https://orcid.org/0000-0002-2485-1796
Abstract
Brain tumor diagnostics have significantly evolved with the use of positron emission tomography (PET) and advanced magnetic resonance imaging (MRI) techniques. In addition to anatomical MRI, these modalities may provide valuable information for several clinical applications such as differential diagnosis, delineation of tumor extent, prognostication, differentiation between tumor relapse and treatment-related changes, and the evaluation of response to anticancer therapy. In particular, joint recommendations of the Response Assessment in Neuro-Oncology (RANO) Group, the European Association of Neuro-oncology, and major European and American Nuclear Medicine societies highlighted that the additional clinical value of radiolabeled amino acids compared to anatomical MRI alone is outstanding and that its widespread clinical use should be supported. For advanced MRI and its steadily increasing use in clinical practice, the Standardization Subcommittee of the Jumpstarting Brain Tumor Drug Development Coalition provided more recently an updated acquisition protocol for the widely used dynamic susceptibility contrast perfusion MRI. Besides amino acid PET and perfusion MRI, other PET tracers and advanced MRI techniques (e.g. MR spectroscopy) are of considerable clinical interest and are increasingly integrated into everyday clinical practice. Nevertheless, these modalities have shortcomings which should be considered in clinical routine. This comprehensive review provides an overview of potential challenges, limitations, and pitfalls associated with PET imaging and advanced MRI techniques in patients with gliomas or brain metastases. Despite these issues, PET imaging and advanced MRI techniques continue to play an indispensable role in brain tumor management. Acknowledging and mitigating these challenges through interdisciplinary collaboration, standardized protocols, and continuous innovation will further enhance the utility of these modalities in guiding optimal patient care.
In both clinical neuro-oncology and clinical trials, anatomical magnetic resonance imaging (MRI) is currently the investigation of choice for diagnosis and follow-up of patients with a brain tumor. It is exceptional in providing detailed structural information on CNS anatomy and brain neoplasms. Nevertheless, the capacity of anatomical MRI to identify non-enhancing tumors or differentiate neoplastic tissue from nonspecific treatment-related changes is limited and is related to its low specificity for neoplastic tissue.1–5
Frequently used advanced MR techniques in neuro-oncology such as perfusion-weighted imaging (PWI) techniques including dynamic contrast-enhanced (DCE) and dynamic susceptibility contrast (DSC) PWI and arterial spin labeling (ASL), diffusion-weighted imaging (DWI) including diffusion tensor imaging (DTI) tractography, functional MRI (fMRI), and proton MR spectroscopy (MRS)6–9 yield additional information regarding tumor biology, especially at the molecular, physiological, and functional level. Of these imaging techniques, a variety of imaging parameters can be derived which are of considerable clinical value for the management of patients with a brain tumor (eg, prediction of tumor genotype, differential diagnosis, response assessment, and estimation of outcome).10
In addition, positron emission tomography (PET) using the most established radiolabeled amino acids such as [11C]-methyl-l-methionine (MET), O-(2-[18f] fluoroethyl)- l-tyrosine (FET), 3,4-dihydroxy-6-[18f]-fluoro-l-phenylalanine (FDOPA) and the more recently established synthetic amino acid analog anti-1-amino-3-[18f] fluorocyclobutane-1-carboxylic acid (Fluciclovine) in combination with anatomical MRI have shown a great potential for more accurate brain tumor diagnostics.11 The advantage of amino acid PET is a relatively tumor-specific tracer uptake even in the absence of blood-brain barrier disruption.
In primary and secondary brain tumors, amino acid PET is able to describe the location, extent, and heterogeneity of the metabolically active tumor, which can be used for improved targeting of diagnostic biopsy as well as planning of resection and radiotherapy by better visualization of tumor margins.12,13 Furthermore, following neuro-oncological therapy (eg, chemoradiation with alkylating agents, immunotherapy with checkpoint inhibitors, and antiangiogenic drugs), amino acid PET has value for distinguishing tumor recurrence from pseudoprogression1,14,15 and tumor response from pseudoresponse in patients undergoing antiangiogenic therapy,16,17 and for identifying metabolic responders which may have a favorable outcome.18,19
Due to the rapidly growing importance and regular use of these techniques in many neuro-oncology centers for patients with gliomas or brain metastases, clinicians are not infrequently confronted with equivocal or even discrepant imaging findings, which are difficult to interpret. The aim of this review is to highlight challenges, limitations, and pitfalls, especially of PET and advanced MRI diagnostic techniques for brain tumors.
Search Strategy
A PubMed search of the published literature with the most relevant search terms “glioma,” “glioblastoma,” “astrocytoma,” “brain metastases,” “PET,” “positron,” “MRI,” “perfusion,” diffusion,” “spectroscopy,” “CEST,” “inflammation,” “ischemia,” “epilepsy,” “treatment-related changes,” and combinations thereof prior to and inclusive of August 2023 was performed. Additionally, articles identified through searches of the authors’ own files were included in the search.
Overview—Advanced MRI Techniques
An overview of the most frequently used advanced MRI techniques is provided in the Supplementary Material.
