Revisiting Concepts of Magnetic Resonance Spectroscopy in the Evaluation of Brain Lesions: An Institutional Experience

Introduction

The ability of magnetic resonance imaging (MRI) in tissue characterization is often limited. Tumors can mimic appearance of abscesses, and vice versa. In vivo magnetic resonance spectroscopy (MRS) is used for monitoring both disease progression and treatment response.¹ Unfortunately for patients, many tumors have advanced to an unresectable or untreatable stage by the time they show typical imaging characteristics. In brain tumor patients especially, though the prognosis in general is relatively poor, there is a subpopulation in which aggressive therapy at an early stage has proved effective in improving prognosis.² For such an approach the early detection and characterization of the lesion, diagnosis, grade, type, and volume is of critical importance.³ Thus the need to assess beyond the purely anatomic aspects, such as biochemistry and tissue physiology, requires the development of functional techniques such as functional MRI, perfusion-weighted imaging, diffusion-weighted imaging, and MRS. Even though extensive studies and research have been done on the topic, brain tumors show poor prognosis. Unique metabolism of brain tumors and their resistance to conventional treatment demand for more research in search of new vulnerabilities, internal biology, and biomarkers.

Evaluating metabolite concentrations is a basis for a system of classification and its reprogramming has been hallmarks of cancer. The potential cause behind an altered metabolism is to sustain rapid proliferation and growth in brain tumors. Since the approval of MRS by the United States Food and Drug Administration in 1995, a new era in neuroradiology had emerged.⁴ MRS is a noninvasive technique that enables tissue characterization on a biochemical level surpassing that of conventional MRI. It works on the same physical principles as MRI, the difference being that the radiofrequency signals emitted by nuclei in tissues are used to determine the concentration of different metabolites in the tissue instead of creating gray-scale images.⁵ It can be performed in 10 to 15 minutes and combined with conventional imaging protocols. Thus, MRS offers the possibility of not just visualization of the lesion but also biochemical characterization simultaneously.⁶

MRS does not replace conventional MRI but compliments the information provided by it.⁷ It may provide information as a prognostic indicator, help follow progress of the disease, and evaluate response to treatment.⁸ Since its introduction, many trials and studies have been done to assess the usefulness of MRS in brain pathologies ranging from tumors to infections to degenerative processes.⁹ Although most of these studies indicate MRS can be a good adjuvant to MRI, there are at present insufficient data on the actual positive predictive values or negative predictive values of MRS for it to be used routinely. This study aimed to evaluate the pattern and spectrum of metabolite peaks on MRS in different brain tumors and also to see role of MRS in grading of gliomas and its role in primary versus secondary brain lesions.

Materials and Methods

This is a retrospective observational study in a tertiary care institute in India. Record of postoperative histology or stereo-tactic biopsy-proven cases of brain tumors in 1 year duration was observed. MRI data of these cases were retrieved through hospital PACS (picture archiving and communication system). Cases in whom MRI with MRS, either single or multivoxel, was performed in our institute were included in study. Cases with inconclusive or noisy MRS spectrum were excluded. All scans were performed with 3T MRI (GE Healthcare) using a head coil. A total of 61 cases were included in the study. Appropriate anesthesia was given in pediatric and uncooperative patients. Single voxel and multivoxel studies were done in 34 and 48 cases, respectively. Both single and multivoxel studies were done in 21 cases. Time of echo applied was short (30 ms) and medium (140 ms) in all cases. Long echo time (270 ms) was also applied in eight cases. Assignment of resonance was done according to literature reports. Resonance at 3.2 ppm was assigned as choline, 3.0 ppm as creatine, and 2.0 ppm as N-acetylaspartate (NAA). A peak at 0.9 ppm with short spin echo with reversal at medium echo was assigned as lactate and without reversal as lipid and resonance at 1.5 ppm as alanine. The pattern of metabolite spectrum was observed in different types of brain tumors. Means with standard deviation of different metabolite ratios (choline/creatine, choline/NAA, and NAA/creatine) were calculated in different types of tumors. Gliomas of different grades on the basis of histological examination were analyzed for statistically significant differences in the values of these metabolite ratios with p-value <0.005. Any statistically significant difference between these metabolite ratios and for presence of other metabolites was also analyzed for glioma versus metastases (►Fig. 1).

