Potential for differentiation of glioma recurrence from radionecrosis using integrated 18F-fluoroethyl-L-tyrosine (FET) positron emission tomography/magnetic resonance imaging: A prospective evaluation
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/neuroindia.NI_101_16
Source of Support: None, Conflict of Interest: None
Purpose: To assess the utility of 18F-fluoroethyl-L-tyrosine (FET) positron emission tomography/magnetic resonance imaging (PET/MRI) in distinguishing recurrence from radionecrosis.
Keywords: Amino acid PET, 18F-fluoroethyl-L-tyrosine (18F-FET), glioma, radiation necrosis, recurrence, simultaneous PET/MRI
Traditionally, the management of high-grade gliomas (HGG) comprises early detection and prompt maximal surgical resection followed by postoperative radiotherapy (RT). Relatively recently, compared to RT alone, Stupp et al., demonstrated the advantage of concurrent temozolomide (TMZ) based chemoradiotherapy (chemoRT) establishing it as the standard of care in glioblastoma multiforme (GBM) management. Despite these advancements, tumor recurrence is frequent, with more than 90% cases showing recurrence at or within 2 cm of the primary site.
However, as a consequence of this therapy intensification, pseudoprogression (PsP), defined as increased contrast enhancement on follow-up magnetic resonance imaging (MRI) with subsequent resolution without alteration of therapy, is being increasingly recognized to the tune of 6–31%., These changes are attributable either to the increased rates of RT necrosis  and therapy-induced alterations in the blood–brain barrier affecting contrast leakage. Conventional MRI has little value in differentiating PsP from true early progressive disease (ePD) as both recurrence and radiation necrosis present as enhancing lesion(s) on a contrast enhanced MRI (CE MRI). The established use of anti-angiogenic agent bevacizumab in recurrent GBM further complicates disease progression assessment. However, distinguishing radiation necrosis/PsP from ePD is essential because, while ePD indicates treatment failure necessitating a change in therapy, chemoRT-induced necrosis is often a marker of treatment success.
In this context, advanced MRI techniques have become available such as diffusion weighted imaging (DWI), diffusion tensor imaging (DTI), perfusion MR imaging, and magnetic resonance spectroscopy (MRS) allowing differentiation of recurrence from radiation necrosis/pseudoprogression, albeit with some associated problems.,,,
Further, among several traditional and novel PET tracers such as 2-deoxy-2-(18F) fluoro D-glucose (18F-FDG), 11C-L-methionine (11C-MET), 18Ffluoroethyl-L-tyrosine (18F-FET), 6-[fluoride-18] fluorolevodopa (18F-DOPA), or 3'-deoxy-3'-[18F] fluorothymidine (18F-FLT) investigated,, amino-acid based tracers such as 11 C-MET and 18 F-FET appear most promising. While the former suffers from the drawback of a short half-life of 11 C (20 minutes) and requirement of an on-site cyclotron, the latter has been extensively investigated in primary diagnosis, tumor therapy assessment, and recurrence detection.18 F-FET PET in conjunction with MRI has been found to be useful for diagnostic and therapeutic assessment for both neurosurgery  and RT planning.
Hence, it appears logical that combination of 18 F-FET PET with MRI might significantly aid in discriminating recurrence from RT necrosis. As fusion of images across modalities requires additional software capabilities and is riddled with registration artifacts, a single session multiparametric analysis is now feasible with the advent of simultaneous integrated PET/MRI. This has ensured accurate spatiotemporal co-registration, significantly reduced time of measurement, with improved patient compliance. Hence, we performed this prospective evaluation aiming to assess the utility of simultaneous 18 F-FET PET/MRI in distinguishing treated cases of glioma with suspected recurrence from radiation necrosis.
In a prospective study, 32 patients (25 males, 7 females; mean age ± SD: 52.53 ± 15.75 years; range: 17–80 years) with a histologically diagnosed glioma after their debulking surgery and after having undergone chemoRT were enrolled for a simultaneous 18 F-FET PET/MRI evaluation. At the time of recruitment in the study, all had a high index of suspicion for recurrence clinically and/or when assessed on the follow up CE MRI.