Challenges and Limitations—Specific Considerations for Advanced MRI
Challenges in Image Acquisition, Processing, and Interpretation
Perfusion MRI.—
For perfusion MRI, there are many points of potential variability, affecting the ability to implement this modality across imaging sites, including pulse sequence parameter selection, details of gadolinium-based contrast agent (GBCA) injection, preprocessing, processing, and post-processing algorithms within software analysis programs, and the details of how parameter maps are specifically qualitatively or quantitatively interpreted (eg, median or histogram analysis, either within a “hotspot” or total tumor region of interest).20 Specific examples that impact DSC accuracy include GBCA injection rate (5 mL/s from the right arm is recommended, with a saline flush to follow the GBCA), use of gradient echo–echo versus spin echo-echo planar imaging (GRE-EPI vs. SE-EPI), and other parameters (eg, choice of repetition and echo time and flip angle).20
In post-processing, normalization (ie, comparing the rCBV values within a specific region or voxel of interest to a reference region) versus standardization (pertains to establish consistent protocols, methodologies, and reference values for measuring and interpreting rCBV) takes into account the inherent variability related to a contrast bolus injection in a particular patient at a particular time, and standardization has been suggested as superior to normalization.21,22 GRE-EPI is currently recommended for the imaging of brain tumors, as it is sensitive to a greater range of blood vessel sizes including the larger, more disorganized vessels frequently seen in higher-grade gliomas, and less sensitive to the diffusion component of perfusion.20,23
Another issue is the effect of contrast agent leakage on accuracy. A primary assumption in DSC modeling is that the blood-brain barrier is intact and that there is no leakage of GBCA outside of the vasculature. Given that most higher-grade gliomas are contrast-enhancing, which reflects extravascular GBCA leakage into the tumoral interstitium, this assumption is frequently violated. This can cause T1 and T2* shortening in the enhancing tissue, leading to inaccurate CBV estimation and potentially a misjudging of tumor progression versus treatment-related abnormalities. Thus, steps must be taken to mitigate leakage effects in brain tumor DSC perfusion imaging. In addition to software-based leakage correction algorithms, it is also recommended that steps during actual image acquisition are taken to minimize these leakage effects. For example, a preloaded dose of GBCA can saturate the interstitium and create a new baseline of brain tissue signal intensity from which the first-pass GBCA bolus curve can then be more accurately measured.23 With a single dose of GBCA and no preload dose, a lower flip angle in the DSC pulse sequence can minimize T1 leakage effects, although this also decreases signal and is thus preferably used for imaging at 3T or higher.23
DCE appears technically more demanding than DSC, in large part due to a signal change related to the passage of a GBCA bolus which is approximately 10 times less than that with DSC and because of its poorer temporal resolution relative to DSC. It is technically difficult to achieve repetition times short enough to accurately image the first-pass signal versus the time curve of a GBCA bolus with DCE. Therefore, the rate of GBCA wash-in and wash-out and any plateau in between are typically analyzed in DCE, either in model-dependent or -independent ways.24,25 However, pharmacokinetic modeling is more complex in DCE than in DSC. To acquire the full range of output parameter maps, imaging for 6–7 minutes is required (though less for Ktrans alone). For full quantification, additional T1 mapping is needed.
For tumoral blood volume estimates using DSC, care must be taken to not include large cortical vessels, or CBV will be overestimated. An issue leading to underestimation of CBV is a proximal (eg, carotid) arterial stenosis which may lead to GBCA bolus dispersion.
For the estimation of the CBF using ASL, it has to be considered that the signal change detected from labeled blood water molecules is dependent on the blood velocity.26 In addition, the accuracy of CBF estimation is also dependent on an appropriate choice of post-labeling delay, which will vary from patient to patient, including substantially by age.27,28 Relatedly, a proximal arterial (eg, carotid) stenosis may lead to substantial underestimation of CBF with ASL in that vascular territory with the use of a typical post-labeling delay. Furthermore, the signal difference between the labeled and unlabeled states is also small, and so ASL is challenged by possessing lower signal-to-noise ratios than DSC or DCE.
Diffusion MRI.—
Despite its inherently quantitative nature and routine availability of apparent diffusion coefficient (ADC) maps, it is mostly used qualitatively by visual inspection only.29 This may provide challenges to interpreting images where tumor tissue is commonly surrounded by vasogenic edema with high ADC, ie, bright signal, the tumor area may, by contrast, appear dark, even when ADC is not reduced compared to the normal brain parenchyma. This issue can be simply resolved by comparing the measured ADC values within the tumor area with ADC of the normal white matter.
Another issue in using DWI is ignoring the ADC maps altogether and only assessing the averaged DWI images themselves. While there is great value in inspecting these images, as diffusion restriction can be readily appreciated as bright signal, the averaged DWI image is also T2-weighted and thus also appears bright on DWI even if no diffusion restriction is present. The pitfall of this so-called “T2-shine-through” phenomenon can be easily avoided by correlating any DWI-bright area with the ADC map where only the truly diffusion-restricted areas are dark and have lower values than the normal-appearing white matter (Figure 1).
Fifty-year-old female patient with a neuropathologically confirmed IDH wild-type glioblastoma. The lesion is hyperintense on T2-weighted imaging (T2w) and shows rim enhancement on the post-contrast T1-weighed image (CE-T1w). On the diffusion-weighted image (DWI), the lesion appears predominantly hyperintense suggesting diffusion restriction (arrowheads). However, on the apparent diffusion coefficient (ADC) map the lesion is also predominantly hyperintense which is consistent with T2-shine-through rather than diffusion restriction (arrowheads). There are also some small areas where high signal on DWI corresponds with low signal on ADC (arrows) indicating true diffusion restriction.