Close

Fig. 1

Homogenously enhancing dural based cerebral convexity extra-axial lesion (A) which shows reduced diffusivity (B),veryhighrCBVvalues (8.5) on PWI (C), & shows peculiar slow rise time intensity curve that crosses the baseline on late phase (D).MultivoxelMRspectroscopy demonstrates elevated lipids and choline within the lesion with absent NAA & Creatinine (E). Classical of a meningioma.

Export to PPT

Results

A total of 61 patients were included in the study and MRS data were obtained. The age range of patients was from 4 to 76 years. Majority of patients in the study belonged to the 21 to 30 years age group, constituting 29.5% of the cases (Chart 1 and ►Table 1). Out of 61, 31 patients were male and 30 were female patients (Chart 2 and ►Table 2).

Close

Chart 1

Age distribution of patients in the study.

Export to PPT

Close

Chart 2

Gender distribution of patients in the study.

Export to PPT

Most common presenting complaint was headache, seen in 65% of the patients, followed by seizures (51%). Symptoms of projectile vomiting were seen in 29% of the patients. Weakness and hemiparesis were the major complaints in 15% of the patients. Fever (8%) and dementia (5%) were seen in minority of patients only (Chart 3 and ►Table 3).

Close

Chart 3

Presenting complaints of patients with intracranial space-occupying lesions.

Export to PPT

The most common cause of intracranial space occupying lesions detected in the study was Gliomas, seen in 20 (32.7%) cases. Meningioma was present in 9 (14.7%) cases (►Figs. 1 and 2) Out of 20 cases of glioma, 7 were grade II gliomas (►Fig. 3), 4 were grade III glioma (►Fig. 4)and9weregrade IV gliomas (►Fig. 5). The next most common tumour was metastases (►Fig. 6), which was seen in 11 (18%) cases, pituitary macroadenoma in 6 (9.8%) cases, each of dysembryoplastic neuroepithelial tumour and trigeminal schwannoma in 4 (6.5%) cases, craniopharyngioma in 3 (4.9%) and each of acoustic schwannoma and medulloblastoma in 2 (3.2%)cases (Chart 4 and ►Table 4).

Close

Chart 4

Percentage distribution of brain tumors.

Export to PPT

Close

Fig. 2

Midline frontal extra-axial mass extending on either side of falx, appearing cellular due to reduced diffusivity (A), heterogenous with small internal cystic areas indistinct margins & significant vasogenic edema in left frontal white matter (B). It shows heterogenous post contrast enhancement with central areas of necrosis (C). PWI show significant hyperperfusion (D),MR spectroscopy (E) shows significantly elevated lipids, mildly elevated choline. Post operative histopathology s/o malignant meningioma (Rhabdoid subtype WHO grade-III).

Export to PPT

Close

Fig. 3

Afairlywelldefined intra-axial non- enhancing mass in left posterior frontal lone & subinsular cortex (A & B). It is appearing hypoperfused on rCBV maps of PWI (C). Multivoxel MR spectroscopy shows elevated choline without elevation of lipid lactate. (D). Biopsy confirmed – WHO grade II astrocytoma.

Export to PPT

Close

Fig. 4

An ill defined non-enhancing glial neoplasm is noted in left temporal lobe and insular cortex (A & B). It shows mildly hyper-perfused with an rCBV value of 3.164 (C). Metabolite Map showing choline containing areas within the lesion(red color in D)&Multivoxel MR spectroscopy showing significantly elevated choline & an inverted lactate peak with significant reduction of NAA & Creatinine (E) Biopsy reveals grade III oligoastrocytoma.