The details of the patients are available in [Table 1]. Written informed consents for conducting an 18 F-FET PET/MRI scan and scientific evaluation of data was obtained from each individual participant. The study and analysis of data was approved by the local institutional review board.
Patients were excluded from the study if they had any primary cerebral pathology, non-glial histopathology, proven cerebral metastasis or a known primary other than a glioma, pregnancy and presence of any standard contraindications for an MRI examination such as magnetic metal implants, pacemakers, etc.
Post 18 F-FET PET/MRI examination, patients having tumors and showing congruent contrast enhancement with FET uptake underwent neurosurgical resection for histopathological confirmation, whereas in patients without FET uptake or patients deemed unfit for surgery, imaging and clinical follow up was performed.
Positron emission tomography/magnetic resonance imaging
Simultaneous PET/MRI was performed on the Biograph mMR (Siemens, Erlangen, Germany), which consists of a 3-T MRI scanner harboring a fully functional PET system equipped with avalanche photodiode technology. MR scanner featuring a high-performance gradient system (45 mT/m) with a slew rate of 200 T/(ms) is equipped with total imaging matrix coil technology, covering the entire body with multiple integrated radiofrequency surface coils.
After fasting for at least 4 hours, 207.2 ± 25 MBq (5.6 ± 0.67 mCi) of 18 F-FET was injected intravenously in the antecubital vein. Brain 18 F-FET PET/MRI was acquired immediately after tracer injection for a period of 25 minutes, during which various MRI sequences were carried out over continuous PET acquisition including a transversal T1-weighted ultrashort echo time (UTE) for attenuation correction, transversal two-dimensional (2D)-encoded fluid attenuated inversion recovery (FLAIR) sequence, T2-weighted turbo spin-echo sequence, sagittal three-dimensional (3D)-encoded magnetization-prepared rapid-acquisition gradient-echo (MPRAGE) sequence, diffusion weighted imaging (DWI), susceptibility weighted imaging (SWI), and perfusion echo planar imaging (EPI). In each case, 3D 1 H MRS was acquired and included the following parameters: a point-resolved spectroscopy sequence (PRESS), which included six presaturation bands placed around the volume of interest (VOI) to minimize contamination by air filled sinuses, osseous, and cerebrospinal fluid (CSF) containing structures. The details of acquisition protocol and parameters are available in [Table 2].
Post-acquisition, PET, MRI, and PET/MRI images were reviewed at syngo.via platform (Siemens, Germany) using the mMR general workflow.
The 18 F-FET PET-MRI data were analyzed by a radiologist and two nuclear medicine physicians, with more than 5, 10, and 5 years of MRI and Positron Emission Tomography computed tomography (PET/CT) experience, respectively. Diagnosis was made by reaching a consensus.
Region of interest selection
For enhancing lesions, software based T1-weighted post-contrast subtraction series was generated by image-by-image subtraction of the pre-contrast series from each post-contrast series of each patient, hence excluding necrosis/altered blood products. A free hand manual region of interest (ROI) was drawn over the area of maximum enhancement in the most representative slice and stored. Relative cerebral blood volume (rCBV) maps were derived using T2* perfusion imaging data and subsequently normalized with contralateral white matter generating a normalized rCBV (N rCBV) map. For diffusion image, data apparent diffusion coefficient (ADC) maps were generated. The stored free hand ROI drawn on the T1-weighted subtracted image (described above) was duplicated using a copy-paste function over the ADC and N rCBV maps and the FET PET images yielding corresponding simultaneous ADCmean, N rCBVmean and standardized uptake values (SUVmax and SUVmean), respectively. The most representative spectra observed within this ROI was employed for choline/creatine (Cho/Cr) ratio calculation.
ROI size, hence, differed for each patient. In cases with absent/minimal enhancement, PET images showing the area of maximum FET uptake was chosen for drawing the ROI.
For background FET uptake value calculation, the same ROI as the ROI of the brain lesion was placed over the contralateral hemisphere (mirror region), which was obviously not affected.
