Neuroradiology (2002) 44: 825–829 DOI 10.1007/s00234-002-0821-z

J.K. Smith A. Londono M. Castillo L. Kwock

Received: 28 March 2001 Accepted: 2 June 2002 Published online: 10 August 2002
Springer-Verlag 2002

J.K. Smith Æ A. Londono M. Castillo (&) Æ L. Kwock CB# 7510, Department of Radiology, University of North Carolina School of Medicine, Chapel Hill, NC 27599-7510, USA E-mail: [email protected] Tel.: +1-919-9663087 Fax: +1-919-9661994

DIAGNOSTIC NEURORADIOLOGY

Proton magnetic resonance spectroscopy of brain-stem lesions

Abstract The imaging findings of brain-stem lesions are often nonspecific and histological diagnosis is limited because of fear of complications associated with biopsy. A noninvasive method for tissue characterization is therefore highly desirable. We undertook a review of proton magnetic resonance spectroscopy (MRS) of patients with solitary brain-stem lesions to determine if MRS could characterize them. We carried out single- or multivoxel proton MRS using long echo times (135 or 270 ms) on 34 patients with solitary brain-stem lesions. We analyzed the following peaks: choline (Cho), creatine (Cr), N-acetylaspartate (NAA), and lipids/lactate (Lip) and calculated peak height ratios for Cho/Cr, NAA/Cr and Lip/Cr. The results were compared with histology in nine patients

Introduction Recently, proton magnetic resonance spectroscopy (MRS) has assumed a role in imaging of brain lesions [1, 2, 3, 4]. The information it provides complements that obtained from MRI and often provides greater tissue characterization. Proton MRS can be used to distinguish between normal and abnormal tissues, aid in the grading of brain tumors, and may enable the distinction between tumors and inflammatory and infectious processes [4]. Technically, proton MRS of lesions in the cerebral and cerebellar hemispheres is relatively easy. Near the base of skull, proton MRS is technically difficult due to the

and with the presumptive diagnosis in 25. We also performed singlevoxel proton MRS on the brain stem of five normal volunteers. There were differences in all ratios between controls and the patients with neoplastic and non-neoplastic lesions: Cho/Cr was low in non-neoplastic and high in neoplastic lesions (control: 1.8±0.1; non-neoplastic: 1.4±0.2; neoplastic: 2.0±0.2); NAA/Cr was low in non-neoplastic, and lower in neoplastic lesions (control: 2.3±0.1; non-neoplastic: 1.4±0.2; neoplastic: 1.2±0.1), and Lip/Cr was elevated in both neoplastic and non-neoplastic lesions (control: 0.04±0.02; nonneoplastic: 1.9±0.7; neoplastic:1.9±0.7). Keywords Brain stem Æ Magnetic resonance spectroscopy

smallness of some structures and magnetic susceptibility artifacts from adjacent bone, fat, and air interfaces. We undertook a review of the proton MRS studies on 34 patients who, during their initial clinical presentation, showed solitary lesions in the brain stem. We compared the results with the histological or presumed diagnoses.

Materials and methods We reviewed 34 patients. All spectroscopic studies were obtained one day after contrast-enhanced MRI. Axial T2-weighted images were repeated and used to guide voxel placement, which was also influenced by the position of any areas of contrast enhancement.

826

We performed single-voxel or 2-dimensional chemical-shift proton spectroscopic imaging, using voxels of 0.5·0.5·0.5, 1.0·1.0·1.0, 1.5·1.5·1.5 or 2.0·2.0·2.0 cm3, depending on the size of the lesion. All lesions occupied at least 80% of the voxel. Voxel localization was accomplished after water suppression using point-resolved spectroscopy (PRESS). All studies were obtained with a long TE (135 or 270 ms), TR 1500–1600 ms, and 128–256 signal averages. Using a similar technique and voxel size of 1.0·1.0·1.0 cm3, we also studied the brain stem in five normal volunteers. Because we used a long TE, only the following peaks were analyzed: choline (Cho) assigned at 3.2 ppm, creatine (Cr) at 3.0 ppm, N-acetyl aspartate (NAA) at 2.0 ppm, and lipids/lactate (Lip) at 1.0–1.6 ppm. Since the studies were obtained over a long period, quantitation was not possible in all of them and we analyzed peak height ratios by measuring the highest point on each peak and normalizing to that of Cr.

Histological proof of the nature of the lesion was available in nine patients. In those in whom histology was not available a presumptive diagnosis was made based on clinical and laboratory findings and follow-up MRI. Comparison between group means was performed using analysis of variance, with statistical differences further analyzed by post hoc pairwise comparison based on the Tukey method. Since the data may not be normally distributed, the group differences were confirmed with a non-parametric Kruskal-Wallis test.

