We retrospectively enrolled 75 patients who underwent MET PET and MRS on the same day at our institution between January 2013 and June 2020. All patients had known pathological diagnoses of grade II or III astrocytomas, based on the 2016 WHO classification.11 Three neurosurgeons (EO, YI, and KM) independently assessed the patterns observed via fluid-attenuated inversion recovery (FLAIR) or T2- weighted image (WI) imaging to differentiate the group with a clear boundary between the tumor and the normal brain and the group with an unclear boundary. At the time of imaging assessment, the reviewers established three categories for classifying tumors: clear boundary, unclear boundary, or neither. Assessments were based solely on imaging, as physicians were blinded to clinical data such as patient name, age, sex, and histological diagnosis.
From this initial cohort of 75 patients, we sought to analyze only the outcomes of patients for whom all three reviewers or two of the three reviewers could agree on tumor classification. The physicians’ independent assessments aligned in 44 of the 75 (58.7%) cases studied. Specifically, their assessments aligned as follows: 14 patients were classified as being in the clear boundary group, and 30 patients were classified as being in the unclear boundary group. In 10 (13.3%) cases, two of the three physicians agreed on classification, and the remaining physician was unable to decide; of these ten patients, the clinicians classified seven patients as being in the clear boundary group and three patients as being in the unclear boundary group. From this cohort of 54 patients, we further excluded seven patients with an unknown IDH status; thus, 47 cases (24 men, 23 women; average age = 47.0 years; age range = 19–89 years) were ultimately included in our analysis,
Of these 47 cases, we noted 17 cases of grade II astrocytoma and 30 cases of grade III astrocytoma, 23 cases of IDH-mut and 24 cases of IDH-wt (Table 1). There were 18 patients in the clear boundary group and 29 patients in the unclear boundary group (Table 1). The average age was significantly higher for the IDH-wt group than the IDH-mut group (P = .005) (Table 1). No significant differences existed in sex and WHO grade between the IDH status groups; however, the IDH status and MRI findings for boundary status were significantly different (P < .001; Table 1).
PET was conducted prior to MRI. Eminence STARGATE (Shimadzu Corporation, Kyoto, Japan) was used, which was equipped with a three-dimensional acquisition system (Shimadzu Corporation) that provided 99 transaxial images at 2.65 mm intervals. The in-plane spatial resolution (full width at half-maximum) was 4.8 mm, and the scans were conducted in three-dimensional mode. MET was injected intravenously at 3.5 MBq/kg through the cubital vein.
During PET data acquisition, head position was corrected using laser beams projected onto ink marks drawn on the forehead, and images were reconstructed using an ordered subset expectation-maximization algorithm. Tracer accumulation in the region of interest (ROI) was analyzed using the standard uptake value (SUV), defined as the activity concentration in the ROI at a fixed time point divided by the injected dose and then normalized to the patient’s weight. The tumor-to-normal region (T/N) ratio of MET was calculated by dividing the maximum SUV of the tumor by the mean SUV of the contralateral normal frontal cortex. The ROI for the maximum tumor SUV was selected based on the areas with the highest tracer accumulation. The maximum tumor SUV was used instead of the mean SUV for the tumor to minimize the effect of tumor heterogeneity. Because of high and unexplained intersubjective SUV variability, we used the T/N ratio instead of the absolute SUV.
Coregistration of PET and MRI was conducted using the Dr. View image analysis software package (AJS, Tokyo, Japan). In this study, fusion images of PET and MRI are referred to as MET PET.
We used the Achieva 3.0T TX QD MRI system (Philips, Amsterdam, Netherlands) for transaxial T1-WI (repetition time (TR), 2200 ms; inversion time (TI), 950 ms; echo time (TE), 9.5 ms; field of view (FOV), 230 ×230 mm2; matrix, 512×512); T2-WI (TR, 4000 ms; TE, 90 ms; FOV, 230×230 mm2; matrix, 512×512); and FLAIR imaging (TR, 8000 ms; TI, 2400 ms; TE,120 ms; FOV, 230×230 mm2; matrix, 512×512). The slice thickness was 5 mm with a 1-mm slice gap. A gadolinium-based contrast agent, gadoteridol (Eizai, Japan), was injected intravenously at 0.1 mL/kg body weight for contrast-enhanced studies.
To quantify the extent of the lesion, we measured the area of the hyperintense lesion on the FLAIR images and the area of the entire brain in the same cross-section by using ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA). We considered the area ratio as the spreading tumor ratio. For the T2/FLAIR image used for the reviewers’ assessment, we selected the same slice as the cross-section showing the highest accumulation of MET in the tumor on MET PET.
Proton MRS was conducted simultaneously with conventional MRI using the single-voxel point-resolved spectroscopy technique with a TR of 2000 ms and TE of 288 ms. The total acquisition time required to obtain these parameters, including scanner adjustments, was <5 minutes. A cubic voxel with a side length of 2.0 cm was manually placed on the lesion, for which MET PET showed the highest accumulation. Spectra were generated using an internal scanner software, thereby providing automatic peak assignment and ratio calculations. The NAA/ Cr, Cho/Cr, and Cho/NAA peak ratios were recorded. Cr was used as the benchmark.
We prepared formalin-fixed paraffin-embedded sections of labeled tissue for histology. The specimens underwent H&E staining and immunohistochemistry to determine IDH status using an anti-IDH1R132H monoclonal antibody (1:20; Dianova; Hamburg, Germany 1:20). Immunoreaction was considered positive when the tumor cells showed strong and diffuse staining for IDH1R132H.
The specimens were also stained with an anti-Ki-67 antibody (1:100; Dako, Tokyo, Japan) to evaluate tumor proliferation. The Ki-67 labeling index was visually quantified by counting the number of immunopositive nuclei in areas with the highest Ki-67 immunoreactivity as the percentage of Ki-67 positive cells per 1000 tumor cells.
The antigen was retrieved in an autoclave (121°C for 15 min). We used the Envision kit (Dako) as a source of secondary antibodies and 3,3-diaminobenzidine as the chromogen.
We used Fisher’s exact test to assess associations between IDH status and sex, WHO grade, and tumor-brain boundary status, and the Mann–Whitney U test to compare the spreading tumor ratio, MRS parameters, MET T/N ratio, and Ki-67 labeling index between IDH-wt and IDH-mut groups.
Statistical significance was set at P<.05. Receiver operating characteristic (ROC) analysis was conducted to determine the optimal thresholds of the Cho/Cr, Cho/NAA, and MET T/N ratios for discriminating IDH-wt from IDH mut and the area under the curve. We determined the optimal sensitivity and specificity from the highest sum of the ROC curve. R software (R Project for Statistical Computing, Vienna, Austria, version 4.0.3) was used for all statistical analyses.