Export to PPT

Close

Fig. 5

Large heterogeneous signal intensity lesion is noted extending across the genu of corpus callosum in bilateral frontal lobes (A) with thick nodular peripheral enhancement and central areas of necrosis (B). The lesion appears significantly hyper-perfused on PWI (C) &exibits arCBV value of 6.24. metabolite map (D) and multivoxel MR spectroscopy (E) shows predominant lipid-lactate in the lesion corresponding to area of necrosis with mildly elevated choline. Biopsy confirmed WHO Grade IV glioblastoma.

Export to PPT

Close

Fig. 6

Large heterogeneous signal intensity right temporo-parietal mass is seen on T2WI (A), which shows irregular peripheral enhancement with areas of necrosis (B). rCBV map of PWI reveals hyper-perfusion along the periphery of the lesion (C).MultivoxelMRspectroscopy (D) from peritumoral edema) & metabolite map (E) shows elevated choline within peritumoral edema s/o infiltrative high grade glial neoplasm than metastases. Biopsy showed Metastasis.

Export to PPT

Means with standard deviation for each of Cho/Cr, Cho/NAA, and NAA/Cr were calculated for each type of tumors and results were tabulated (Chart 5 and ►Table 5). The mean Cho/Cr ratio was high in gliomas, metastasis, meningiomas, medulloblastoma, pituitary macroadenomas, and schwannomas, while it was normal in DNETs and craniopharyngiomas. The highest mean Cho/Cr value was seen in metastasis (3.58). The mean Cho/NAA ratio was high in all the lesions, with the highest values seen in the meningiomas (7.84) and gliomas (4.66). The mean NAA/Cr ratio was below normal in gliomas, metastasis, meningiomas, medulloblastoma, pituitary macro-adenomas, and acoustic schwannomas. The lowest values are seen in the meningiomas (0.53) and gliomas (0.67) (Chart 5). The mean NAA/Cr ratio was highest in trigeminal schwannomas (1.79) and craniopharyngiomas (1.71).

Close

Chart 5

Average metabolite ratios in the various brain tumors.

Export to PPT

Lactate was seen in 81% of metastases, 69% of all high-grade gliomas, 66% of craniopharyngioma, 50% of medulloblastomas, and 30% of meningiomas. Lipid peak was found in the high-grade gliomas, metastases, meningiomas, and medulloblastomas, particularly in grade IV gliomas (88.8%). Alanine was seen in 66.6% of all the meningiomas, but not in any other malignancy. There was absence of lactate and lipid resonance in pituitary macroadenomas, DNET, and schwannomas (Chart 6 and ►Table 6).

Close

Chart 6

Other metabolites detected in tumors.

Export to PPT

Means with standard deviation for each of Cho/Cr, Cho/NAA, and NAA/Cr were calculated for different grades of glioma and their values are listed in Chart 7 and ►Table 7.The mean Cho/Cr ratio was highest in grade IV gliomas (3.27) and lowest in grade II gliomas (1.48). The Cho/Cr and Cho/NAA ratios increase with increasing grade of malignancy, with the maximum mean Cho/Cr ratio (3.27) and Cho/NAA ratio (7.96) seen in the grade IV gliomas. The NAA/Cr ratio was lower in the higher grade gliomas than in the low-grade gliomas; lowest in grade IV gliomas (0.45) (Chart 8). Statistically significant differences in values of these metabolite ratios were found for different grades of gliomas. Statistically significant difference in values of Cho/NAA and NAA/Cr ratios was also found between grade IV gliomas and metastasis (Chart 9 and ►Table 8).

Close

Chart 7

Comparison of metabolite ratios in various grades of gliomas.

Export to PPT

Close

Chart 8

Comparison of metabolite ratios in the three grades of gliomas.

Export to PPT

Close

Chart 9

Comparison of metabolite ratios in gliomas and metastatic lesions.