ADC findings in brain tumors may also be variable and commonly conflicting, which is in large part due to underlying tumor heterogeneity. In higher-grade gliomas, the presence of necrosis, (micro)cystic changes, and vasogenic edema (all with high ADC) affect the expected low ADC in areas with high cellularity. Such effects also occur at the microscopic level, resulting in averaging of ADC values. In addition, other pathologies than tumor can also reduce diffusion, most notably acute ischemia, postictal changes, and intracerebral hemorrhage.
MRS.—
Proton MRS methods have methodical challenges such as the dependence on the magnetic field homogeneity, the low concentration of measurable compounds compared to water signal, and the limited spectral resolution in the field strengths of clinical scanners. In addition, the low signal intensity of metabolite protons implies long measurement times and poor spatial resolution.
DTI tractography and fMRI for presurgical mapping.—
DTI tractography accuracy can be diminished by peritumoral edema and large tumor volumes and when fibers cross, only 1 direction can be detected per voxel. Another disadvantage of DTI tractography in general is that its seeding points are often based on anatomical landmarks and may thus, similar to anatomical MRI, not always accurately reflect functional tissue in individual patients.
Other MRI techniques.—
It has to be pointed out that quantitative MRI techniques are dependent on a high MR field strength. For example, CEST and sodium MR image quality benefits considerably from acquisition using ultrahigh field MRI scanners (7 T or higher),30 which are currently available only in a few centers. On the other hand, there is a growing body of literature suggesting that CEST imaging at 3 T is sufficient enough to provide valuable information for clinical glioma management.31 Regarding sodium MRI, the 23Na signal is still 10 000 times lower than the 1H signal, resulting in a decreased signal-to-noise ratio compared with conventional MRI, which is a major obstacle for routine clinical use.32 Sodium MRI also suffers from partial volume averaging, long scan times, and difficulty with differentiating the tumor from peritumoral compartments.33
Artifacts
Not only anatomical but also advanced MRI techniques such as MRS, fMRI, and DTI techniques, whose image quality can be improved by higher field strengths (>1.5 T), can be affected by artifacts. Artifacts are defined as any image (signal or loss of signal) that has no anatomical basis, but is the result of distorted, additional, or suppressed MR image information.34 Overall, the majority of artifacts in the brain represent blood flow artifacts, cerebrospinal fluid pulsation artifacts, magnetic susceptibility artifacts, and motion artifacts.34
For example, all 3 perfusion MRI techniques can have their measured signal change diminished or destroyed by susceptibility effects which alter the local magnetic field, such as near the skull base (eg, air in the paranasal sinuses and mastoid air cells, and bone; Figure 2), any metallic material from surgery (eg, vascular clips and craniotomy closure plates), or paramagnetic blood products such as hemosiderin which are not uncommon after tumor resection. Out of these 3 techniques, DSC, usually being T2*-weighted, is particularly sensitive to these effects. Conversely, several ASL implementations are SE- rather than GRE-based, thus less sensitive to these susceptibility effects and particularly useful for post-treatment assessment or tumors near the skull base.
A left temporal chemoradiation-treated glioblastoma on post-contrast T1-weighted imaging (arrows in A and B). Relative cerebral blood volume (rCBV) maps through this lesion from dynamic susceptibility contrast (DSC) perfusion-weighted imaging are shown in C and registered with and superimposed on post-contrast T1-weighted imaging in D. However, susceptibility-related signal loss in proximity to the petrous and mastoid temporal bones artifactually makes rCBV in the inferior temporal lobes appear to be zero (arrows in C and D). Since the treated glioblastoma is within this area of artifactual signal loss, it could falsely appear to be hypoperfusing and so it is unable to be evaluated with DSC.
ASL is a subtraction technique, as CBF is estimated by subtracting signal from the water-labeled state relative to the baseline, unlabeled state. This also makes ASL particularly vulnerable to patient motion, and scan times are typically longer (around 4–5 minutes at 1.5T) than with DSC. On the other hand, ASL can easily be repeated after patient motion.
Standardization and Comparability of MRI Acquisition Protocols
The clinical value and potential of advanced MRI techniques in neuro-oncology as discussed above is undisputed. Nevertheless, in addition to the general availability of these techniques, the clinical success, and dissemination particularly depend on a high reproducibility and comparability of the results, which require a greater standardization across image acquisition, reconstruction, post-processing methods, and reporting of the results. This is not only important for advanced MRI techniques such as PWI, DWI, MRS, or CEST, but also for anatomical MRI. Accordingly, general recommendations for a standardized brain tumor imaging protocol were published35 and extended by Kaufmann and colleagues for clinical trials in patients with brain metastases.36 Therein, an “ideal” brain tumor imaging protocol including structural (T1-weighted pre- and post-contrast, T2-weighted, and FLAIR) and advanced MRI (DWI and PWI) as well as a “minimum” protocol omitting PWI are provided also acknowledging different MRI field strengths. Considerable work remains to be done with standardization, as reflected in their recommendations to perform follow-up measurements in the same patients on the same scanner platforms using the same imaging protocols to ensure an accurate evaluation of imaging changes over time.
