INTRODUCTION

Glioblastoma (GBM) and brain metastases are the two most common tumors in adults, with GBM accounting for 16% of primary intracranial tumors and 55% of gliomas.[1] Brain metastases account for 10–26% of intracranial tumors, of which only 25–30% are monofocal.[2,3] Cases with multifocal brain metastases or previous malignancies are diagnosed on routine magnetic resonance imaging (MRI). However, discriminating between solitary brain metastasis (SBM) and GBM when no primary malignancy has been detected is quite difficult because both tumors appear similar on images, with central necrosis, irregular margins, surrounding edema, ring enhancement, and often calcification and bleeding.[4,5] The differential diagnosis of these two tumors is essential for treatment planning and patient outcome prediction.[6-9] The treatment for GBM is maximal resection, radiation therapy, and chemotherapy with temozolomide.[10] In contrast, SBM is treated with stereotactic radiosurgery or surgical resection combined with whole-brain radiation after surgery, depending on the size and location of the tumor.[11,12] Pathology is the gold standard for discriminating between these two tumors at the cellular level. However, accessing tumors in deep or functional areas for surgery or biopsies can be challenging.[13,14]

Conventional MRI with contrast injection is an effective method of diagnosing brain tumors. However, its diagnostic capacity is limited. The rate of incorrect differential diagnosis of GBM and SBM on routine MRI is up to 40%.[15,16] The area around the tumor is critical for distinguishing between these two types. In GBM, the peritumoral region is a combination of vasogenic edema and tumor cell infiltration, whereas, in SBM, the peritumoral part is purely vasogenic edema.[12,17] Some advanced techniques, especially magnetic resonance perfusion (MR perfusion) and magnetic resonance spectroscopy (MR spectroscopy), have been applied to analyze the peritumoral region to differentiate between GBM and SBM. MR perfusion allows quantitative and qualitative assessment of tumor angiogenesis and peritumoral edema without being affected by blood-brain barrier disruption. Relative cerebral blood volume (rCBV) maps are used to quantify areas of angiogenesis.[18,19] MR spectroscopy provides metabolic information on brain tumors by measuring specific amino acids, such as N-acetylaspartate (NAA), choline (Cho), creatine (Cr), lipid, and lactate and their ratios in the tumor area and the peritumor edema area.[20,21]

Several studies have focused on the use of MR perfusion and MR spectroscopy and their combination with other techniques, such as diffusion-weighted imaging and diffusion tensor imaging, in the differential diagnosis of GBM and SBM.[12,15,18-20] However, only a few studies have considered the value of combining MR perfusion and MR spectroscopy in the differential diagnosis of these two types of tumors, and the results are inconsistent.[18,22-24] Therefore, we assessed the value of MR perfusion and MR spectroscopy in 3.0-Tesla MRI for distinguishing between GBM and SBM.

MATERIAL AND METHODS Study population

This cohort study consisted of 36 patients who were preoperatively imaged using 3.0-Tesla MRI with conventional sequences, dynamic susceptibility contrast (DSC) MR perfusion, and multivoxel MR spectroscopy at Viet Duc Hospital, Hanoi, Vietnam, from January 2020 to October 2022. All patients underwent biopsy or surgery and were histopathologically diagnosed. Patients were divided into two study groups based on the histopathological diagnosis of GBM (24 patients) and SBM (12 patients). Our study was approved by the Ethics Committee (QD: 2682/QDĐHYHN July 13, 2021) and conformed to the principles of the Declaration of Helsinki. The Hanoi Medical University Review Board waived the patient or guardian consent requirement because the study was a retrospective and involved analysis of anonymized image data.

MRI

The imaging examinations were performed on a 3.0-Tesla MRI scanner (GE SIGNA™ Pioneer, GE Healthcare, Chicago, IL, USA) using a head coil with conventional sequences including T1-weighted (T1W) spin echo, fluid-attenuated inversion recovery (FLAIR), T2-weighted (T2W) gradient echo, and diffusion-weighted imaging with apparent diffusion coefficient map reconstruction with the protocol applied to all patients [Table 1]. DSC perfusion was performed during the first pass of a bolus contrast injection before T1W post-contrast images were acquired. Dynamic T2*-weighted gradient-recall echo-planar imaging was used with the same slide thickness as for FLAIR or T2W axial sequences. The correlation between the level of tumor enhancement on T1W after contrast injection and the area of hyperperfusion on the rCBV map was determined. The average value of rCBV was calculated for each group of tumors.