Export to PPT

Discussion

MRS is a new and emerging imaging modality which offers a level of tissue characterization that can match histological and biochemical diagnosis. Since its introduction, various studies have been done and are being performed to determine its applications in the field of diagnostic radiology. This study was performed in the Department of Radiodiagnosis in a tertiary care hospital in India. We found that MRS has potential application in the diagnosis and characterization of various intracranial space-occupying tumors. Besides differentiation of benign from malignant lesions, it also enables the grading of malignancies. It not only compliments the information available by conventional imaging, but can also predict the prognosis and help in management of these lesions.

In the present study, most of the patients presenting with brain tumors were in the 21 to 30 years age group. The majority of patients were males, forming 50.8% of the study population. We found the predominant cause of intracranial space-occupying lesions to be tumors, gliomas being the most prevalent. The predominant presenting symptom was headache, seen in 65% of the patients, followed by seizures (33%).

Single voxel proton MRS, using STEAM (STimulated Echo Acquisition Mode) or PRESS (Point RESolved Spectroscopy), has been performed to study brain lesions in clinical settings.^10,11 The utility of these sequences in the differentiation of lesions has been assessed by various workers in the past. Proton MRS is useful only when the voxel of interest is taken from well within the lesion. For very small lesions, the possibility of partial averaging from surrounding tissues and hence obtaining a misleading spectrum is a limiting factor.¹² The presence of prominent areas of hemorrhage, calcification or sometimes necrosis, or the peripheral location of the lesion close to calvarium or cerebrospinal fluid also result in poor spectrum, which does not serve any diagnostic purpose. The study was done from the most homogenous part of the lesion in case of predominantly solid lesions. For cystic lesions, the cystic or necrotic parts were analyzed separately.

In most of the studies in the past on intracranial lesions a short echo time sequence (30 or 20 ms) was used along with a long echo time (270 ms) sequence,¹³ or two long echo time sequences were used employing PRESS or STEAM for localization. Some studies^10-12 have also used SE 135 and 270 ms sequences for better signal-to-noise ratio and a smooth spectrum devoid of contributions from short T2 metabolites, such as inositols and lipids. In the present study, a combination of STEAM 30 ms and 135 ms sequences was used. The 135 ms sequence was used to obtain the metabolite ratios. The short TE 30 ms sequence was used to observe the metabolites with short T2 such as lipids and inositol.

The results of the present study reveal that the spectral pattern of tumors is markedly different from the normal brain. In general, the results are consistent with the earlier studies on brain tumors.^10,12,14 Most of the tumors revealed an elevated choline (Cho) peak along with a decrease in NAA resonance.

All malignant tumors like gliomas, metastasis, and meningiomas were characterized by increased Cho, decreased NAA and creatinine (Cr) along with the presence of lactate (Lac), lipid (Lip), or lipid resonances in all the cases. Increased Cho has been observed in most brain tumors, attributed to the increased membrane turnover and cell proliferation.¹⁴ Presence of Cho was an essential feature of all gliomas, even when cystic; though the choline levels in cystic neoplasms were lower than in the solid lesions. The latter appears to be due to the degraded pool of Cho in the cystic fluid with no active membrane turnover.

In the in vivo proton MRS, the peak at 3.22 ppm is composed mainly of Cho, PC (phosphatidylcholine), and GPC (glycerol-phosphatidylcholine).¹⁵ The choline pool is expected to increase in tumors because of an increased choline esterase activity which catalyzes the first step of phospholipid biosynthesis. On the other hand, increased PDEs (phosphodiesters) containing GPC may be an indicator of the necrotic fraction in tumors as a consequence of phospholipid degradation. Thus, the increased Cho may either indicate increased cell proliferation (increased PC) or a necrotic process (increased GPC).^14,15

NAA is predominantly located in neurons. It has been shown that NMR-visible NAA peak at 2.02 ppm arises mainly from NAA and only a small component of it is from other N-acetyl-containing compounds. Presence of NAA as seen in most of the tumors has been attributed to the difference in the cellular composition and nature of the tumor.^15,16 Higher grade tumors, especially with tissue necrosis, naturally have lower NAA levels due to neuronal loss or replacement.