DWI is part of the abovementioned brain tumor imaging protocols and is widely used in neuro-oncology. In addition, diffusion MRI is the only 1 that is currently included in a neuro-oncological response assessment guideline, ie, for pediatric glioma.37 Nevertheless, since in most cases only a visual evaluation of the ADC maps is performed,29 further efforts for standardization are necessary to fully exploit the potential of quantitative DWI in the future.10 One of the advanced MRI techniques for which standardization has advanced significantly in recent years is DSC PWI. cutoff values for rCBV and CBF for clinical diagnoses of brain tumors are; however, highly variable and can thus far neither be used nor compared between institutions.38,39 The consensus recommendations published by Boxerman et al.23 aid towards greater generalizability of PWI parameters and should be emphasized and followed in future neuro-oncological studies utilizing advanced neuroimaging.
Besides DSC PWI, consensus recommendations have also been published for the use of proton MRS by an expert committee of the International Society of Magnetic Resonance in Medicine MRS study group.40 In this paper, guidance on how to perform MRS at different field strengths and indications for MRS as well as recommendations on post-processing, analysis, and quality assurance are provided.
Standardization for other more advanced MRI techniques such as CEST, sodium MRI, or X-nuclei MRS has yet to be established. Generally, although several international researcher-driven committees put much effort into the standardization of advanced MRI techniques in brain tumor diagnostics, the role of the vendors should not be ignored. Historically, MRI was not aimed at quantification and many sequence implementations are deliberately different between vendors. With the current paradigm shift towards MRI-based imaging (bio)markers, a dedicated effort from the vendors is needed to aid in the standardization of pulse sequence implementation, acquisition protocols, and post-processing. Only then, a true translation of these techniques to obtain quantitative imaging markers in routine clinical use and to apply them in multicenter studies could be achieved.
Diagnostic Performance of Advanced MRI
For the differentiation of tumor relapse from treatment-related changes, a meta-analysis including more than 1700 patients concluded that the diagnostic performance of PWI parameters is predominantly within the range of 80%–90% in terms of pooled sensitivity and specificity.38 On the other hand, the authors reported a pronounced heterogeneity across all studies (n = 28). For example, rCBV threshold values varied considerably (range of maximum rCBV values, 1.5–3.1), indicating that the observed wide span of threshold values may challenge this clinically important differentiation (Figure 3).
Fifty-two-year-old patient with a history of an acute lymphoblastic leukemia was diagnosed more than 35 years ago and treated with methotrexate-based chemotherapy and whole-brain radiotherapy. Subsequently, a secondary and atypical parasagittal meningioma was diagnosed in the right frontal lobe. After resection, the patient had undergone adjuvant proton radiotherapy (radiation dose, 60 Gy). Thirty months after radiotherapy, anatomical MR imaging showed a nodular contrast enhancement at the rim of the resection cavity (sagittal view) with an increased blood volume on 3 T dynamic susceptibility contrast perfusion MRI (relative cerebral blood volume after contrast agent leakage correction, 3.0). At this time, the patient had a Karnofsky performance status of 90% and a mild but worsening left-sided hemiparesis, which required dexamethasone (maximal dose, 8 mg). After resection of the nodular lesion, histology revealed widespread necrotic areas without tumor cells (hematoxylin and eosin staining).
Furthermore, a limitation of 2-HG MRS was highlighted by Suh and colleagues.41 Their study in 82 IDH-wild-type glioblastoma patients suggested that 21% of cases had false-positively increased levels of 2-HG, which was associated with necrosis, highlighting accuracy difficulties of 2-HG MRS. Another limitation of 2-HG MRS is that it measures total 2-HG levels without distinction between its enantiomers (ie, D- and L-2-HG). The contribution of nonspecific elevations of L-2-HG, which has been observed in glioblastomas,42 can obscure the D-2-HG signal that is the actual product of mutant IDH.43
Challenges and Limitations—Specific Considerations for PET
PET Tracers
Glucose PET.—
A common pitfall relates to the high physiological FDG uptake in gray matter brain structures that may reduce the tumor-to-background contrast and detectability despite precise PET to MRI image registration, eg, in brain metastasis around the gray–white matter boundary or in patients with non-enhancing CNS WHO grade 2 or 3 gliomas (Figure 4).
A 41-year-old male with a newly diagnosed brain lesion suspicious of a glioma without contrast enhancement and widespread FLAIR signal abnormalities in the right frontal lobe. In contrast to the hypometabolic FDG PET in this region, FET PET revealed a pathologically increased tracer uptake. Histological tissue evaluation obtained from biopsy confirmed an anaplastic astrocytoma.
Furthermore, interpretation difficulties arise in any condition with high regional FDG uptake above the level of white matter that can mimic active tumor tissue. This includes inflammation of any cause, eg, postsurgical trauma in skin, any infection, granulomatous diseases, hemorrhage (<4 days), radiotherapy-related effects, encephalitis, and, less commonly, regional epileptic seizure activity at the time of FDG injection and (sub)acute stroke (<14 days).44
Amino acid PET tracers.—
For the differentiation of glioma relapse from treatment-related changes using amino acid PET, by far the most frequent indication in clinical routine for this imaging modality, its diagnostic performance is largely comparable to perfusion MRI. For example, for the radiolabeled amino acid FET, several studies reported sensitivities and specificities predominantly around 90% (Supplementary Table 1).