Table 1:: MRI sequences.

Sequence Plane TR (ms) TE (ms) Thickness (mm) Matrix FOV Others FLAIR Axial 8500 117 5 184×256 240×240 IR 2500 ms T2 TSE Axial, coronal 2500 100 5 360×288 220×220 T2*/SWI Axial 360 10 5 µ DWI Axial 5202 78 5 230×240 T1 SE Axial, sagittal 2325 24 5 240×240 300×224 T1 SE CE+ Axial, sagittal 2325 24 5 240×240 300×244 α DSC-MRI GRE EPI 1250 45 5 88×87 338×240 β MR spectroscopy 2D multivoxel CSI 8500 144 240×240 Voxel 3D CSI 2000 144 Voxel size of 0.9×1.6×1.2 cm3

Multivoxel MR spectroscopy was localized using a FLAIR or T1 GE 3D sequence. The average density of metabolites (Cho, NAA, and Cr) and the ratios of metabolites (Cho/NAA and Cho/Cr) of the solid tumor component and peritumoral region were computed to assess the infiltrative nature of the tumor.

Image analysis

The MR findings were read by a neuroradiologist, with more than 10 years of experience, without clinical pathology information. Related lesions on conventional MRI included the following:

The solid tumor component was the enhanced part of post-contrast T1W

The peritumoral region was defined as the region around the solid part of the tumor, hypointense on T1W, and hyperintense on FLAIR, with no contrast enhancement

The contralateral normal white matter area was the normal white matter portion located in the contralateral brain parenchyma on the same plane as the lesion

The cystic degenerating or necrotic tumor region was the lesion significantly hypointense on T1W images and hyperintense on T2W images and without enhancement after injection

Hemorrhagic and calcified areas in the tumor were indicated by T2* hypointensity.

Measurement of cysts, calcifications, necrosis, and juxta-skull base was avoided due to possible signal interference, leading to reduced accuracy. The solid tumor component, pCBV, and contralateral benign white matter are shown in [Figure 1].

Figure 1:: Left frontal-parietal glioblastoma in a 58-year-old female patient. The locations of the solid tumor component (yellow arrow), peritumoral region (green arrow), and contralateral benign white matter (red arrow) on magnetic resonance imaging. (a) Axial fluid-attenuated inversion recovery image; (b) Non-enhanced axial T1-weighted (T1W) image; and (c) Contrast-enhanced axial T1W image. The tumor is heterogeneously hypointense on T1W with surrounding edema on fluid-attenuated inversion recovery and causes a mass effect that compresses the midline to the right. After contrast, the lesion is strongly enhanced.

Export to PPT

Each patient underwent MR spectroscopy with two-dimensional multivoxel chemical shift imaging to measure metabolites in the solid tumor component, peritumoral region, and contralateral benign white matter. Three regions of interest (range of interests [ROIs], ~30 mm2) were located on each of these sites. The concentrations and ratios of metabolites were determined using spectroscopic analysis of the peaks for NAA at 2.18–2.01 parts per million (ppm), Cr at 3.15–3.0 ppm, and Cho at 3.36–3.12 ppm. The software automatically calculated the Cho/NAA and Cho/Cre ratios for the results, as shown in [Figure 2].

Figure 2:: Left frontal-parietal glioblastoma in a 58-year-old female patient. (a-c) Positions of three regions of interest in the solid tumor component, peritumoral region, and contralateral benign white matter (indicated by red circles). (d-f) Magnetic resonance (MR) spectroscopy in the respective regions. (d) MR spectroscopic data of the solid tumor component reveals choline (Cho) elevation, N-acetylaspartate (NAA) reduction, and an elevated Cho/creatine ratio suggesting a tumor lesion. (e) MR spectroscopy of the peritumoral region shows increasing Cho and an increased Cho/NAA ratio in keeping with tumor infiltration. (f) Normal MR spectroscopy of contralateral benign white matter.

