18F-FDG and L-methyl-11C-methionine (MET) are the most effective tracers for PET imaging of brain tumors.1–4 FDG accumulates in proportion to tumor malignancy, which is useful to differentiate high-grade from low-grade malignant tumors.1,4–6 However, FDG also accumulates in gray matter depending on neuronal activity, which often interferes with tumor evaluation. In contrast, MET distribution in normal gray matter is very low; therefore, MET-PET enables easy identification of tumor extent. In clinical settings, these PET images are interpreted by nuclear medicine physicians who consider information regarding the degree of tracer uptake, the extent of their altered distribution, and the differences between these 2 tracers to diagnose malignancy grade.
Semiquantitative metabolic indices provide objective evaluation and more detailed stratification of tracer uptake intensity. However, these metabolic indices do not always successfully separate high- from low-grade gliomas because of the considerable overlap in metabolic indices between these 2 grades.6 This may, in part, be due to the method of calculating metabolic index, which has focused solely on the maximum intensity of tumor uptake, namely, the tumor core area. These metabolic indices do not contain an important aspect of the invasive characteristic of glioma.7 Tumor cell infiltration interferes with normal neuronal activity, and decreases FDG uptake in normal gray matter; however, it is difficult to depict the tumor infiltrative area as higher uptake on PET imaging due to a microscopic partial volume effect. Attempts to estimate the extent of the infiltrative area have been made by some authors. Kinoshita et al8,9 characterized infiltrative area according to MET uptake, which was even lower than the cutoff value, but was significantly higher than the predicted value estimated by the FDG uptake at the same area. Histopathologically, MET accumulation has high sensitivity for tumor infiltration (89%).10 Bergstrom et al11 reported the concordance of MET distribution and tumor infiltration at autopsy. These previous reports suggest that optimal analysis of both MET-PET and FDG-PET may lead to more precise estimation of the infiltrative area, which may provide additional information about glioma malignancy.
In the present study, we first determined the most optimal metabolic index focusing on tumor core area because there are various indices, such as tumor-to-normal (T/N) ratios with various calculation methods for reference value or peak SUV itself. Then, using the most optimal metabolic index focusing on tumor core, we characterized them based on their pathological grade and molecular markers according to the 2016 World Health Organization (WHO) classification.12 To separate tumors that were indistinguishable according to T/N ratio, we then developed a new method, focusing on metabolic alteration on the surrounding tumor core area, probably reflecting tumor cell infiltration. We hypothesized that infiltration areas exhibited increases in MET uptake, even slightly, and lower FDG uptake than normal gray matter. Supporting this hypothesis, we calculated the volume of the potential infiltration area using voxel-based analysis with both FDG-PET and MET-PET, and compared the volume with their pathological grade.
PATIENTS AND METHODS
Sixty-three patients who underwent both FDG-PET and MET-PET for the evaluation of brain tumor(s) in the authors' department between March 2009 and August 2016 were identified. All PET images were obtained before any oncologic treatments were administered. FDG-PET and MET-PET were performed within 2 weeks of one another. The pathological diagnosis was verified histologically in surgical specimens and reperformed using genetic methods according to the 2016 WHO guidelines. The interval between the date of surgery and PET scans ranged from 2 days to 30 weeks (mean, 29 days). All of these patients (39 men, 24 women; mean age, 50 ± 18 years) had untreated primary glioma (Table 1). The patients were divided into 2 groups: 44 high-grade gliomas including 22 glioblastoma and 22 anaplastic gliomas; and 19 low-grade gliomas including 4 diffuse astrocytoma, 13 oligodendroglioma, one ependymoma, and one dysembryoplastic neuroepithelial tumor (DNT). Written informed consent was obtained from all patients for taking MET-PET images before any administration of oncologic treatment as well as for the PET data to be utilized for this study. This retrospective study was approved by the institutional review board of the authors' hospital.
The patients fasted for at least 5 hours before FDG-PET imaging. Patients rested in the supine position with an eye mask in a quiet PET room to minimize the confounding factors of environmental noises. Each patient was administered 296 MBq (8 mCi) 18F-FDG intravenously up to January 2011, and 4.5 MBq/kg (0.12 mCi/kg) thereafter. Emission scans were obtained 45 minutes later in 3-dimensional mode for 10 minutes using a PET/CT scanner (Aquiduo; Toshiba Medical Systems, Otawara, Japan). Photon attenuation correction was performed using a low-dose CT scan. The PET scanner was equipped with 24,336 lutetium oxyorthosilicate crystals in 39 detector rings, and had an axial field of view of 16.2 cm, and 82 transverse slices of 2.0 mm thickness. The intrinsic full width at half-maximum (FWHM) spatial resolution at the center of the field of view was 4.3 mm, and the FWHM axial resolution was 4.7 mm. PET images were reconstructed using Fourier rebinning ordered subset expectation maximization iterative reconstruction, with 2 iterations and 8 subsets; a 4-mm FWHM Gaussian filter was applied. The data were collected in a 256 × 256 × 41 matrix with a voxel size of 0.95 × 0.95 × 4 mm.