Nevertheless, it should be considered that approximately 30% of gliomas of CNS WHO grade 2 exhibit no amino acid uptake.17 Of note, brain lesions without FET uptake and MRI findings suspicious for non-enhancing gliomas (eg, CNS WHO grade 2 astrocytomas) may even present photopenic defects on FET PET with uptake visually lower than the healthy background uptake but harbor gliomas with higher CNS WHO grades.45 This phenomenon has also been described for the radiolabeled amino acids MET and FDOPA.46
On the other hand, oligodendrogliomas may show disproportionately elevated amino acid uptake compared to other gliomas47 and may thus mimic high tumor aggressiveness. Of note, a more recent study suggested that only around 50% of non-enhancing oligodendrogliomas exhibit increased fluciclovine uptake,48 which is considerably lower than that was observed with the more established radiolabeled amino acids such as FET or MET.47,49
Although increased uptake of the radiolabeled amino acids FET, MET, and FDOPA in non-neoplastic brain lesions is rare, it should be considered in clinical practice.2 The most relevant non-tumoral etiologies with increased amino acid uptake are discussed below.
Other tracers.—
Further PET tracers directed at various target molecules expressed in tumor cells or the tumor microenvironment may also be related to potential pitfalls in interpretation of tracer uptake. For example, the 18 kDa mitochondrial translocator protein (TSPO) is not only expressed by tumor cells, but also in activated microglia and infiltrating macrophages, which leads to increased TSPO tracer uptake in non-tumoral areas such as inflammatory lesions or treatment-related changes.50 The prostate-specific membrane antigen, already used routinely as target for PET imaging and theranostic treatment of prostate cancer and under investigation for CNS tumors, has been described to be expressed primarily in endothelial cells of gliomas. Thus, prostate-specific membrane antigen PET uptake may reflect vascular alterations rather than direct tumor cell activity51 and may be found in rare cases of pseudoprogression.52 Furthermore, binding of PET ligands to somatostatin receptors, usually performed for the delineation of meningiomas and only rarely in glioma patients, may be increased in rare cases of pseudoprogression.53
Physiological Variants
Several anatomical structures may demonstrate mild to moderate physiological uptake of radiolabeled amino acids that should not be interpreted as neoplastic tissue. Increased physiological uptake of FDOPA is present in the basal ganglia since the tracer is a precursor of dopamine synthesis.54 FET shows occasionally increased physiological uptake in the pineal region while increased MET uptake in the pituitary gland is a common finding due to the incorporation of MET in cerebral protein synthesis55 which is increased in this gland. In addition, the pituitary gland shows high physiological uptake following injection of radiolabeled somatostatin receptor ligands, which may serve as a positive control but limit the exact meningioma delineation in its close proximity.56 Since FET has a delayed elimination after injection and therefore remains longer in the blood pool, an increased signal in large venous vessels (eg, sagittal, transversus, and cavernosus sinus) and in vascular malformations (eg, arterio-venous malformations, angiomas, cavernomas, and developmental venous anomalies) can be observed.
Technical Aspects
Scanner resolution.—
It should be considered that the spatial resolution of PET is in the range of 3–5 mm (full width at half maximum) and is hence inferior to that of structural MRI with a spatial resolution of 1 mm or less. Nevertheless, PET is not primarily designed to provide detailed anatomical information, but rather information about the metabolism or molecular characteristics. Therefore, the combination of PET and MRI provides both anatomical and metabolic information, leveraging the strengths of both modalities. Of note, PET may even detect lesions below the scanner resolution providing that the lesion-to-background ratio is sufficiently high.
Image reconstruction.—
Analytical image reconstruction techniques such as filtered back projection directly reconstruct the images from the acquired data by a mathematical process called back projection. Although computationally fast, analytical image reconstruction techniques have been widely replaced in PET by iterative image reconstruction techniques such as the ordered subset expectation maximization which are computationally more demanding but offer several advantages. Iterative reconstruction techniques are known to generally improve the quality of the reconstructed image compared with analytical reconstruction techniques like filtered back projection, especially in low-count areas.57 Specifically, the noise level is reduced by iterative reconstruction techniques resulting in an improved signal-to-noise ratio and, hence, an improved lesion-to-background contrast. Furthermore, a lower noise level, which means lower variations in the tracer uptake values in individual voxels, allows for a more robust assessment of parameters that are based on a few voxels such as the maximum tumor-to-brain ratio, potentially improving the comparability between different scanners and institutions.
Standardization and Comparability of PET Acquisition Protocols
The determination of PET parameters may be influenced by various factors such as different correction methods (eg, scatter or attenuation), image reconstruction parameters, and positioning of regions of interest in the tumor and reference region. A bicentric study suggested that different PET reconstruction parameters affect the comparability of maximum tumor-to-brain ratios and tumor volumes obtained from FET PET.58 Another study demonstrated the influence of a different segmentation of the background region on the reproducibility and comparability of FET PET tumor-to-brain ratios.59
Consequently, guidelines to ensure comparability and reproducibility of PET study results and to assist nuclear medicine practitioners in terms of performing, interpreting, and reporting PET results in glioma patients using glucose and amino acid PET were established for the first time in 2006.60 In 2016, the PET task force of the Response Assessment in Neuro-Oncology (RANO) Working Group and the European Association of Neuro-oncology (EANO) provided evidence-based recommendations for the use of PET imaging in patients with glioma.12 The most recent procedural guidelines for adults were published in 201961 and represented a collaborative effort of the EANO, the European Association of Nuclear Medicine, the Society for Nuclear Medicine and Molecular Imaging, and the PET/RANO working group, with a similar effort for pediatric gliomas in 2022.62 Besides definitions, clinical indications, examination procedures and general precautions, the administered activity for patients for glucose and amino acid PET tracers is specified, which eliminates one source of challenge to reproducibility. In addition, details and recommendations on the attenuation correction methods, also in the setting of PET/MR imaging, are provided. These guidelines also aid in standardizing data acquisition and interpretation and provide suggestions for how details of the PET scans and image interpretation should be reported. Even though these detailed recommendations exist and are continually being expanded and improved, different ways of evaluating data are still possible, so comparability may not be assured despite adherence to the guidelines. And while there are guidelines, they are not necessarily always followed. To further standardize and harmonize data evaluation and image noise characteristics across sites, methods from the field of artificial intelligence are likely to become increasingly important in the future. Initial studies in adult and pediatric populations have shown that an automated and thus standardized detection and segmentation of brain tumors based on FET PET is possible.63,64 Even though studies with larger numbers of cases involving different centers are necessary, these methods already hold great potential for a better standardization of the evaluation of PET in brain tumor patients prior to a possible translation into clinical routine.