Export to PPT

On the CBV map, three ROIs (~30 mm2) were delineated on the solid tumor component, pCBV, and contralateral benign white matter and compared to conventional sequences. The solid part of the tumor was selected as the most hyperperfused part presenting as highly intense on the CBV color scale, excluding confounding areas. The hemodynamic parameter was calculated as rCBV by normalizing the maximum values obtained in the solid part of the tumor and the peritumoral region against contralateral benign white matter [Figure 3].

Figure 3:: A right frontal solitary brain metastasis originating from lung cancer in a 50-year-old female patient. The positions of three regions of interest (yellow circles) in the solid tumor component (yellow arrow), peritumoral region (green arrow), and contralateral benign white matter (red arrow). (a) Fluid-attenuated inversion recovery image; (b) Non-enhanced axial T1-weighted image; (c) Contrast-enhanced axial T1-weighted image; and (d) Cerebral blood volume (CBV) map. The tumor is hypointense on T1W with moderate surrounding edema on axial fluid-attenuated inversion recovery, causing a mass effect that compresses the midline to the left, and heterogeneous enhancement. The CBV map shows hyperperfusion of the tumor compared to the surrounding brain tissue.

Export to PPT

Statistical analysis

Data were analyzed using SPSS 20.0 (SPSS Inc., Armonk, NY, USA). Quantitative parameters are given as the mean ± standard deviation and qualitative parameters as number (n) and percentage (%). The Mann–Whitney U-test for non-normally distributed variables and independent samples t-test for normally distributed variables were used to compare differences between the GBM and SBM groups. P < 0.05 was considered significant. Indicators found to be significant for differentiating between the two groups and the combinations of these parameters were analyzed using the receiver operating characteristic (ROC) curve to demonstrate their prognostic value.

RESULTS

Patient characteristics are presented in [Table 2]. No significant differences were found in mean age and sex between GBM and SBM (P > 0.05).

Table 2:: Patient characteristics.

Characteristic Tumor -value GBM (n=24) SBM (n=12) Age 57.3±14.9 59.8±7.2 0.113* Sex Men: 12 (50.0%) Men: 5
(41.7%) 0.637** Women: 12
(50.0%) Women: 7
(58.3%)

[Table 3] shows the MR perfusion and MR spectroscopy values in the solid tumor component, pCBV, and contralateral benign white matter in GBM and SBM.

Table 3:: MR perfusion and MR spectroscopy in the solid tumor component, peritumoral region, and contralateral benign white matter in GBM and SBM.

Parameter GBM SBM P-value tCBV 8.972±2.169 9.946±3.668 0.410* pCBV 1.978±0.976 1.012±0.554 0.003* tCho 116866±55991 90498±50897 0.179* tNAA 37328±21059 48530±29535 0.198* tCre 34453±20912 28558±15734 0.476** pCho 90726±43134 61175±40632 0.057* pNAA 67691±34812 70980±37344 0.560* pCre 45352±23443 40529±22586 0.796*

The CBV of the peritumoral region (pCBV) in GBM cases was significantly higher than that in SBM cases (P < 0.05). The area under the curve (AUC) for pCBV was 0.823 [Figure 4]. A cutoff value of 1.41 was established to differentiate between GBM and SBM, with a sensitivity of 70.83%, specificity of 83.33%, negative predictive value (NPV) of 89.47%, and positive predictive value (PPV) of 58.82%. The remaining indices were not significantly different between these two tumors (P > 0.05).

Figure 4:: Receiver operating characteristic curves for cerebral blood volume (CBV) of the peritumoral region (pCBV), the ratio of CBV of the pCBV to CBV of normal white matter (rCBVp/n), the ratio of choline (Cho) to N-acetylaspartate (NAA) in the solid tumor component (tCho/NAA), the ratio of Cho to NAA in the pCBV (pCho/NAA), and the combination of these parameters for differentiating between glioblastoma and solitary brain metastasis.

Export to PPT

[Table 4] shows the ratios parameters determined from MR perfusion and MR spectroscopy images of the solid tumor component, pCBV, and contralateral benign white matter in GBM and SBM.