Results Key information on individual subjects is presented in Table 1. Mean results for peak height ratios of Cho/Cr,

Table 1. Key data on subjects studied Age (years)

Sex

Control subjects 22 M 23 M 40 F 42 F 50 M

Site

TR (ms)

Pathology

Metabolite ratios Cho/Cr

NAA/Cr

Lip/Cr

135 135 135 135 135

1500 1500 1600 1500 1500

Normal Normal Normal Normal Normal

1.94 1.67 1.67 1.54 2.14

2.69 2.17 2.33 1.97 2.43

0.12 0.00 0.00 0.06 0.00

Non-neoplastic lesions 16 M Pons 25 M Brain stem 41 F Pons 59 F Brain stem 9 M Midbrain 26 F Pons 38 M Pons 41 M Pons 71 F Pons 26 M Pons 43 M Brain stem 62 F Brain stem 63 M Pons 34 F Midbrain

270 270 270 270 270 270 270 270 135 270 135 135 135 135

1600 1500 1500 1600 1500 1500 1500 1500 1600 1500 1500 1500 1500 1600

Fungal mass (biopsy) Gliosis (presumeda) Gliosis (presumed) Gliosis (presumed) Hamartoma (presumed) Inflammatory (presumed) Inflammatory (presumed) Inflammatory (presumed) Inflammatory (presumed) Multiple sclerosis (presumed) Multiple sclerosis (presumed) Myelinolysis (presumed) Myelinolysis (presumed) Necrosis (biopsy)

1.00 0.00 1.13 1.50 1.32 1.48 2.00 1.22 1.40 2.60 1.16 1.25 1.33 1.63

0.06 0.50 1.50 1.13 1.08 1.90 1.69 1.30 2.20 2.50 2.11 1.25 1.11 1.50

2.50 9.00 0.13 1.50 0.20 0.00 0.15 0.09 0.40 0.05 3.21 5.13 0.22 4.00

Neoplasms 4 33 49 20 5 5 6 6 6 7 10 18 27 68 72 36 39 65 1 70

270 270 270 270 270 270 270 270 270 270 270 270 270 135 270 270 135 270 135 270

1500 1500 1500 1500 1600 1500 1500 1500 1600 1600 1500 1500 1500 1500 1600 1500 1600 1600 1500 1500

Astrocytoma (biopsy) Astrocytoma (biopsy) Astrocytoma (biopsy) Ganglioglioma (biopsy) Glioma (biopsy) Glioma (presumed) Glioma (presumed) Glioma (presumed) Glioma (presumed) Glioma (presumed) Glioma (presumed) Glioma (presumed) Glioma (presumed) Metastatic lung cancer (presumed) Metastatic lung cancer (presumed) Lymphoma (biopsy) Metastatic melanoma (presumed) Melanoma Met (presumed) Primitive neuro-ectodermal tumor (biopsy) Transitional-cell metastasis (presumed)

3.27 2.00 1.54 1.60 4.38 1.78 0.36 1.80 2.93 1.31 1.50 2.61 1.77 0.91 1.50 2.00 2.00 2.31 3.50 1.40

1.87 0.40 0.92 1.60 1.13 1.22 1.60 0.44 0.89 1.31 1.00 0.70 1.89 1.18 1.50 1.87 1.23 1.92 0.50 0.50

4.13 8.00 0.77 1.00 1.88 0.56 3.00 0.49 1.29 0.50 1.67 0.00 0.65 0.45 4.00 0.00 4.07 0.15 0.40 4.05

a

M M F F F F M M F F F M M M F F M M M M

Pons Pons Pons Pons Pons

TE (ms)

Pons Brain stem Pons Pons Pons Pons Pons Pons Pons Pons Pons Midbrain Midbrain Pons Pons Midbrain Pons Pons Pons Pons

based on clinical and laboratory findings and follow up imaging

827

Table 2. Mean metabolite ratios (mean±standard error of the mean) by group. Abbreviations as in Table 1 Group

