Introduction

The neuroradiological diagnosis of brain tumors has advanced progressively within the 4 decades since the emergence of CT and MRI. In addition to these conventional morphological neuroimaging methods, metabolic analysis of brain tumors with PET, which has been widespread in the clinical setting during the last decade, has made it feasible to obtain more characteristic features of each brain tumor for diagnosis [1,2,3,4]. Among PET tracers used for the diagnosis of brain tumors, 11C-methionine (MET) is one of the most popular amino acid tracers as well as O-(2-[18F]-fluoroethyl)-L-tyrosine(FET) and 3,4-dihydroxy-6-[18F]-fluoro-L-phenylalanine(FDOPA) [5,6,7,8,9]. Although the use of MET-PET remains limited in routine clinical work-ups because of the short half-life (20 min) of MET, its usefulness for brain tumor diagnosis is undoubted according to previous clinical studies and is widely recognized [5,6,7,8,9,10,11,12,13]. The response assessment in neuro-oncology (RANO) working group and the European association for neuro-oncology (EANO) recommendations for the clinical use of PET imaging for glioma indicate that MRI is the standard to monitor both treatment and response, but PET can provide additional insight beyond MRI into the biology and treatment response [9]. MET-PET can give information on malignancy and extent of tumors in the brain. That is, MET uptake is known to represent the biological activity of brain tumors, and also the microstructural tumor infiltration and tumor vasculature affecting MET dynamics. Such information about the dynamics of MET uptake can be easily obtained with dynamic PET studies, and may provide more specific profiles of each brain tumor, which cannot be obtained from only static PET studies, to aid in making a more precise differential diagnosis.

In static MET-PET, the uptake is higher in gliomas of WHO grades III/IV compared with WHO grade II gliomas; however, there are limitations due to overlap between WHO grades as well as histological subtypes [9].

In 2012, Aki et al. [14], reported the usefulness of dynamic MET-PET with a 2-dimensional mode (2D) in the differential diagnosis of brain tumors as an aid to the static MET-PET examination. They showed that a significant dynamic decrease in the maximum MET standardized uptake value (SUV) tumor/normal frontal lobe ratio occurred with meningiomas and oligodendrocytic tumors, whereas a significant dynamic increase occurred with glioblastomas and malignant lymphomas. In such a study using 2D-PET, however, values of MET uptake in tissues in only three consecutive phases can be obtained, and more detailed information on MET uptake values over time is limited. More consecutive MET uptake values may reveal the histological background of brain tumors affecting MET dynamics, which may aid in making a more precise differential diagnosis. Compared to 2D-PET, PET scanning with a 3-dimensional mode (3D-PET) has potential for higher image resolution over time, and can yield a time-activity curve (TAC) of MET based on the MET uptake values in tissues measured at 14 time frames using a list mode data collection system. Several FET 3D-PET studies of brain tumors and radiation necrosis have been published [15,16,17,18,19], but few of MET 3D-PET. Moulin-Romsee et al. reported that dynamic MET-PET does not allow WHO grading of glioma [20].

The aim of this study was to assess whether MET-TAC obtained from a MET-PET dynamic study with 3D-PET could aid in differential diagnosis of brain tumors, and in understanding the histological background related to tumor biological activity and vasculature in each common brain tumor affecting MET dynamics.

Patients and methods

Patient population

One hundred sixty patients with newly diagnosed brain tumors (8 patients, diffuse astrocytomas [DA]; 27, oligodendrogliomas [OD]; 21, oligoastrocytomas [OA]; 23, anaplastic astrocytomas [AA]; 10, anaplastic oligodendrogliomas [AO]; 19, anaplastic oligoastrocytomas [AOA]; 31, glioblastoma multiforme [GBM]; 9, meningiomas; 4, hemangioblastomas; and 8, primary central nervous system lymphomas [PCNSL]) who underwent MET-PET following the acquisition of written informed consent in the Chubu Medical Center for Prolonged Traumatic Brain Dysfunction, Kizawa Memorial Hospital, between March 2012 and October 2015, were enrolled in this retrospective study. No patient received prior surgery, and diagnoses were made from tumor specimens obtained during surgery or biopsy. Tumors were histologically classified according to the fourth edition of the World Health Organization (WHO) classification of tumors of the central nervous system, published in 2007. Patient profiles are summarized in Table 1.