Table 4:: The ratios of parameters determined from magnetic resonance perfusion and magnetic resonance spectroscopy images of the solid tumor component, peritumoral region, and contralateral benign white matter in GBM and SBM.

Parameter GBM SBM P-value rCBVt/n 5.354±2.461 7.080±2.042 0.044* rCBVp/n 1.099±0.484 0.745±0.459 0.043* tCho/NAA 4.600±4.744 2.319±1.483 0.041** pCho/NAA 1.402±0.452 0.995±0.903 0.006** tCho/Cre 4.115±2.544 3.469±1.735 0.476** pCho/Cre 2.145±0.743 1.569±0.996 0.435*

The ratio of CBV of the solid tumor component to CBV of normal white matter (rCBVt/n) in GBM cases was significantly lower than that in SBM cases (P < 0.05). The AUC for pCBV was 0.757 [Figure 5]. A cutoff value of 5.21 was established to differentiate between GBM and SBM, with a sensitivity of 62.50%, specificity of 91.67%, NPV of 55.00%, and PPV of 93.75%.

Figure 5:: Receiver operating characteristic curve for the ratio of cerebral blood volume (CBV) of the solid tumor component to CBV of normal white matter for differentiating between glioblastoma and solitary brain metastasis.

Export to PPT

The ratio of CBV of the peritumoral region to CBV of normal white matter (rCBVp/n), the ratio of Cho/NAA in the solid tumor component (tCho/NAA), and the ratio of Cho/NAA in the peritumoral region (pCho/NAA) in GBM cases were significantly higher than those in SBM cases (P < 0.05). The AUC for rCBVp/n, tCho/NAA, and pCho/NAA were 0.743, 0.712, and 0.781, respectively [Figure 4].

[Table 5] shows the cutoff value, AUC, sensitivity, specificity, PPV, and NPV of each of the parameters significantly different between GBM and SBM. The combination of all four significantly different metrics (pCBV, rCBVp/n, tCho/NAA, and pCho/NAA) resulted in an AUC of 0.872 [Figure 4]. The sensitivity, specificity, NPV, and PPV of this combination were 90.91%, 71.43%, 83.33%, and 83.33%, respectively.

Table 5:: The diagnostic performance of significantly different parameters between GBM and SBM.

Parameters Cutoff AUC Se (%) Sp (%) NPV (%) PPV (%) pCBV 1.41 0.819 70.83 83.33 89.47 58.82 rCBVt/n 5.21 0.757 62.50 91.67 55.00 93.75 rCBVp/n 1.04 0.743 54.16 91.67 92.86 50.00 tCho/NAA 2.11 0.712 87.50 66.67 84.00 72.73 pCho/NAA 1.02 0.781 87.50 75.00 87.50 81.82 pCBV+rCBVp/n+tCho/NAA+pCho/NAA 0.55 0.868 95.45 78.57 87.50 91.67 DISCUSSION

Differentiation between GBM and SBM is important for treatment planning and patient outcomes.[6-9] Due to the similar clinical and imaging features of these two tumors, GBM is often confused with SBM and vice versa during diagnosis.[15,16] We studied the differences in peritumoral areas between the two tumor types. Peritumoral edema in GBM is a combination of vasogenic edema and tumor cell infiltration, whereas that in SBM is pure vasogenic edema.[12,17,25] Advanced MR techniques such as perfusion and spectroscopy, which evaluate peritumoral tumor infiltrates, help to differentiate between these two tumors.