Cho/Cr

NAA/Cr

Lip/Cr

Control Non-neoplastic Neoplastic

1.8±0.1 1.4±0.2 2.0±0.2

2.3±0.1 1.4±0.2 1.2±0.1

0.04±0.02 1.9±0.7 1.9±0.7

NAA/Cr, and Lip/Cr within each group are presented in Table 2. Individual peak height ratios are presented in Fig. 1. From 1996 to 2002, 34 patients (19 males, 15 females, age 1–72 years) presented with solitary lesions in the brain stem. Of the nine who underwent biopsy, seven had neoplasms: three low-grade astrocytomas, one high-grade glioma, one ganglioglioma, one primitive neuroectodermal tumor, and one lymphoma, and two had non-neoplastic lesions: a fungal mass and radiation necrosis. We assigned presumptive diagnoses to 25 lesions. These included 12 presumed non-neoplastic lesions and 13 presumed neoplasms. The former consisted of four inflammatory lesions which resolved with time or after antibiotic or antiviral therapy. There were two lesions in patients who later developed typical clinical and imaging findings of multiple sclerosis. Two patients had typical lesions that developed after correction of serum osmolyte abnormalities. In one patient found to have neurofibromatosis a lesion remained stable and was presumed to be a hamartoma. Three lesions followed surgery, trauma or stroke and remained stable for two or more years and were presumed to represent gliosis. Presumed neoplasms were diagnosed in eight patients with typical imaging findings of a mass that progressed on follow-up studies and was presumed to be a glioma. There were five patients who developed diffuse brain metastases (two from melanoma, two from lung cancer, and one from transitional-cell cancer of the bladder) on follow-up and we assumed that the initial solitary brainstem lesion was also a metastasis. MRS of the brain stem was of good quality in all normal volunteers. Analysis showed that Cho/Cr was 1.8±0.1, somewhat higher than reported in the cerebral hemispheres (Fig. 2). At most trace levels of lipids were observed. On average, non-neoplastic lesions showed lower levels of Cho (Cho/Cr 1.4±0.2) and NAA (NAA/Cr 1.4±0.2) and higher levels of Lip (Lip/Cr 1.9±0.7) than normal subjects. Any one ratio in individual lesions c

Fig. 1a–d. Metabolite peak height ratios for individual lesions, compared to normal control. a choline/creatine b N-acetylaspartate (NAA)/creatine c lipids/creatine d choline/creatine vs NAA/ creatine ratios. If both NAA and choline levels are considered, normal subjects can be separated from all those with neoplastic and most of those with non-neoplastic lesions (box), although there is significant overlap between neoplastic and non-neoplastic lesions

828

Fig. 2. Normal volunteer proton spectra (MRS) show that the choline (Cho)/creatine (Cr) peak ratio is slightly higher than usual in the cerebral hemispheres. NAA/Cr ratio is similar in the cerebral hemispheres

showed considerable overlap with both normal controls and neoplasms (Fig. 1). As a group, neoplastic lesions showed higher levels of Cho (Cho/Cr 2.0±0.2) and lower levels of NAA (NAA/ Cr 1.2±0.1) than controls or non-neoplastic lesions, and higher levels of Lip (Lip/Cr 1.9±0.7) than controls. Again, ratios in individual lesions showed considerable overlap with both normal controls and neoplasms. With both high Cho and low NAA, individual neoplasms could be separated from controls, but there was still overlap with individual non-neoplastic lesions (Fig. 1D).

Discussion In our institution, many brain-stem lesions are treated based on clinical and laboratory findings, without histological proof. A noninvasive test that could improve their characterization would be highly desirable. If such a test suggested a tumor, biopsy might be more aggressively pursued, so as to choose the most appropriate therapy. To the best of our knowledge, no series of patients with solitary brain-stem lesions studied by proton MRS has been described. Proton MRS in patients with neoplasms showed high Cho and low NAA, findings compatible with neoplasia and reported previously for tumors in other parts of the brain [1] (Fig. 3). The high Cho peak is thought to be

Fig. 3. A neoplasm: brain-stem glioma. MRS shows marked elevation of Cho with respect to Cr; NAA is reduced, and no lipids are seen

related to increased cell proliferation and/or cell membrane turnover [1, 2, 3]. The low NAA may be due to replacement and/or dysfunction of the neurons and axons. Lipids are usually found in untreated high-grade tumors with necrosis or in previously treated tumors [5]. Lymphoma is known to show increased lipids/lactate [6]. The brain-stem lymphoma reported here did not, and the reason for this is not clear. Based on our limited experience, we believe that proton MRS adequately reflects the neoplastic nature of many brain-stem lesions and may aid presumptive diagnosis. MRS was also useful in characterizing the non-neoplastic lesions. In most, Cho levels were normal or slightly low, NAA was reduced, and lipids/lactate were frequently present. In both patients with brain-stem plaques who later developed multiple sclerosis, the spectra also showed resonances at 2.0–2.6 ppm, presumed to represent the marker peaks known to be present in demyelinating plaques [7] (Fig. 4). Thus, we believe that proton MRS was helpful in suggesting the non-neoplastic nature of these lesions. Such findings may help dictate less aggressive investigation and management of some brain-stem lesions. In the patients with pontine osmotic myelinolysis, MRS showed low Cho, low NAA, and Lip. One textbook shows three cases of pontine osmotic myelinolysis but two of the spectroscopy studies were obtained in the occipital gray matter and not in the pons [8]; the results were nonspecific and thought to be related to hypo-osmolar encephalopathy. One patient did have spectroscopy of the pontine lesion,