Table 1 Profiles of 160 patients with brain tumors
Full size table

PET imaging acquisition

The PET scanner used in this study was the Eminence STARGATE (Shimadzu Corporation, Kyoto, Japan) equipped with a 3D acquisition system, which provides 35 transaxial images at 2.65-mm intervals. The in-place spatial resolution (full width on transaxial images) was 4.8 mm. Participants were placed in the PET scanner so that axial slices were parallel to the canthomeatal line. Immobility was checked by alignment of three laser beams, with lines drawn on the participants’ faces. A 4-min transmission scan was performed, and the radiopharmaceutical (MET: 3.5 MBq/kg) was injected intravenously through the cubital vein. With the dynamic MET-PET studies, we started a 35-min emission scan in the 3D mode at the time of administration. The emission recording consisted of 14 time frames (time frames 1–5, 1 min; 6–8, 2 min; 9–11, 3 min; 12–14, 5 min) covering the period up to 50 min after injection. Static scans were reconstructed with attenuation correction with data from the transmission scan, dead time, random, and scatter. Regional MET uptake was expressed as an SUV (tissue activity/ml)/[injected radioisotope activity/body weight (g)] and the dynamic row-action maximum likelihood algorithm (DRAMA) reconstruction method was applied.

Image analysis

PET scans were co-registered to MRI for better anatomical determination. Image fusion was performed with an image analysis program in combination with Dr. View/Linux image analysis software (INFOCOM CORPORATION, Tokyo, Japan) and a method described by Kapouleas et al. [21] First, the 90% iso-contour threshold ROIs were defined on the summation images for the interval 5–35 min after the tracer injection and applied to the dynamic images. ROIs of the normal confluence, internal jugular vein, and normal frontal cortex in the 139 patients with gliomas were traced at the same time.

Second, the maximum tumor MET-SUV was measured consecutively to construct the TAC of the normal confluence, internal jugular vein, normal frontal cortex and brain tumors.

To classify the characteristics of the MET-TAC pattern of each brain tumor, the MET-TAC was divided into three phases: initial phase, one minute after MET administration; early phase, from one to ten minutes after MET administration, in which a rapid change of MET-SUV can be seen; and late phase, 10 min after MET administration, in which MET-SUV change is slow. Time-to-peak (TTP; time in minutes from the beginning of the dynamic acquisition up to the maximum SUV of the lesion) was determined. The slope of the curve was determined by a regression coefficient (SLOPE; h−1) in early and late phases [19].

Statistical analysis

To compare the MET-SUV among tumors, statistical analyses were performed with ANOVA. Differences of P < 0.05 were considered statistically significant. Receiver operating characteristic (ROC) curves were generated from the TTP and the SLOPE in the late phase. The area under the curve, cutoff point, sensitivity, and specificity were calculated. All data were analyzed with SPSS 21 for Windows.

Results

The MET-TAC patterns could be divided into four types for differential diagnosis of tumors (Fig. 1).

Fig. 1
figure 1

Schematic pattern of MET-TAC types. Type A shown by astrocytic tumors was a MET-TAC with a MET-SUV with a low or middle level in the initial phase, a rapid increase in the early phase, and a consistent or limited/slow increase in the late phase. The MET-TAC of the normal frontal cortex belonged to this type. Type O shown by tumors that included an oligodendroglial component was a MET-TAC with a MET-SUV with a low or middle level in the initial phase, a rapid increase in the early phase, and a limited/slow decrease in the late phase. Type L shown by PCNSL was a MET-TAC with a MET-SUV with a low level in the initial phase, a rapid increase in the early phase, and a continuous increase in the late phase. Type M/H shown by meningiomas and hemangioblastomas was a MET-TAC with a MET-SUV with a high level in the initial phase, a rapid decrease in the early phase, and a continuous decrease in the late phase. The MET-TACs of the confluence and internal jugular vein belonged to this type. MET-TAC time-activity curve of maximum tumor MET-standardized uptake value, MET-SUV maximum tumor MET-standardized uptake value