For MET-PET imaging, a 740-MBq (20 mCi) dose of MET was injected intravenously, and a 10-minute emission scan was started 30 minutes after the injection. The PET/CT scanner and image reconstruction protocols were the same as the protocols used for FDG-PET imaging.
To conveniently analyze PET images, all voxel values from PET images were normalized to SUV using patient body weight (grams), injected radioactivity (becquerel/millileter), and a cross-calibration factor (becquerel/counts per second), assuming a specific gravity of 1 g/mL.
Metabolic Indices of the Tumor Core
In the first step, focusing on the highest metabolic value within tumor core area, 5 types of metabolic indices were calculated including various T/N ratios and maximum value within the tumor itself. A circular region of interest (ROI), 10 mm in diameter, was placed on the “hottest” tumor area to calculate both tumor maximum value (T_max), which refers to the highest voxel value within the ROI, and tumor mean value (T_mean), which refers to the average voxel values of the ROI. A normal reference value was calculated using 2 methods: one was a manual method (N_manu), in which 4 circular ROIs, 10 mm in diameter, were manually placed on contralateral normal cortex identified by operator visual inspection concerning CT or MRI; and the other was a fully automated method (N_aut), in which normal cortex voxels were automatically identified using a method developed by the authors in a previous study.13 Briefly, N_aut was calculated using a fully automated method, in which the normal gray matter was identified using voxel-based analysis based on 2 assumptions: normal gray matter exhibits a relatively high uptake of FDG, and the tumor does not exceed more than one half of the brain cortex area on MET-PET in most clinical settings. This method provides the same coordination of voxels, which are most likely to be normal gray matter on intrasubject coregistered FDG-PET and MET-PET images, and demonstrated excellent agreement with those determined by a conventional ROI method by experts' visual inspection. The pathological results and these indices were then compared.
The best indices for distinguishing between high- and low-grade glioma were identified; these metabolic indices were then characterized according the 2016 WHO classification.
Estimation of Infiltrative Area
For distinguishing high- and low-grade glioma, the tumor core metabolic index has been helpful if the tumor clearly includes high uptake areas, such as FDG/N ≥ 0.9 and/or MET/N ≥ 3.0; however, investigators often encounter tumors with an equivocal tumor core metabolic index such as FDG/N < 0.9 and MET/N < 3.0. In fact, the metabolic changes and its extent of surrounding tumor core, as well as the tumor core itself, were used to differentiate these equivocal tumors.
To distinguish high- and low-grade glioma in tumors that exhibit neither FDG N ≥ 0.9 nor MET/N ≥ 3.0, we hypothesized that the altered metabolic area exhibiting FDG/N < 0.9 and MET/N < 3.0 was divided into 3 parts: tumor core, infiltrative area, and other benign metabolic changes (Fig. 1). Tumor infiltrative areas surrounding the tumor core exhibit higher-than-normal MET uptake, even if slight. At the same time, infiltrating tumor cells interfere with normal neuronal activities; therefore, these areas exhibit lower FDG uptake than normal gray matter. An attempt was made to develop a method to calculate tumor core and infiltration area based on this hypothesis, in which the image processes are illustrated in Figure 2.
Image Processes of the New Method
To estimate tumor core and infiltrative area surrounding the tumor core, FDG-PET and MET-PET images were processed as follows:
- Coregistration between FDG-PET and MET-PET in each patient was performed to analyze FDG-PET and MET-PET images with same coordination.
- Voxel intensity was normalized by dividing the voxel value for each voxel by N_aut. These were designated as “MET/N map” for MET-PET and “FDG/N map” for FDG-PET, respectively.