Grading of gliomas has been done on the basis of NAA/Cho,¹⁰ Cho/Cr, and NAA/Cr ratios. NAA/Cr and Cho/ Cr ratios have shown a consistency in predicting tumor grade,^10,11 although Kugel et al have not found any signifi-cant difference in the metabolite ratios between various grades of tumors.¹² In the present study, a significant difference was seen in the Cho/NAA ratios of all three grades of gliomas. Increasing Cho/Cr and decreasing NAA/Cr values were also seen with increasing grade of malignancy. The NAA/Cr levels in grade IV gliomas were very low, NAA peak being absent in some of the GBMs. However, NAA/Cr ratios in low-grade gliomas and anaplastic group showed overlapping values. The Cho/Cr levels also showed overlapping values in the higher grade gliomas. From the present study, it can be concluded that Cho/Cr, Cho/NAA, and NAA/Cr ratios can be used in the grading of malignancies as suggested by Sutton et al and Ott et al.^10,11 Of these ratios, Cho/NAA appears to be the most significant in determining tumor grade.

Presence of lactate has been used as criteria for grading of astrocytomas. Lactate in tumor cells is believed to be formed by conversion from pyruvate, which accumulates as a result of the increased glycolytic activity in the tumor cells, due to the decrease in the tricarboxylic acid cycle activity.¹⁶ Lactate may also be formed by anaerobic glycolysis in tumors with hypoxia.¹⁷ High-grade tumors showed the presence of lac-tate as compared with low-grade tumors which did not show its presence. Hence the presence of lactate appears to correlate significantly with the grade of malignancy, i.e., higher levels of lactate correlated with higher grade tumors. Similar results were obtained by Fulham et al¹⁷ in their study of brain tumors.

Lipid resonances have been observed in high-grade glioma in vivo studies using different echo times by Gupta et al.¹³ Ex vivo studies have confirmed the presence of mobile lipids in necrotic and viable tissues of high-grade gliomas and have suggested a correlation between the presence of lipids and the degree of necrosis as seen on histology. Few authors have concluded that the presence of lipid signals may be useful in the discrimination of low- and high-grade tumors. Necrosis distinguishes the high- from low-grade tumors in majority of the histological classification systems and correlates inversely with survival time. In the present study, lipid signals were seen in tumors with and without visible necrosis. None of the low-grade gliomas showed presence of lipids, while it was present in all the cases of grade IV and most of the grade III gliomas, all of which showed varying levels of necrosis on histological examination. Hence it can be concluded that lipid resonances indicate necrosis, and presence of lipid correlates with higher degrees of malignancy.

From the present study, it is evident that Cho/NAA and Cho/Cr ratios can be used to determine the histological grade of malignancy. But when used alone, they may prove inconclusive, especially in the case of cystic or necrotic gliomas. However, when combined with the presence or absence of lipid signals, the grade of malignancy can be predicted with greater accuracy.^13,14,18

Reduced or absent NAA peaks along with variable signal intensities from Cho, Cr, Lac, and Lip have been observed in metastasis.^13,17,19 As metastasis often contain nonneuronal tissue, a low or absent NAA peak is expected. However, in most cases a NAA peak is obtained. This may be explained by the nature of the tumor, that is, NAA may be seen if the tumor grows invasively or is unable to displace all the neurons or due to partial volume from adjoining brain tissue. Kugel et al have observed increased lipid signals in metastasis.¹² Most studies have failed to demonstrate any spectral variations in the different histological types of metastasis, and have not found any differences in metastasis versus glioblastomas or abscesses based on MRS. In concordance with these observations, the present study also revealed spectral patterns in metastasis similar to and hence indifferentiable from high-grade gliomas.

