Brain tumors are the most prevalent type of solid cancer in childhood, are the most common cause of pediatric cancer death, and are a significant cause of long-term disability [6]. Histopathology is still the gold standard in diagnosing pediatric brain tumors; however, even image-guided biopsies have an appreciable morbidity and mortality. Additionally, the heterogeneity of some tumors introduces problems with “sampling” error and undergrading of the tumor [7].
Noninvasive and accurate differentiation between neoplastic and non-neoplastic pediatric brain lesions as well as low- and high-grade neoplasms is challenging and important in determining the correct treatment plan. MRS provides information about biochemical characteristics, metabolic heterogeneity of the lesion, and the surrounding brain tissue [8].
Previous studies have shown the ability of various metabolite ratios to diagnose different brain lesions with varying sensitivities and specificities. Most of the studies used Cho/NAA, Cho/Cr, and NAA/Cho ratios; however, relying on these ratios alone proved to have certain fallacies [9]. In order to improve the diagnostic accuracy of MRS, the current study also evaluated the Cho/Naa+Cr and the myo-inositol/Cr ratios.
In this study, the Cho/NAA+Cr and Cho/NAA ratios had the highest sensitivity and specificity for differentiation of neoplastic from non-neoplastic lesions. Using ROC curve analysis, the Cho/NAA+Cr had 88% sensitivity and 100% specificity for the prediction of neoplastic lesions at a cutoff point > 0.8 with AUC = 0.91 and P value = 0.009. The Cho/NAA ratio showed similar results at a cutoff point > 2 having 88% sensitivity and 75% specificity with AUC = 0.91 and P value = 0.01. The Cho/Cr ratio at a cutoff point > 1.86 had 85% sensitivity and 75% specificity with AUC = 0.85 and P value = 0.02.
These results were found to be in agreement with the results of Karatag et al. [10], who reported that a Cho/NAA ratio > 1.83 showed 87.2% sensitivity and 100% specificity while a Cho/Cr ratio > 1.98 showed 71.8% sensitivity and 100% specificity for differentiation of neoplastic from non-neoplastic lesions. McKnight et al. [11] also reported that Cho/NAA ratio correlates with cell density and cell proliferation index. They also found that a ratio greater than 2 shows 96% sensitivity and 70% specificity for differentiating neoplastic from non-neoplastic lesion. Butzen et al. [12], however, reported a lower cutoff value of the Cho/NAA ratio being > 1 to indicate a neoplastic process with a sensitivity of 79% and specificity of 77%. On the other hand, the NAA/Cr ratio in this study showed no statistically significant values for the prediction of neoplastic lesions with low AUC as well as sensitivity and specificity.
The radiologic grading of gliomas using conventional MRI has a sensitivity ranging from 55 to 83% due to similar imaging characteristics of some low-grade and high-grade neoplasms. Attempts to grade tumors using MRS suggest that with increasing grade, there is an increase in the Cho/Cr ratio, a reduction in NAA, and increase in lactate/lipid peak [13]. Kapsalaki et al. [14] also demonstrated that the higher the Cho/NAA ratio is, the higher the astrocytoma grade.
In this study, the ROC curve analysis showed that the Cho/NAA at a cutoff point > 3.3 had 92% sensitivity and 72% specificity for the prediction of high-grade neoplasms while the Cho/Cr ratio at a cutoff point > 3.5 had 83% sensitivity and 78% specificity. Furthermore, the Cho/NAA+Cr ratio at a cutoff point > 1.3 had 83% sensitivity and 72% specificity and the NAA/Cho ratio showed 92% sensitivity and 95% specificity at a cutoff of < 0.5.
This was in concordance with the results of Karatag et al. [10], who reported that a Cho/NAA ratio > 3.2 shows sensitivity of 82.6% and specificity of 100% while a Cho/Cr ratio > 2.2 shows 95.7% sensitivity and 84.6% specificity in differentiation of low-grade versus high-grade neoplastic lesions. Oz et al. [15] also reported that Cho/NAA ratio greater than 2, a Lac/NAA ratio greater than 0.25, and the presence of lipid at MR spectroscopic imaging are characteristics of a high-grade tumor.
Myo-inositol is one of the most abundant metabolites visible on MRS at short TE. It is involved in the production of proteolytic enzymes found in aggressive primary tumors through activation of protein C [16]. Consequently, the MI/Cr ratio was measured in this study to help improve the prediction of tumor grade by MRS. The level of myo-inositol was significantly higher in low-grade gliomas compared with high-grade gliomas. At a cutoff point < 1.5, the MI/Cr ratio had 83% sensitivity and 79% specificity for prediction of high-grade neoplastic lesions.