Full size image

Type A, shown by astrocytic tumors, was a MET-TAC with a MET-SUV with a low or middle level in the initial phase, a rapid increase in the early phase, and a consistent or limited/slow increase in the late phase. The MET-SUV was markedly higher in the initial phase and had a tendency to increase more in early and late phases in GBM compared to AA and DA (Fig. 2a). MET-SUVs were significantly higher in early and late phases in GBM compared to AA, DA and the normal frontal cortex (*P < 0.05) (Fig. 2a). There was no significant difference in MET-SUVs between AA and DA (Fig. 2a). The MET-TAC of the normal frontal cortex belonged to this type, and the MET-SUVs of the normal frontal cortex were significantly lower than those of GBM, AA and DA (**P < 0.05) (Fig. 2a).

Fig. 2
figure 2

MET-TACs in brain tumors, the normal frontal cortex, confluence, and internal jugular vein. MET-SUVs were significantly higher in early and late phases (*P < 0.05) and showed a more marked increasing tendency in the early phase in GBM compared to AA, DA and normal frontal cortex (a). There was no significant difference in MET-SUVs between AA and DA (a). The MET-SUVs of the normal frontal cortex were significantly lower than those of GBM, AA and DA (**P < 0.05) (a). MET-SUVs were higher in the initial phase and showed a tendency to increase more in the early phase in AO compared to AOA, OD, and OA (b). There was no significant difference in MET-SUVs among AO, AOA, OD and OA (b). A tendency to decrease more was seen in the late phase in AOA, OD, and OA compared to AO (b). The MET-SUVs of OA were markedly lower than those of AOA and OD (b). MET-SUV with a low level in the initial phase, a rapid increase in the early phase, and a continuous increase in the late phase in PCNSL (c). MET-SUV with a high level in the initial phase, a rapid decrease in the early phase, and a continuous decrease in the late phase in meningiomas, hemagioblastomas, the confluence, and internal jugular vein (d). Error bars represent ± SEM, the 95% confidence interval. MET-TAC time-activity curve of maximum tumor MET-standardized uptake value, MET-SUV maximum tumor MET-standardized uptake value, DA diffuse astrocytoma, OD oligodendroglioma, OA oligoastrocytoma, AA anaplastic astrocytoma, AO anaplastic oligodendroglioma, AOA anaplastic oligoastrocytoma, GBM glioblastoma multiforme, PCNSL primary central nervous system lymphoma

Full size image

Type O, shown by tumors that included an oligodendroglial component, was a MET-TAC with a MET-SUV with a low or middle level in the initial phase, a rapid increase in the early phase, and a limited/slow decrease in the late phase. The MET-SUV was higher in the initial phase and had a tendency to increase more in the early phase in AO compared to AOA, OD, and OA (Fig. 2b). There was no significant difference in MET-SUVs among AO, AOA, OD and OA (Fig. 2b). A tendency to decrease more was seen in the late phase in AOA, OD, and OA compared to AO (Fig. 2b). The MET-SUVs of OA were markedly lower than those of AOA and OD (Fig. 2b).

Type L, shown by PCNSL, was a MET-TAC with a MET-SUV with a low level in the initial phase, a rapid increase in the early phase, and a continuous increase in the late phase (Fig. 2c).

Type M/H, shown by meningiomas and hemagioblastomas, was a MET-TAC with a MET-SUV with a high level in the initial phase, a rapid decrease in the early phase, and a continuous decrease in the late phase (Fig. 2d). The MET-TACs of the confluence and internal jugular vein belonged to this type (Fig. 2d).