829

Fig. 4. A non-neoplastic lesion: a brain-stem plaque of multiple sclerosis. MRS shows no significant changes in Cho, Cr, or NAA. There is a large lipid (L) resonance and marker peaks(*). This spectrum suggests a non-neoplastic lesion with demyelination

showing low NAA, normal Cho, low myoinositol, and high Lip. The fact that this type of lesion is characterized by numerous lipid-laden macrophages may explain the large Lip peak in that patient and in one of ours. It is conceivable that the presence of numerous macrophages may also lead to a mild elevation of the Cho peak [9]. We believe that the lipids/lactate seen in some of the lesions in our patients are not due to contamination.

Although extraneous lipid contamination may occur when proton MRS is obtained close to the base of the skull, contamination does not occur if one uses a technique similar to ours, even when the voxel is close to fatcontaining structures [10]. In addition, not all the lesions we studied demonstrated the presence of lipids. If lipids were due to contamination, we would have expected to see them in most studies. There was considerable overlap between certain metabolite ratios in normal controls and individual nonneoplastic and neoplastic lesions. When typical changes in Cho and NAA are considered, individual neoplasms could be separated from controls, but there was still overlap between individual non-neoplastic lesions and between non-neoplastic lesions and controls (Fig. 1D). This emphasizes the need to consider the entire spectrum, and not rely on any one metabolite level. One limitation of our study is that only peak height measurements were obtained. Peak area measurements would more accurately reflect true metabolite concentrations. Unfortunately, these were not obtained in the earlier patients studied. Therefore, for comparing between groups, we used peak height ratios. In some patients in both the neoplastic and non-neoplastic groups, the levels of Cho, Cr and NAA were very low, with large Lip peaks; these were presumably areas of necrosis. In such cases, the peak height ratios may be significantly affected by noise. However, to avoid introducing bias, we included these patients in the group comparisons. Another limitation is that we used only long echo times. At the time some of these data were collected we were unable to use short echo times, which may allow one to detect metabolites with a short relaxation time and aid in characterizing a lesion. Despite these shortcomings, we believe our data indicate that proton MRS may be useful in investigation of brain-stem lesions.

References 1. Tien RD, Lai PH, Smith JS, Laseyras F. Single-voxel proton brain spectroscopy exam (PROBE/SV) in patients with primary brain tumors. AJR 1996; 167: 201–209 2. Cecil KM. Jones BV. Magnetic resonance spectroscopy of the pediatric brain. Topics in Magnetic Resonance Imaging. 12(6):435–52, 2001 Dec. 3. Tzika AA, Zarifi MK, Goumnerova L, et al (2002) Neuroimaging in pediatric brain tumors: Gd-DTPA-enhanced, hemodynamic, and diffusion MR imaging compared with MR spectroscopic imaging. AJNR 23:322–333

4. Rand SD, Prost R, Haughton V, et al (1997) Accuracy of single-voxel proton MR spectroscopy in distinguishing neoplastic from non-neoplastic brain lesions. AJNR 18: 1695–1704 5. Taylor JS, Langston JW, Reddick WE, et al (1996) Clinical value of proton magnetic resonance spectroscopy for differentiating recurrent or residual brain tumor from delayed cerebral necrosis. Int J Radiation Onc Biol Phys 36: 1251–1261 6. Bizzi A, Movsas B, Tedeschi G, et al (1995) Response of non-Hodgkin lymphoma to radiation therapy: early and long term assessment with H-1 MR spectroscopic imaging. Radiology 194: 271–276

7. Bitsch A, Bruhn H, Vougiokas V, et al (1999) Inflammatory CNS demyelination: histopathologic correlation with in vivo quantitative proton MR spectroscopy. AJNR 20: 1619–1627 8. Danielsen ER, Ross B (1999) Magnetic resonance spectroscopy diagnosis of neurological diseases. Dekker, New York, pp 80–82, 118, 119 9. Krouwer HGJ, Kim TA, Rand SD, et al. (1998) Single-voxel proton MR spectroscopy of nonneoplastic brain lesions suggestive of a neoplasm. AJNR 19: 1695–1703 10. Kwock L, Brown MA, Castillo M (1997) Extraneous lipid contamination in single-volume proton MR spectroscopy: phantom and human studies. AJNR 18: 1349–1357