Tumor to background ratios, TBRmax and TBRmean, values were calculated by dividing maximum and mean SUV values of tumor FET uptake with the mean background FET uptake. These values were used for further statistical analysis
Correlations between MRI parameters, N rCBVmean, ADCmean, Cho: Cr ratio, and PET parameters, and FET uptake (TBRmax and TBRmean), were evaluated using Pearson product-moment correlation coefficient. Sensitivity, specificity, and accuracy of individual parameters for recurrence detection were also evaluated using independent samples Student's t-test (two-tailed) and receiver operating characteristic (ROC) analysis using binary logistic regression. Decision thresholds were considered optimal when the sum of paired values for sensitivity and specificity reached the maximum. In addition, multivariate analysis was performed using multivariate analysis of variance (MANOVA) to decide whether a model considering a combination of parameters was better than any single parameter. Further, the combination of MRI and PET/MRI parameters was evaluated for diagnostic accuracy. All statistical analyses were performed using the Statistical Package for the Social Sciences (version 18.0, SPSS for Windows, 2009).
Thirty-two patients enrolled in this study yielded 32 lesions for evaluation. Appropriate neurosurgical resection by an experienced neurosurgeon was performed in 12 patients, with histopathology showing viable tumor tissue in all. Eight patients were classified as radiation necrosis based on their stable clinical state for extended follow-up periods, absence of recent-onset neurologic symptoms, and no significant enlargement of the lesion observed on follow-up MRI images up to a mean period of 12 ± 2 months. Example of one such case is demonstrated in [Figure 1]. Twelve patients were classified as having a recurrent tumor based on the clinical deterioration of neurologic symptoms and/or progressive increase in size of the lesion on CE MRI or PET/MRI during the follow-up period [Table 1] and [Figure 2].
In evaluation of correlations between FET uptake and MRI parameters, significant moderate correlations were observed between TBRmax and N rCBVmean (ρ =0.503; P = 0.003) [Figure 3]a; and, TBRmean and N rCBVmean (ρ =0.414; P = 0.018). In addition, significant moderate negative correlations were observed between TBRmax and ADCmean (ρ = −0.52; P = 0.002) [Figure 3]b and TBRmean and ADCmean (ρ = −0.518; P = 0.002). No correlation was noted among TBR and Cho: Cr ratio values.
Significant difference was observed between the recurrent vs. radiation necrosis group using all four parameters [Figure 3]c: (a) TBRmax 3.97 ± 1.21 vs. 2.02 ± 0.97, P = 0.001; (b) TBRmean2.54 ± 1.09 vs. 1.24 ± 0.61, P = 0.000; (c) Cho: Cr ratio, 2.72 ± 1.01 vs. 1.02 ± 1.00, P = 0.000; (d) N rCBVmean, 3.24 ± 2.05 vs. 1.51 ± 1.17, P = 0.03; (e) ADCmean1149.57 ± 284.92 (10−6 mm 2/s) vs. 1642.31 ± 519.21 (10−6 mm 2/s), P = 0.03.
The diagnostic accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (area under the curve; AUC) of the evaluated individual parameters for the identification of tumor recurrence are demonstrated in [Table 3]. The corresponding ROC curves for these parameters are shown in [Figure 3]d.
MANOVA was carried out for all of the above mentioned parameters in order to evaluate if a combination of these parameters demonstrated better evaluation of the recurrence vs. necrosis diagnostic dilemma than any single parameter statistically. There was a statistically significant difference between the two groups, Hotelling's Trace = 0.957, F (8, 24) =3.282, P = 0.014; partial η2= 0.489.
The diagnostic accuracy, sensitivity, and specificity for recurrence detection using all three MRI parameters were 93.75%, 96%, and 85.7%, respectively. Addition of FET PET TBR values improved these values further to 96.87%, 100%, and 85.7%, respectively.
The current benchmark to distinguish true recurrence from RT necrosis remains histopathology, which is not only invasive but is also marred by complications such as infection, hematoma formation, and neurologic sequel. Hence, it has become an imminent necessity to develop novel noninvasive imaging techniques with high accuracy to reliably perform this distinction. With the development of integrated simultaneous PET/MRI, we aimed at assessing this clinical predicament using multimodal image analysis in a prospective design.