Similar to the results of Jung et al.,[17] the findings of our 36-patient study showed that both types of tumors are more common in individuals more than 50 years of age and are found in both sexes with no significant difference. In DSC perfusion, the difference in tCBV values was not statistically significant (P > 0.05), the tCBV of GBM was 8.972 ± 2.169, and the latter’s group value was 9.122 ± 4.442. This conclusion is similar to those of the previous studies.[12,25,26] On the other hand, the pCBV of the GBM group was significantly higher than that of the SBM group (1.978 ± 0.976 vs. 0.943 ± 0.615; P < 0.05). This difference is due to the histology of the peritumoral area of these two tumor types. This region is a combination of vasogenic edema and tumor cell infiltrates in GBM and, therefore, has increased angiogenesis, which is indicated by a higher rCBV than that of SBM, which shows only vasogenic cerebral edema.[12,17,25] Our pCBV results for GBM were higher than those of Tong et al. (1.8 ± 0.7),[27] Tsougos et al. (1.68 ± 0.59),[26] and Romano et al. (1.46).[25] pCBV of SBM in our study was also higher than that of Tong et al. (0.6 ± 0.1),[27] and Romano et al. (0.87)[25] but lower than Tsougos et al.’ result of 1.06 ± 0.38.[26] Although there is a difference in results between our study and previous studies, this difference is not remarkable. pCBV, higher in GBM than in SBM, is an important indicator for distinguishing between these two types of tumors. When calculating rCBV for the tumor area and the peritumor area with normal white matter on the opposite side, we also obtained ratios with statistically significant differences between tumor types (P < 0.05; Table 4).

Magnetic resonance spectroscopy showed that the mean values of metabolites were not significantly different between GBM and SBM (P > 0.05; Table 3). However, tCho/NAA of GBM was significantly higher than that of SBM (4.600 ± 4.744 vs. 2.319 ± 1.483; P < 0.05). Compared to Tsougos et al.’s study, tCho/NAA of GBM in our study was slightly higher and that of SBM was lower.[26] In addition, we also found that GBM had a significantly higher pCho/NAA ratio than SBM (1.402 ± 0.452 vs. 0.995 ± 0.903; P < 0.05). These differences can also be explained by the differences in the histology of the pCBV of these two types of tumors described above. This ratio represents the infiltration and spread of tumor cells and is helpful in the differential diagnosis of GBM and SBM. We simultaneously calculated the average value of the ratio of Cho to Cr in the tumor region (tCho/Cre) and peritumor area (pCho/Cre) of the GBM group and the SBM group; the results showed that the GBM group had higher tCho/Cre and pCho/Cre ratios than the SBM group. This result is similar to the findings of Arévalo-Sáenz[28] and Server et al.[29]

We used ROC curves to analyze the sensitivity and specificity of cutoff values of statistically significant factors. The pCBV cutoff at 1.41 had an AUC, sensitivity, and specificity of 80.30%, 70.83%, and 83.33%, respectively. In Tsougos’ study of 49 patients, the cutoff was 1.7, which had higher sensitivity and specificity (80% and 94%, respectively) than our cutoff.[26] In our study, the cutoff of tCho/NAA at 2.11 had an AUC of 71.2%, sensitivity of 87.50%, and specificity of 66.67%, with a PPV of 84.00% and an NPV of 72.73% [Table 5]. In contrast, Tsougos did not find a statistically significant difference between the two groups.[26] The pCho/NAA ratio with a cutoff of 1.02 in our study had an AUC of 78.1%, sensitivity of 87.50%, and specificity of 75.00%. The cutoff of 1.115 for pCho/NAA in Arévalo-Sáenz’s study had higher sensitivity and specificity (93.87% and 93.33%, respectively).[28] In addition, the pCho/NAA cutoff of 1.10 in Tsougos et al.’s study had lower sensitivity (78%) but higher specificity (93%) than that in our study.[26]

Finally, we analyzed the diagnostic value of combining MR spectroscopy and MR perfusion in differential diagnosis of GBM and SBM tumors [Table 5, Figures 4 and 5] by combining the significantly different values. We found an increased AUC of 86.80%, with a cutoff of 0.55 having a sensitivity of 95.45% and specificity of 78.57%. This showed that combining sequences increased the reliability of the indicators in distinguishing between the two tumors.[25]

Our study has some limitations. First, it had a small sample size, which might have led to bias in assessing diagnostic efficiency. Second, the measurements might have been affected in some cases due to tumoral heterogeneity, including hemorrhage, calcification, necrosis, or adjacent bone.

CONCLUSION

Our study showed that advanced MRI techniques, including perfusion and spectroscopy, can be used to differentiate between GBM and SBM. The parameters with high diagnostic value for differentiating between these two tumors were pCBV and pCho/NAA, with cutoff values of 1.41 and 1.02, respectively. The combination of these ratios increases the diagnostic performance of MRI in distinguishing between the two tumors.