Reimbursement and Availability Issues of Amino Acid PET
US perspective.—
At present, no radiolabeled amino acids have approval from the FDA to be marketed for use in diagnostic imaging of brain tumors. Accordingly, amino acid PET is largely limited in the US to use in clinical research at major academic institutions. A small handful of institutions currently offer amino acid PET for routine clinical diagnosis, usually when funding has been obtained to cover the cost of the radiotracer and PET imaging. Only one US institution is known by the authors to utilize amino acid PET in routine clinical care (via an Expanded Access Investigational New Drug approval) with recovery of procedure costs, but not radiopharmaceutical costs, through patient insurance (Indiana University).
With regard to reimbursement, there is no published information regarding coverage of PET imaging in the United States. Coverage is known to vary greatly between insurance companies without an obvious underlying rationale or pattern. Even in one geographic region, an insurance company may cover certain exams that are not covered by the same company in another region of the country. For the most part, amino acid radiotracers for brain tumor imaging are not covered routinely by US insurance companies.
A second major consideration US medical practices have to overcome is the process by which a PET/CT or PET/MRI exam can be approved through insurance once it can be considered for coverage. Insurance authorization prior to receiving the exam (so-called pre-authorization) is often required by private insurance companies before a patient can be considered for a brain PET/CT or PET/MRI exam. This has many challenges that are almost universally detrimental to the patient, the ordering physician, and the healthcare system. Unfortunately, the process of pre-authorization is becoming more and more frequent.
From a national standpoint, a more centralized electronic pre-approval system would maximize efficiency across insurance providers. It would also permit better transparency and accountability by insurance companies when they decide not to cover an imaging exam recommended by treating physicians for advancement of their patients’ care. PET/CT and PET/MR imaging with an emphasis on amino acid PET imaging would very likely benefit from such an arrangement.
European perspective.—
The tracer FDOPA (product name DOPAVIEW) is approved for the detection of suspicious recurrence or residual disease in primary brain tumors of all grades in the following Member States of the European Economic Area: Austria, Belgium, France, Germany, Italy, Luxemburg, Netherlands, Portugal, Spain, and United Kingdom.
For FET, approval as medical drug for gliomas exists in Switzerland and in France (product name IASOglio). Reimbursement for FET PET is available in Switzerland and in France. Clinical use of FET and MET is possible in Germany under the Regulation on Radioactive Drugs or Drugs Treated with Ionizing Radiation, ie, in centers with a cyclotron unit and a manufacturing license for the tracer.65 In December 2021, the German Federal Joint Committee approved reimbursement for PET or PET/CT with radiolabeled amino acids in gliomas to differentiate treatment-related changes from tumor relapse as an amendment to the guideline on outpatient specialist care.66 Thus, a reimbursement of MET, FDOPA, and FET by statutory health insurance is now granted. General approval of FET as a medical drug by the German Federal Institute for Drugs and Medical Devices is not yet available. In Denmark, MET, FDOPA, and FET PET scanning may be used clinically for all common indications in neuro-oncology on a compassionate use basis and is reimbursed by the public health system.
Overall, the possibility of reimbursement of amino acid PET in brain tumors by statutory health insurance is not yet uniformly regulated in Europe and varies from country to country. Studies proving a clinical benefit for patients regarding therapeutic management, outcome, and health economics (eg, by saving additional diagnostics or tailoring therapy) may support reimbursement endeavors.