The presence of alanine (Ala) is specific for meningiomas¹⁴; meningeal cells are known to have very high amounts of alanine as compared with the neuronal and glial cells.¹⁹ Kugel et al however found Ala in only 62.5% of the meningiomas examined.¹² In the present study, a prominent Ala peak was seen in six out of nine (66.6%) cases of meningiomas, along with strong resonance of Cho. The NAA and Cr levels were low, giving a high Cho/NAA and increased Cho/Cr ratios. Specificity of alanine for meningiomas was confirmed.

The correlation between the tumor grade and prognosis is well established. Histological classification from biopsy specimens is the reliable and standard method for this purpose.

In the present study, 6.5% of the patients had been diagnosed with DNETs. These patients were young adults presenting with chronic intractable seizures. DNETs are low-grade (World Health Organization Grade I) cortical neuroepithelial tumors which usually present with unresponsive seizure in young adults. They appear as T1 hypointense and T2 hyperin-tense lesions in the temporo-parietal cortex without signifi-cant mass effect. MRS findings on DNETs were described by Lee et al.²⁰ Both reported that the spectrum from DNETwas similar to that of normal brain with no increase in Cho or decrease in NAA values. We found normal Cho/Cr, Cho/NAA, and NAA/Cr values in all cases of DNETs, with presence of lactate in two cases, probably related to the cystic component. These findings are consistent with earlier reports and indicate a useful role of spectroscopy in confirming the diagnosis, ruling out a more malignant pathology.

Proton MRS has been used to predict the presence of histopathological features important for brain tumor diagnosis that can be used to plan stereotaxic biopsies and selective tumor resections.²¹ Dowling et al found that areas of abnormal spectrum indicating higher degree of malignancy within a lesion correlated with regions of viable cancer.²¹ This may be valuable for guiding surgical biopsies and focal therapy.

Thus, the present study has shown that MRS has potential applications in grading of malignancies; it complements the information obtained from conventional MRI and contrast studies, proving particularly useful when these studies are inconclusive.

Conclusion

A total of 61 patients with brain tumors detected by MRI were investigated with proton MRS, using single voxel and multivoxel techniques. The most common age group of intracranial space-occupying lesions is 21 to 30 years. The most common presenting complaint in these patients is headache, followed by seizures. The most common intracranial tumors were gliomas. With conventional MRI, many of these lesions, such as necrotic tumors and infections, showed similar and often confusing imaging appearances. After MRS, the following conclusions were drawn:

• Grading of gliomas could be done reliably on the basis of Cho/Cr and Cho/NAA ratios. However, in certain cases these ratios were not sufficient to predict grade, and presence of lipids was a differentiating feature for the tumor grade in these lesions.

• It was not possible to differentiate the metastasis from high-grade gliomas accurately as both lesions showed overlapping values of metabolite ratios. The spectrum from metastasis shows high choline, low NAA, and often the presence of lipids similar to that from high-grade gliomas.

• Presence of alanine in 66% of meningiomas makes it a useful marker for these tumors.

• DNETs show normal spectrum, a characteristic that enables its differentiation from more malignant lesions.

• On multivoxel spectroscopy, the regions of high choline and Cho/NAA ratios correlated with a higher degree of malignancy on histopathology.

• The postsurgery and radiotherapy patients with recurrent lesions have high Cho/NAA levels, allowing differentiation from the radiation-induced changes (►Fig. 7).

Close

Fig. 7

Post operative & post RT temporo-parietal GBM, multivoxel MR spectroscopy shows elevated lipids as well as choline within the lesion (A & B). PWI rCBV map super-imposed on corresponding T1WI (C). Shows hypo-perfused areas superficially along anterior and posterior aspect of lesion (red arrows) corresponding with areas of radiation necrosis admixed with relatively hyper-perfused areas (yellow arrow) in the middle deep portion of the lesion suggesting recurrence.

Export to PPT

Hence, it is concluded that MRS is an integral tool in evaluation and helps in better tissue characterization of different brain tumors. It has a complimentary role to MRI not only in providing a better diagnosis, but also in directing treatment and follow-up of such lesions.