The quantified parameters of MET-TAC are shown in Table 2. All parameters represent mean ± SEM, with a 95% confidence interval. The TTP was significantly shorter in tumors that included an oligodendroglial component (OD, OA, AO and AOA) compared to astrocytic tumors (DA, AA and GBM) (P = 0.001). The SLOPE in the late phase was significantly lower in tumors that included an oligodendroglial component compared to astrocytic tumors (P < 0.001). The TTP was significantly shorter in low grade tumors that included an oligodendroglial component (OD and OA) compared to DA (P < 0.05). The SLOPE in the late phase was significantly lower in low grade tumors that included an oligodendroglial component (OD and OA) compared to DA (P < 0.01). The TTP was significantly longer in PCNSL compared to GBM (P < 0.01). The SLOPE in the late phase was significantly higher in PCNSL compared to GBM (P < 0.01).

Table 2 The time-to-peak and regression coefficient (SLOPE) of the MET-TAC in each phase
Full size table

Diagnostic performances of parameters are summarized in Table 3. When we set the cutoff of the TTP to 15.5 min for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors, diagnostic accuracy was 77.4% sensitivity and 54.4% specificity. The area under the ROC curve was 0.672 (Fig. 3a). When we set the cutoff of the SLOPE in the late phase to − 0.04 h−1 for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors, the diagnostic accuracy was 74.2% sensitivity and 64.9% specificity. The area under the ROC curve was 0.731 (Fig. 3b). When we set the cutoff of the TTP to 27.5 min for the differentiation of low grade tumors that included an oligodendroglial component (OD and OA) from DA, diagnostic accuracy was 75.0% sensitivity and 81.2% specificity. The area under the ROC curve was 0.783 (Fig. 3c). When we set the cutoff of the SLOPE in the late phase to − 0.34 h−1 for the differentiation of low grade tumors that included an oligodendroglial component (OD and OA) from DA, diagnostic accuracy was 100% sensitivity and 47.9% specificity. The area under the ROC curve was 0.802 (Fig. 3d). When we set the cutoff of the TTP to 27.5 min for the differentiation of GBM from PCNSL, diagnostic accuracy was 100% sensitivity and 54.8% specificity. The area under the ROC curve was 0.819 (Fig. 3e). When we set the cutoff of the SLOPE in the late phase to 0.54 h−1 for the differentiation of GBM from PCNSL, diagnostic accuracy was 100% sensitivity and 67.3% specificity. The area under the ROC curve was 0.862 (Fig. 3f).

Table 3 Diagnostic performances of parameters
Full size table
Fig. 3
figure 3

a The ROC curve generated from the TTP for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors. b The ROC curve generated from the SLOPE in the late phase for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors. c The ROC curve generated from the TTP for the differentiation of low grade tumors that included an oligodendroglial component (OD and OA) from DA. d The ROC curve generated from the SLOPE in the late phase for the differentiation of low grade tumors that included an oligodendroglial component (OD and OA) from DA. e The ROC curve generated from the TTP for the differentiation of GBM from PCNSL. F, The ROC curve generated from the SLOPE in the late phase for the differentiation of GBM from PCNSL. When we set the cutoff of the TTP to 15.5 min for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors, the diagnostic accuracy was 77.4% sensitivity and 54.4% specificity. The area under the ROC curve was 0.672 (a). When we set the cutoff of the SLOPE in the late phase to − 0.04 h−1 for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors, the diagnostic accuracy was 74.2% sensitivity and 64.9% specificity. The area under the ROC curve was 0.731 (b). When we set the cutoff of the TTP to 27.5 min for the differentiation of low grade tumors that included an oligodendroglial component (OD and OA) from DA, the diagnostic accuracy was 75.0% sensitivity and 81.2% specificity. The area under the ROC curve was 0.783 (c). When we set the cutoff of the SLOPE in the late phase to − 0.34 h−1 for the differentiation of low grade tumors that included an oligodendroglial component (OD and OA) from DA, the diagnostic accuracy was 100% sensitivity and 47.9% specificity. The area under the ROC curve was 0.802 (d). When we set the cutoff of the TTP to 27.5 min for the differentiation of GBM from PCNSL, the diagnostic accuracy was 100% sensitivity and 54.8% specificity. The area under the ROC curve was 0.819 (e).When we set the cutoff of the SLOPE in the late phase to 0.54 h−1 for the differentiation of GBM from PCNSL, the diagnostic accuracy was 100% sensitivity and 67.3% specificity. The area under the ROC curve was 0.862 (f). TTP time-to-peak, SLOPE late regression coefficient of the MET-TAC in late phase

