterizes changes in vascular leakage associated with angiogenesis whereas ¹⁸F-FDG uptake is associated with cellular glucose metabolism where ¹⁸F-FDG freely crosses the blood–brain barrier. Of note, children with supratentorial HGG had slightly higher correlation than children with LGG, suggesting that correlation of Kps_max with T/W_max may also depend on tumor histology. However, a larger number of patients would be needed for a covariate analysis of Kps_max, T/W_max, and PFS by tumor histology. We found that baseline Kps_max was not significantly associated with CBV_max, whereas baseline CBV_max was significantly associated with T/W_max. Because radiotracer uptake is related to the delivery of radiotracer to the tumor, which in turn is reflected by blood flow and CBV, this result was not unexpected. Further, we found that CBV_max was slightly higher in the cohort of children with HGG and BSG than in those with LGG. This correlation likely reflects the correlation of higher grade tumors with increased CBV_max. There was no such association of tumor type with vascular permeability or ¹⁸F-FDG uptake. This result could be in part related to our small sample size. Law et al. previously suggested that there was a stronger correlation between rCBV and tumor grade than vascular permeability and tumor grade (20); therefore, there is a possibility that although our sample size was sufficiently large to show an association with CBV_max, it was not sufficient to show an association with Kps_max or T/W_max. Finally, T1 permeability and CBV were not significantly associated with PFS. ¹⁸F-FDG uptake also was not significantly associated with PFS; however, there was a suggestion that such an association may be shown with an increase in the number of subjects.

The main limitations of this study were the small sample size and the mixture of tumor types. However, this was an exploratory evaluation of the feasibility of an approach for combining ¹⁸F-FDG PET with MR permeability images after independent acquisition in the pediatric population that resulted in good to excellent coregistration of the tumor ROI across a spectrum of new and recurrent disease. In future studies, a multivariate analysis of both MR and PET parameters with PFS would be helpful using this approach in a homogeneous histologic group. However, there were not sufficient numbers of subjects in this preliminary investigation to test such a hypothesis. Future studies could also investigate associations of MR imaging with ¹⁸F-FDG PET on hybrid PET/MR systems or the benefits of other PET radiopharmaceuticals such as ¹¹C-labeled methionine, ¹⁸F-labeled choline, or ¹⁸F-labeled fluorothymidine, among others. In addition, we used a simple ROI analysis to illustrate proof-of-principle that there is good anatomic correlation between the ¹⁸F-FDG PET and MR permeability in a heterogeneous group of pediatric brain tumors. However, voxel-based analysis could also be considered in future expanded studies using this approach to characterize tumor tissue, to possibly identify active tumor regions and to monitor changes during therapy.