- The candidate area of tumor was defined as MET/N > 1.1, excluding outside of the brain by FDG/N > 0.3. Among these areas, the tumor core pulse potential infiltrative area was defined as FDG/N < 0.9, to exclude normal gray matter and tumor core area as MET/N > 1.4:
- Tumor core and potential infiltrative area = (0.3 < FDG/N < 0.9) and (MET/N > 1.1);
- Tumor core area = (0.3 < FDG/N) and (MET/N > 1.4)
Accordingly, potential infiltrative area = area (A) – area (B)
If there is no candidate area of tumor (MET/N > 1.1), this new method cannot be applied.
These processes were programmed using Statistical Parametric Mapping 8 (SPM8) and MATLAB version R2014a (MathWorks Inc, Natick, MA). In addition, they were verified visually, and the continuous area of the tumor core was extracted by manually specifying 2 or 3 points clearly within the tumor core, which were added to the program using the MATLAB software.
The selected areas were converted to volume by multiplying voxel volume (0.003627 mL/voxel). Tumors with a larger potential infiltrative area suggested high-grade glioma. This was subsequently applied to the method for tumors that did not have apparently high metabolic areas suggestive of high-grade glioma.
To determine the optimal index for discriminating high- from low-grade gliomas, receiver operating characteristic (ROC) curves were plotted, and the areas under the ROC curve (AUC) were calculated. Comparisons of index values or potentially infiltrative areas between high- and low-grade gliomas were performed using the 2-tailed t test; P < 0.05 was considered to be statistically significant. Statistical analysis was performed using SPSS version 20.0 (IBM Corporation, Armonk, NY).
Metabolic Indices of the Tumor Core
ROC curves are shown in Figure 3, and AUC, sensitivity, and specificity, using cutoffs derived from the ROC curves, are presented in Table 2. The greatest AUC was obtained for FDG-PET of T_mean/N_aut; MET-PET of T_mean/N_aut was also satisfactory. Therefore, the authors proceeded to the following investigations using T_mean/N_aut for both FDG-PET and MET-PET.
A scatter plot of T_mean/N_aut according to the 2016 WHO classification is shown in Figure 4. T_mean/N_aut was 0.9 or greater on FDG-PET and/or 3.0 or greater on MET-PET in 30 tumors, 28 of which were high-grade gliomas, with a positive predictive value of 93%. The other 33 tumors exhibited T_mean/N_aut less than 0.9 on FDG-PET and less than 3.0 on MET-PET, including 16 high-grade and 17 low-grade gliomas, in which the T_mean/N_aut overlapped between the high- and low-grade gliomas. The T_mean/N_aut values according to subtype are presented in Table 1.
Estimation of Infiltrative Area
The 33 tumors had equivocal tumor core metabolic index, namely, T_mean/N_aut less than 0.9 on FDG/N and less than 3.0 on MET/N. Nine of these tumors, composed of 4 high-grade tumors and 5 low-grade tumors, did not have areas exhibiting MET/N > 1.1. Therefore, the new method was applied to the remaining 24 tumors. The difference in calculated tumor core volume (milliliters) between high- and low-grade gliomas was not statistically significant (18.6 ± 24.6 vs 8.3 ± 27.5, respectively; P = 0.17 [t test]; Fig. 5A). The difference in potential infiltrative volumes between high- and low-grade gliomas was statistically significant (43.8 ± 30.2 vs 14.0 ± 12.6 mL, respectively; P = 0.005 [t test]; Fig. 5B). Representative images are shown in Figure 6.
As shown in Figure 5B, 11 of 12 low-grade tumors had potential infiltrative volume less than 20.0 mL, whereas only 3 of 12 tumors were high-grade tumors. Combining tumor core evaluation with a cutoff value of T_mean/N_aut 0.9 or greater on FDG-PET and/or 3.0 or greater on MET-PET and potential infiltrative volume with a cutoff value of 20.0 mL yielded a diagnostic accuracy of 89% in distinguishing between high- and low-grade gliomas; this was under the condition that 9 patients, who did not have candidate area of tumor (T_mean/N_aut ≥ 1.1 on MET-PET), were excluded. The diagnostic flowchart of our results is shown in Figure 7.