Full size image

Discussion

We identified some characteristic features of the MET-TACs obtained in this MET-PET dynamic study of the normal confluence, internal jugular vein, normal frontal cortex, and common brain tumors. We correlated the TTP and the slopes in the late phase of the MET-TACs with the classification of glioma subtypes and the differentiation between PCNSL and GBM.

Several static MET-PET studies have reported a higher diagnostic accuracy compared to MRI, but still found limited accuracy due to high overlap between WHO grades [22]. Dynamic analysis of FET-PET uptake improved the differential diagnosis between WHO grade II and WHO grades III/IV gliomas. WHO grade II gliomas typically showed steadily increasing time-activity curves as compared to WHO grades III/IV gliomas, in which the early peak activity occurred around 10–20 min after injection, and was followed by a decrease of FET uptake [9]. However, even with dynamic analysis of MET-PET, the uptake characteristics did not allow for classification of low-grade versus high-grade gliomas [20]. In our study, although there were no significant differences in MET-SUVs among Type O, MET-SUVs were significantly higher in all phases in GBM compared to AA and DA (Fig. 2). When limited to astrocytic tumors, MET-TAC analysis may be useful for WHO grading.

Jansen et al. correlated combined static and dynamic FET-PET uptake characteristics with the presence/absence of oligodendroglial components in WHO grade II and III gliomas. When they set a cutoff of the maximum SUV corrected for the mean background activity in the contralateral hemisphere to ≥ 2.4 and temporally increasing TAC for the differentiation of low-grade tumors that included an oligodendroglial component from astrocytic tumors and set one to ≥ 3.0 and temporally decreasing TAC for that of high-grade tumors, diagnostic accuracy was 62% sensitivity and 85% specificity [23]. However, it has not been reported that dynamic MET-PET is useful for the classification of glioma subtypes. In our study, when we set a cutoff for the SLOPE in the late phase to − 0.04 h−1 for the differentiation of tumors that included an oligodendroglial component from astrocytic tumors, the diagnostic accuracy was 74.2% sensitivity and 64.9% specificity. When we set the cutoff of the TTP to 27.5 min for the differentiation of low-grade tumors that included an oligodendroglial component from DA, diagnostic accuracy was 75.0% sensitivity and 81.2% specificity (Table 3). Our dynamic MET-PET data had comparable sensitivity and specificity to the previous study, which suggests that dynamic MET-PET could be as useful as dynamic FET-PET for classification of glioma subtypes.

GBM and PCNSL sometimes have similar features on conventional MRI. 18F-FDG PET may provide useful information for distinguishing WHO grade III/IV gliomas from other malignant brain tumors, but its specificity is limited. Importantly, the maximum SUV was significantly higher in PCNSL than in GBM [9]. In a dynamic MET-PET study, the ratio of the maximum SUV in late and early phases was considered to be a good diagnostic test when encountering difficulties in the differential diagnosis between GBM and PCNSL [24]. In our study, when we set the cutoff of the SLOPE in the late phase to 0.54 h−1 for the differentiation of GBM from PCNSL, the diagnostic accuracy was 100% sensitivity and 67.3% specificity. Our results were similar to the previous study, which suggests that dynamic MET-PET could be useful for the differential diagnosis between GBM and PCNSL.