Clinically Relevant Diagnostic Pitfalls—Considerations for Amino Acid PET and Advanced MRI
Inflammation and Infection
A few case reports observed that increased FET uptake in brain lesions may be caused by an inflammatory process. These lesions can be caused by several factors, including autoimmune disorders (eg, demyelinating diseases such as multiple sclerosis and primary CNS vasculitis),67,68 granulomatous diseases (eg, neurosarcoidosis),69 and pathogen-related infections (eg, brain abscesses, progressive multifocal leukoencephalopathy, toxoplasmosis, and bacterial meningoencephalitis).67,70,71 Similar results have been reported for MET and FDOPA.72–77 In general, the uptake in inflammatory lesions appears to be mild to moderate and can be reliably differentiated from higher-grade brain tumors using lesion-to-brain ratios.67,74,78
Although advanced MRI metrics are of clinical value to facilitate the diagnosis of tumefactive (atypical demyelinating) lesions, a case series suggested that these lesions may also show perfusion parameters that were more typical for glial tumors (ie, increased rCBV on PWI).79 Another report described advanced MR imaging findings atypical for an infectious process, ie, absence of a centrally reduced diffusion sign,80 no dual rim sign, and increased rCBV, combined with ring enhancement on anatomical MRI suggesting a higher grade glioma or a brain metastasis. Nevertheless, histological work-up after resection revealed a pyogenic Streptococcus milleri brain abscess.81 Vice versa, atypical advanced MRI findings (eg, no increased rCBV as assessed by PWI, normal choline/n-acetyl-aspartate ratio on MRS) were also observed in patients with glioblastoma, initially misdiagnosed and treated for autoimmune encephalitis.82
Cerebral Infarction
Mild to moderate MET or FET uptake in cerebrovascular diseases such as stroke may be observed in or surrounding the lesion in the subacute or chronic stage (Figure 5). For example, Nakagawa et al. reported that the average MET uptake in cerebral infarctions as assessed by lesion-to-brain ratios was 1.07 ± 0.28 (range, 0.67–1.31).83 Of note, FET uptake normalized within the first days in the subacute phase,67,84 whereas MET uptake seems to persist over several weeks.85 An initial study suggested that FDOPA also accumulates in ischemic brain lesions.75
Early postoperative MRI of a glioblastoma patient after 48 hours, and MR and FET PET imaging 3 and 9 weeks after surgery showing the temporal evolution of a perioperative ischemic stroke involving the left middle cerebral artery territory. Neuroimaging 3 weeks after surgery shows a gyral pattern of parenchymal contrast enhancement in the left insula and temporoparietal region (laterally placed arrows) consistent with postischemic hyperperfusion following vessel recanalization with vasodilation, inflammation, and blood-brain barrier injury leading to leakage of contrast material and FET (maximum tumor-to-brain ratio, 2.5) into the extravascular space obscuring any binding to glioma tissue. The left striatocapsular infarct with a low apparent diffusion coefficient (ADC) does not show increased FET uptake (arrows in the midline) as tracer delivery is perfusion-dependent. Follow-up MRI and FET PET at 9 weeks show fading of gyral contrast enhancement and associated FET uptake (maximum tumor-to-brain ratio, 1.8).
Similarly, patients with subacute ischemic stroke may display contrast enhancement, and the presence of luxury perfusion can cause hyperperfusion on PWI, mimicking brain neoplasms. DWI is useful in defining ischemic lesions in the acute phase but might not show restricted diffusion in subacute phases of infarction.86
Intracerebral hemorrhage.—
Spontaneously occurring non-neoplastic intracerebral hemorrhage may result in a slightly increased amino acid uptake in the periphery of the lesion.87 Especially in the subacute stage with hematoma manifestation, the average lesion-to-background uptake for MET was only 1.12 ± 0.04 (range, 1.10–1.17).83 Similarly, in a study evaluating a rat model, increased MET and FET uptake in the periphery of intracerebral hematomas persisted in most cases not longer than 14 days.88 In contrast, after the subacute phase (>14 days), neoplastic hematomas may show a higher MET uptake and extend beyond the contrast-enhancing areas compared to non-neoplastic hematomas.89 A more recent study reported that the tracer FDOPA also accumulates in intracerebral hemorrhages.75
Equivocal Imaging Changes after Therapeutic Interventions
A more recent study reported on transient and significantly increased FET uptake after glioma resection in around 25% of patients.90 These postoperative reactive changes were located adjacent to the resection cavity and occurred within the first 2 weeks after surgery. For the evaluation of residual tumors early after surgery, this “flare” phenomenon should be considered.
Despite the fact that numerous studies have reported that amino acid PET allows a highly reliable differentiation of tumor relapse from nonspecific treatment-related changes in patients with gliomas or brain metastases,91,92 false-positive uptake in irradiated areas may occur in rare cases, particularly when high cumulative radiation doses (>120 Gy) were applied (Figure 6). A false-positive uptake of radiolabeled amino acids in the irradiated region is most probably related to strong reactive astrogliosis.93
Forty-eight-year-old patient with a CNS WHO grade 2 astrocytoma who was heavily pretreated by several radiotherapies with a cumulative dose of more than 160 Gy. At suspected relapse, the corresponding FET PET showed pathologically increased metabolic activity (ie, maximum tumor-to-brain ratio, 3.1) and was consistent with tumor progression. In contrast, histological evaluation revealed a coagulation necrosis (right) accompanied by necrosis of blood vessels with hyalinization of the vessel wall remnants, and reactive astrogliosis at the margin of the necrosis (left; hematoxylin and eosin staining; courtesy of Prof. Joachim Weis, Institute of Neuropathology, RWTH Aachen University Hospital, Aachen, Germany).
The use of alkylating chemotherapy (eg, temozolomide) for glioma patients may affect the normal-appearing brain tissue surrounding the tumor. An initial study suggested that FDOPA uptake is significantly reduced in the background region of patients undergoing temozolomide compared to patients who discontinued chemotherapy, potentially affecting the calculation of tumor-to-brain ratios.94 Nevertheless, data on intraindividual uptake changes are lacking.