We demonstrated that tumors with high uptakes either of FDG or of MET had high positive predictive value for high-grade tumor (93%) with the cutoff value for the T/N ratio of 0.9 or greater on FDG-PET and/or 3.0 or greater on MET-PET. However, only 48% of all tumors exhibited T/Ns higher than these cutoff values in this study. Others were under these cutoffs, and their T/N ratios overlapped between high- and low-grade tumors. The metabolic alteration surrounding the tumor core was investigated to calculate a potentially infiltrative area using FDG-PET and MET-PET. This result demonstrated that the volume was significantly larger in high-grade than in low-grade gliomas. Finally, when we used both the estimated infiltrative volume surrounding the tumor core and the metabolic index of the tumor core, differentiation between high-grade and low-grade glioma reached a sensitivity of 84% and a specificity of 84%. Only the tumor core volume did not statistically differ between high-grade and low-grade gliomas. Our new method estimating infiltrative area can be applied in clinical settings when both FDG-PET and MET-PET are performed, and would be helpful even for visual inspection of these 2 PET images side by side. If the metabolic index of the tumor core is not helpful, and when the glucose hypometabolism area associated with even slightly increased MET uptake is larger than the tumor core area, it may suggest high-grade glioma. This study is the first to demonstrate a metabolic index according to the 2016 WHO classification and to differentiate the metabolic pattern of FDG-PET and MET-PET on surrounding tumor core area between high- and low-grade glioma.
Metabolic indices, especially T/N ratios, have been applied to the characterization of brain tumor(s) and demonstrated satisfactory correlation with tumor malignancies; however, the variety of calculation methods has confounded objective evaluation. In addition, normal reference values are, inevitably, operator-dependent. We developed a new automated method to calculate normal reference value in our previous study and applied it in the present study.13 The T/N ratio, which was calculated using the tumor core ROI mean value divided by the normal reference value of our automated method, demonstrated the best performance. Other T/N ratios also performed better than tumor max alone. Tumor max, namely, the maximum SUV value of the tumor, was also reported to be less effective.14 This is probably because SUV is susceptible to various factors such as body composition, blood sugar level, and image reconstruction methods.15 As a result, T/N ratios in both FDG-PET and MET-PET images were correlated with tumor malignancy grade, except for tumors with isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion. Oligodendroglial differentiation was also reported to positively affect the metabolic index.6 In this study, 2 tumors unsuccessfully classified, which showed higher than the threshold of T/N ratio of 0.9 or greater on FDG-PET and/or 3.0 or greater on MET-PET, were oligodendroglioma, IDH mutant and 1p/19q codeleted.
We hypothesized that areas with MET increases, even if slight, and FDG lower than normal gray matter in the surrounding tumor core area suggest a potentially infiltrative area. An autopsy study reported that MET uptake was larger than FDG uptake, and MET uptake was highly correlated with tumor extent.11 Kinoshita et al8,9 calculated a decoupling score, which suggests that a higher MET uptake than that estimated by FDG uptake, even if it was within the reference range, correlated with tumor cell infiltration. A volumetry MET-PET study also reported that a large extent of MET uptake was a more significant prognostic factor than T/N ratio of the hottest area of the tumor.16 Therefore, MET uptake can be a good predictor for tumor cell extent, and accuracy is probably improved by comparison with FDG-PET. In addition, the larger extent of brain tumor where tumor cells are involved probably reflects a higher malignancy.
We set the threshold of the T/N ratio of MET-PET at 1.4 to identify the tumor core area. The value was concordant with ratios reported in previous studies, such as for differentiating tumors and nontumor lesions resulting in thresholds of 1.4717 and 1.3,10 a mean value of low-grade tumor at 1.5,14 and median threshold value for surgery planning at 1.41.18 Although the precise method for calculating these T/N ratios was different, we believe that 1.4 is reasonable. The cutoff value to extract potentially infiltrative volume was decided simply as 10% higher than the reference value (T/N ratio 1.1 of MET-PET) and lower than that of the normal gray matter value (T/N ratio 0.9 of FDG-PET). T/N of 1.1 on MET-PET is supported by previous studies, in which the nontumor tissues exhibited 1.09 ± 0.36 by biopsy,10 and most normal brain regions were lower than 1.1.19 Elimination of the voxels outside the brain was performed using T/N ratios less than 0.3 of FDG-PET, which is dependent on visual investigation.
Our proposed method for calculating “potential infiltrative tumor volume” cannot be applied for high FDG/N because the delineation of the tumor cannot be determined. In this study, we defined the tumor core as MET/N ≥ 1.4; however, if the area of FDG/N ≥ 0.9 is larger than the area of MET/N ≥ 1.4 around the tumor core, the areas that shows both FDG/N ≥ 0.9 and MET/N ≤ 1.4 may be either high-grade tumor or normal cortex, because MET uptake has been reported to reach approximately MET/N of 1.4 even in the normal cortex.20 We designated the area with 1.1 ≤ MET/N ≤ 1.4 as abnormal (tumor infiltration) because FDG accumulation was reduced, such as FDG/N ≤ 0.9, due to the interference of normal neuronal activity by tumor cell infiltration.