These results were similar to those of Castillo et al. [17], who compared mI/Cr ratio in 20 patients with high-grade glioma, 14 patients with low-grade glioma, and 5 control cases. Castillo found elevated mI/Cr in low-grade glioma with mean mI/Cr 0.8 ± 0.25, lowered mI/Cr in anaplastic astrocytoma 0.33 ± 0.16, and GBM 0.15 ± 0.12 compared with control subjects 0.49 ± 0.07.
Metwally et al. [16], in a study to determine the diagnostic accuracy of MI/Cr ratio in the grading of gliomas, also found that the level of MI/Cr was higher (2 ± 1.5) in low-grade astrocytomas than anaplastic astrocytomas. They reported that the MI/Cr ratio had 100% sensitivity and 92.8% specificity for the grading of gliomas.
Approximately 60% of all pediatric tumors arise from the posterior fossa. In most cases, these tumors are grade IV medulloblastoma, grade I pilocytic astrocytoma, or grade II or III ependymoma [3]. Sometimes, a cystic/necrotic medulloblastoma may have similar imaging features to posterior fossa pilocytic astrocytoma. Furthermore, it is often difficult to distinguish medulloblastoma from ependymoma due to similar appearance on conventional imaging. MRS and diffusion imaging are particularly useful for the differentiation of these lesions [3].
The two cases in this study that were histologically proven to be medulloblastomas, showed higher levels of choline than the other posterior fossa tumors with slight increase of myo-inositol, low NAA levels, lactate peaks, and no lipids. Mittal [18] described elevated choline and taurine peak with dwarfing of other metabolites and no lipid or lactate peak. This was in agreement with the results of this study apart from the taurine peak which is detected in only 60% of medulloblastomas.
Majos et al. [19] in a comparative study of medulloblastoma to low-grade glioma found three to fourfold increase in Cho level with markedly decreased NAA compared with low-grade gliomas. In the study done by Cuellar-Baena et al. [20], they also found that medulloblastomas had higher choline levels than ependymomas and pilocytic astrocytomas with lower NAA levels and lactate peaks. They explained this metabolic profile to be due to increased membrane turnover, low neuronal viability, and glycolysis alterations.
Ependymomas, however, had higher myo-inositol than medulloblastoma or pilocytic astrocytoma. They also had relatively high choline levels, particularly grade III ependymoma, but falling between medulloblastoma and pilocytic astrocytoma. Both ependymomas and medulloblastomas had lactate peaks with no lipids. However, Yuh et al. [21] reported that there is usually choline elevation with reduction of NAA in ependymoma, so there is no significant difference in-between ependymoma and medulloblastoma due to variability of Cho/NAA within each tumor.
In this study, pilocytic astrocytomas had mildly increased Cho and Cho/Cr ratio with low levels of Cr and mI which is consistent with the low cellularity of the tumor. They also showed elevated lactate doublet and no lipid peak. Panigrahy et al. [3] also found that the Cr level and Lipids are characteristically low in pilocytic astrocytoma, but mean lactate levels are higher than in other tumors.
Cecil et al. [22] stated that pilocytic astrocytomas exhibit elevated lactate and Cho and diminished levels of NAA, Cr, and mI. Hwang et al. [23] suggested that the high lactate level in the benign tumor may be due to an abnormal number or dysfunction of mitochondria, which would interfere with the process of oxidative phosphorylation and electron transport, alterations in proportional oxygen delivery, and oxygen extraction or usage by tumor or anaerobic glycolysis by tumor cells.
Diffuse pontine gliomas account for 10–15% of childhood brain tumors [24]. There were two cases of diffuse pontine gliomas in this study diagnosed radiologically and by clinical follow-up. They showed elevation of choline and reduced NAA and creatine thus confirming neoplastic nature. They also showed relatively low Cho/Cr and Cho/NAA ratios with no lipid or lactate peaks.
To differentiate primary high-grade glioma from solitary metastases, Cho level is measured outside the tumor margin or enhancing edge. In the current study, the Cho/NAA and Cho/Cr ratios measured outside the margin of high-grade gliomas ranged from 1.7 to 7 and from 2.4 to 3.5 respectively while in the case of metastasis, they measured 0.5 and 0.7. This was in agreement with the results of Al-Okaili et al. [1] who used a perilesional Cho/NAA ratio of 1 outside the enhanced or T2 defined margin. This showed 100% accuracy in discrimination of primary from secondary neoplasms. Tsougos et al. [25] found that the Cho/Cr ratio measured in the peritumoral region had 89% sensitivity and 62% specificity at a cutoff value 1.4, while the Cho/NAA showed 78% sensitivity and 93% specificity at a cutoff value 1.1.
A limitation of this study is the inclusion of a small number of cases. Further studies are needed to quantify the extent to which MRS facilitates diagnosis, changes pediatric patient management, and its impact on the outcome of these patients.