The major factor that could facilitate MET uptake into tissues is the active transport mechanism, which would be responsible for biological activities including tissue proliferation and metabolic maintenance for tissue survival. The volume of MET uptake from active transport is thought to be positively correlated with tumor growth activity. In this study, the existence of this mechanism is supported by the tendency for a greater increase in the early phase of the MET-SUV in tumors with more malignant biological features. The active transport of MET is known to operate via the MET transporter. Amino acids such as methionine are transported in tissues by the L-transport system. MET is known to be transported predominantly by LAT1, one of the amino acid transporters in the L-transport system [25]. Immunohistochemical analysis indicates that the expression of LAT1 is in the membrane of many cancer cells and also tumor vessels, and it has been reported to increase with the degree of malignancy [26, 27]. Oligodendroglial tumors are known to have rich tumor vessels compared to astrocytic tumors histologically and may express more LAT1 [28]. This may account for our finding that the MET-SUVs of ODs, AOs and AOAs were greater compared to those of DAs and AAs in the early phase.

For brain tumors in which the volume of MET uptake from active transport is superior to the volume of passive outflow with regard to tissue blood volume (TBV) and tissue blood flow (TBF) (TBV/TBF) mechanisms, the MET-SUV is thought to increase during the early phase. Also, in meningiomas and hemangioblastomas, the MET uptake from active transport can be observed [29]. However, the volume of outflow of MET from TBV/TBF is thought to be much greater than the volume of the MET uptake from active transport [30]. Therefore, a marked decrease in the MET-SUV can be seen in the early phase in meningiomas and hemangioblastomas. There may be a certain degree of active transport of MET, even in normal frontal cortex, DAs, and OAs. The amount of TBV/TBF-related factors is much smaller in these tissues compared to the confluence, internal jugular vein, meningiomas, and hemangioblastomas, and the volume of MET outflow in normal frontal cortex, DAs, and OAs is much smaller in the early phase. Therefore, the MET-SUVs in normal frontal cortex, DAs, and OAs may be increasingly offset by the volume of MET outflow in this phase.

In our study, the MET-SUV decreased distinctly over time during the late phase in oligodendroglial tumors regardless of malignancy grading. This can likely be attributed to the rich tumor vessels found by histology in oligodendroglial tumors compared with astrocytic tumors, and may be strongly influenced by TBV/TBF-related mechanisms [28]. Furthermore, in oligodendroglial tumors, the amount of MET outflow from TBV/TBF is highly likely to overcome the amount of MET uptake from active transport. Alternatively, the MET-SUV increased distinctly over time during the late phase in PCNSL, which is a strongly enhanced tumor due to BBB disruption similar to GBM, but has less TBV/TBF than GBM [31]. Therefore, PCNSL is minimally affected by TBV/TBF-related factors, and might be more strongly influenced by active transport. For the differentiation of low-grade tumors from DA that included an oligodendroglial component, the TTP had better accuracy than the SLOPE in the late phase. A reason for this finding could be that DA might be more influenced by active transport than OD and OA because of the decreased TBV/TBF, and thus the TTP of DA was longer than that of OD and OA.

The following limitations were identified in our study. First, the study was retrospective. Second, we did not use the revised fourth edition of the WHO classification of tumors in the central nervous system, published in 2016. Our pathology results were collected from multiple institutions since 2015, and it is very difficult to re-classify tumors pathologically based on the WHO 2016. In fact, although an effort to reevaluate OA and AOA is currently underway, the risk of data error is high. Thus, for this report we have kept the 2007 classification. In a recent study, both static and dynamic 18F-FET PET imaging were helpful in the non-invasive determination of isocitrate dehydrogenase (IDH) mutation status [32]. So, the molecular genetic profile might influence the kinetic MET uptake; further studies are needed based on molecular genetic information.

Conclusions

The results of this study show that quantification of the MET-TAC for each brain tumor obtained from a dynamic MET-PET study could be helpful for non-invasive discrimination of subtypes of brain tumors, in particular gliomas.