We did not confirm the pathological diagnosis (ie, whether the area we estimated as the potential tumor infiltrative area actually demonstrated tumor infiltration). Because this was a retrospective investigation, identification of the pathological examination site on PET images was difficult. A follow-up study for the existence of recurrence within the calculated potential infiltrative area that was not treated by surgery or radiation therapy may be helpful to confirm the efficacy of our method.
1. Kaschten B, Stevenaert A, Sadzot B, et al. Preoperative evaluation of 54 gliomas by PET
with fluorine-18-fluorodeoxyglucose and/or carbon-11-methionine. J Nucl Med
2. Chung JK, Kim YK, Kim SK, et al. Usefulness of 11
in the evaluation of brain lesions that are hypo- or isometabolic on 18
. Eur J Nucl Med Mol Imaging
3. Van Laere K, Ceyssens S, Van Calenbergh F, et al. Direct comparison of 18
F-FDG and 11
in suspected recurrence of glioma: sensitivity, inter-observer variability and prognostic value. Eur J Nucl Med Mol Imaging
4. Borbely K, Nyary I, Toth M, et al. Optimization of semi-quantification in metabolic PET
studies with 18
F-fluorodeoxyglucose and 11
C-methionine in the determination of malignancy of gliomas. J Neurol Sci
5. Borgwardt L, Hojgaard L, Carstensen H, et al. Increased fluorine-18 2-fluoro-2-deoxy-D-glucose (FDG) uptake in childhood CNS tumors is correlated with malignancy grade: a study with FDG positron emission tomography/magnetic resonance imaging coregistration and image fusion. J Clin Oncol
6. Manabe O, Hattori N, Yamaguchi S, et al. Oligodendroglial component complicates the prediction of tumour grading with metabolic imaging. Eur J Nucl Med Mol Imaging
7. Burger PC, Heinz ER, Shibata T, et al. Topographic anatomy and CT correlations in the untreated glioblastoma multiforme. J Neurosurg
8. Kinoshita M, Arita H, Goto T, et al. A novel PET
C-methionine uptake decoupling score, reflects glioma cell infiltration. J Nucl Med
9. Kinoshita M, Goto T, Arita H, et al. Imaging 18
C-methionine uptake decoupling for identification of tumor cell infiltration in peritumoral brain edema. J Neurooncol
10. Kracht LW, Miletic H, Busch S, et al. Delineation of brain tumor
extent with [11
C]L-methionine positron emission tomography: local comparison with stereotactic histopathology. Clin Cancer Res
11. Bergstrom M, Collins VP, Ehrin E, et al. Discrepancies in brain tumor
extent as shown by computed tomography and positron emission tomography using [68
C]glucose, and [11
C]methionine. J Comput Assist Tomogr
12. Louis DN, Perry A, Reifenberger G, et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol
13. Takahashi M, Soma T, Mukasa A, et al. An automated voxel-based method for calculating the reference value for a brain tumour metabolic index using 18
. Ann Nucl Med
14. Utriainen M, Metsähonkala L, Salmi TT, et al. Metabolic characterization of childhood brain tumors: comparison of 18
F-fluorodeoxyglucose and 11
C-methionine positron emission tomography. Cancer
15. Keyes JW Jr. SUV: standard uptake or silly useless value? J Nucl Med
16. Galldiks N, Dunkl V, Kracht LW, et al. Volumetry of [11
C]-methionine positron emission tomographic uptake as a prognostic marker before treatment of patients with malignant glioma. Mol Imaging
17. Herholz K, Holzer T, Bauer B, et al. 11
for differential diagnosis of low-grade gliomas. Neurology
18. Arbizu J, Tejada S, Marti-Climent JM, et al. Quantitative volumetric analysis of gliomas with sequential MRI and 11
assessment: patterns of integration in therapy planning. Eur J Nucl Med Mol Imaging
19. Coope DJ, Cizek J, Eggers C, et al. Evaluation of primary brain tumors using 11
with reference to a normal methionine uptake map. J Nucl Med
20. Uda T, Tsuyuguchi N, Terakawa Y, et al. Evaluation of the accumulation of 11
C-methionine with standardized uptake value in the normal brain. J Nucl Med
Keywords:Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.
brain tumor; 18F-FDG; 11C-MET; PET; tumor-to-normal ratio; voxel-based analysis