The results obtained using individual PET and MRI parameters were found to be in agreement with published literature.
Dynamic susceptibility-weighted CE MRI (DSC) allows one to quantify relative cerebral blood volume reflecting the vascular environment surrounding a tumor. We observed moderate positive correlations between the FET uptake and N rCBV mean that suggested the presence of coupled vascularity and tumor amino acid uptake with mitotic activity and endothelial proliferation. Similar results have been demonstrated using 11 C-Methionine. However, we observed that both these parameters showed inadequate spatial congruence, as also demonstrated earlier by Filss et al. We found N rCBVmean above cut-off value of 1.78 to be statistically significant for tumor recurrence, although with a lower specificity. Other studies have shown 0.71–1.75, as cut-off values to reliably distinguish radiation necrosis from tumor recurrence with variability owing to variation of techniques employed for calculation.
Similarly, we observed moderate negative correlations between TBR values and ADCmean suggesting increased FET uptake in areas of high mitotic potential and consequently increased cellular density yielding lower ADC values. Though this has been proven directly using stereotactic biopsy, several researchers have reported poor correlation  citing factors such as tissue compression and ischemia contributing to restricted diffusion in GBM. Although significant correlations observed are a departure from published literature, they may be attributable to simultaneous acquisition and accurate ROI copied from areas of enhancement. The use of contrast-enhanced SWI as a guide to identify regions for ADC measurements has been reported to enhance the ability of DWI to differentiate tumor recurrence from PsP. Statistically significant difference of ADC values was noted between the recurrence and RT necrosis groups, similar to the result reported retrospectively by Hein et al., and prospectively by Asao et al.
Recurrent brain tumors are known to demonstrate elevated Cho with decreased N-acetyaspartate (NAA) in comparison to RT necrotic areas with a reported diagnostic yield of 96.2%,, which is similar to our results. However, no significant positive correlation was observed between Cho: Cr ratios and TBR, suggesting that they constituted independent parameters evaluating different aspects of tumor biology. In addition, we noted that FET images were useful in guiding MRS voxel selection.
Furthermore, several publications have reported a need for a multimodal analysis combining FET PET and different advanced MRI techniques , such as MRS with DWI., However, the use of FET alone, or a fusion based or simultaneously integrated FET PET/MRI evaluation has remained limited to the assessment of cerebral metastasis. In this regard, in our analysis using Hotelling's Trace MANOVA, we observed a significant multivariate effect for the combined use of all four parameters in distinguishing recurrence from RT necrosis. Further, improved accuracy and sensitivity was noted with addition of FET PET findings to advanced MRI parameters clearly demonstrating the usefulness of a statistical analysis of this combinatorial analysis.
Apart from discerning the role of simultaneous FET PET/MRI in recurrence vs. radiation necrosis, the authors were also in complete agreement regarding the adequacy of both PET and MRI image quality obtained in an acquisition time of approximately 25 minutes, which is considerably shorter than usually required for a separate acquisition and image fusion.
One should exercise caution in interpreting these results owing to our small sample size and the inability to achieve histological confirmation of disease recurrence in 12 patients, although 50% of these showed rapid neurological deterioration and the remaining demonstrated increasing lesion size on follow- up evaluations [Figure 2]. It has been extensively reported that ePD is more likely to be associated with symptoms than PsP. Tumor factors such as O(6)-methylguanine-DNA methyltransferase (MGMT) methylation status, known to be more frequently associated with the development of PsP, were not evaluated and remain a subject of future evaluations.
Simultaneous PET/MRI evaluation for recurrent glioma is feasible and useful with improvement in diagnostic accuracy. Synergetic effect of multiple MR parameters evaluated together in addition to FET PET uptake highlights the fact that integrated18F-FET PET/MRI might have the potential to impact management of patients with a glioma by timely and conclusive recognition of true recurrence from radiation necrosis.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2], [Figure 3]
[Table 1], [Table 2], [Table 3]