Following vascular endothelial growth receptor blockade using bevacizumab or application of anticancer agents with antiangiogenic properties (eg, regorafenib) a pronounced diffusion restriction on DWI in combination with a decreased amino acid uptake may occur.95,96
In patients with radionecrosis, the evaluation of DWI should be interpreted with caution as study results on this topic are contradictory and DWI metrics may reflect opposite phenomena (ie, diffusion restriction due to hypercellularity or coagulative necrosis after radiotherapy).97 For the differentiation of radionecrosis from tumor relapse using PWI it should be considered that an accurate rCBV evaluation may be impeded by tumor heterogeneity, the co-existence of neoplastic and radionecrotic tissue, and susceptibility artifacts due to petechial hemorrhage caused by the irradiation or melanin from malignant melanoma.98
Epilepsy
A case series reported on patients with transient and strictly cortical FET uptake following epileptic seizures over several weeks, potentially mimicking brain tumors.99 Similar observations were reported also for MET.100,101 Conversely, a subsequent study performed in both animal epilepsy models and patients with focal epilepsy revealed no evidence for increased longer-lasting FET uptake in the postictal state, suggesting that epileptic seizures are unlikely to represent a major pitfall in brain tumor diagnostics using FET PET.102
In a subset of patients with gliomas suffering from epileptic seizures, diffusion restriction with low ADC values and transient increased cortical perfusion in DSC-PWI, especially within the first weeks after seizure onset, were reported.99 Interestingly, the tumors showed no signs of progression.
Combined use of Amino Acid PET and Advanced MRI to Increase Diagnostic Performance
In recent years, several studies highlighted an increase in diagnostic performance by combining amino acid PET parameters with advanced MRI metrics, particularly for differentiating tumor relapse from treatment-related changes. A more recent review published in 2023 summarized the added value of amino acid PET and advanced MRI combined for various indications in patients with gliomas,103 and selected studies are presented in Supplementary Table 2.
Outlook and Perspectives
The risk of misinterpretation in case of discrepant findings or imaging variants between PET and MRI, as well as diagnostic pitfalls, can be reduced with increased experience over time. Attention to technical guidelines for imaging acquisition and interpretation provides the basis for comparing findings between institutions and across different disease entities. Intense and vivid exchange of information between imaging and clinical experts will help to detect and unravel indistinct findings which need to be reported to further train and teach the community.
In case of diagnostic uncertainty, biopsies should be performed,104 and unexpected findings thereof be published. Consortial studies or at least multiinstitutional imaging and data repositories may help to speed up the learning process. Currently, radiomics and artificial intelligence applications such as machine or deep learning may help to better detect, discriminate, and interpret complex data patterns of certain imaging phenotypes.105 It is tempting to speculate that this will reduce diagnostic pitfalls in the future.
Supplementary material
Supplementary material is available online at Neuro-Oncology (https://dbpia.nl.go.kr/neuro-oncology).
Funding
None.
Conflicts of interest statement
N.G.: Honoraria for lectures from Blue Earth Diagnostics and for advisory board participation from Telix Pharmaceuticals. T.J.K.: Reported no potential conflicts of interest. P.V.: Reported no potential conflicts of interest. P.L.: Honoraria for lectures from Blue Earth Diagnostics. M.S.: Honoraria (paid to institution) for lectures from AuntMinnie and Fondazione Internazionale Menarini; honoraria (paid to institution) for consultancy from Bracco. M.C.V.: Honoraria for consulting, lectures, or advisory board participation from Telix Pharmaceuticals. Academic-Industry Partnership funding from Telix. Research grant from Blue Earth Diagnostics. K.L.J.: Honoraria for consulting from Telix Pharmaceuticals. R.R.: Reported no potential conflicts of interest. N.L.A.: Honoraria for consulting or advisory board participation from Novartis/Advanced Accelerator Applications, Telix Pharmaceuticals, and Servier, and research grants from Novocure. E.H.: Reported no potential conflicts of interest. I.L.: Reported no potential conflicts of interest. M.H.: Reported no potential conflicts of interest. R.S.: Honoraria for lectures from Servier and for advisory boards from Seagen and Astra Zeneca. M.A.V.: Clinical trial funding from DeNovo, Chimerix, Oncosynergy, Infuseon, and honorarium from Servier. P.Y.W.: Research support from Astra Zeneca, Black Diamond, Bristol Meyers Squibb, Celgene, Chimerix, Eli Lily, Erasca, Genentech/Roche, Kazia, Medicinova, Merck, Novartis, Nuvation Bio, Servier, Vascular Biogenics, and VBI Vaccines. Honoraria for consulting or advisory board participation from Astra Zeneca, Black Diamond, Celularity, Chimerix, Day One Bio, Genenta, Glaxo Smith Kline, Insightec, Kintara, Merck, Mundipharma, Novartis, Novocure, Prelude Therapeutics, Sapience, Servier, Sagimet, Vascular Biogenics, and VBI Vaccines. M.W.: Research grants from Quercis and Versameb, and honoraria for lectures or advisory board participation or consulting from Bayer, Curevac, Medac, Neurosense, Novartis, Novocure, Orbus, Philogen, Roche, and Servier. J.C.T.: Honoraria for lectures or consulting from Seagen and Novartis, and research grants from Novocure and Munich Surgical Imaging.
Authorship statement
Conceptualization: N.G. and J.C.T. Writing of manuscript drafts: all. Revising manuscript: all. Preparation of figures: T.J.K., R.R., M.S., I.L., K.J.L., and N.G. Approving final content of manuscript: all.
