- Email: [email protected]
Proton magnetic resonance spectroscopy of brain tumors correlated with pathology1
Original Investigations
Proton Magnetic Resonance Spectroscopy of Brain Tumors Correlated With Pathology1 Alvaro Magalhaes, MD, William Godfrey, MS, Yimin Shen, PhD, Jiani Hu, PhD, Wilbur Smith, MD
Rationale and Objectives. Evaluate proton magnetic resonance spectroscopy (1H-MRS) for assessing and grading brain tumors. Materials and Methods. The research was done at Detroit Medical Center in a 1.5-T Siemens MR magnet using singlevoxel or multivoxel MRS. This study consisted of 27 patients: 10 females and 17 males ages 22– 83 years (average age 43.8). The data were recorded for three peaks—N-acetyl aspartate (NAA), choline (Cho), creatine (Cr)—which were used to calculate the ratios Cho/NAA and Cho/Cr. Results. Abnormal spectra were seen in 25 patients and normal spectra in 2. In 16 patients with brain astrocytoma of various grades, the pathology grading was correlated with Cho/NAA and Cho/Cr. These values were 6.53 and 3.35 for nine patients with Grade 4 astrocytoma; 1.85 and 1.62 for three patients with Grade 3 astrocytoma; 2.21 and 1.50 for three patients with Grade 2 astrocytoma; and 1.45 and 1.49 for one patient with Grade 1 astrocytoma. The remaining nine patients with abnormal spectra were also correlated with pathology. Conclusion. MRS ratios can be used to differentiate malignant and nonmalignant lesions from normal brain tissue. In general, high-grade astrocytoma have higher Cho/NAA and Cho/Cr ratios compared with low-grade astrocytoma. Key Words. Magnetic resonance spectroscopy; brain tumor; astrocytoma. ©
AUR, 2005
The grading of a tumor has important implication for clinical management. The grade of the tumor defines to a substantial extent survival probability and course of therapy. The gold standard of tumor grading is histopathologic diagnosis requiring a biopsy during an open or stereotactic neurosurgical procedure. There is strong evidence that proton magnetic resonance spectroscopy (1HMRS) signal patterns can serve to identify tumor type and grade (1–3).
Acad Radiol 2005; 12:51–57 1 From the Department of Radiology, School of Medicine, 3990 John R, Wayne State University, Detroit Medical Center, Detroit, MI 48201-2097 (A.M., W.G., Y.S., J.H., W.S.). Received March 17, 2004; revised April 12, 2004; accepted October 23, 2004. W.G. received funding in the radiology summer externship program from the Department of Radiology and the School of Medicine of Wayne State University. Address correspondence to A.M. e-mail: [email protected]
© AUR, 2005 doi:10.1016/j.acra.2004.10.057
1H-MRS
uses radiofrequency excitation in the presence of a magnetic field to obtain signals from tissue chemicals other than water. The resonance frequency position of each peak is dependent on the chemical environment of that nucleus. It can be used to identify metabolites. The area under the peak may be calculated and yield relative measurements of the concentration of protons, when relaxation effects can be ignored—such as zero echo time. The width of the peak at half-height is proportional to 1/T2 when the magnetic field is perfectly homogenous. Many metabolites can be measured by 1H-MRS. Those of concern in this study are well differentiated in 1H-MRS and differ between neuronal and neoplastic brain tissue. Certain metabolites of neurons and abnormal brain tissue that most research is focused on are N-acetyl aspartate (NAA), choline (Cho), creatine (Cr), and lipids. In neoplastic tissue compared with normal neuronal tissue, there is a decrease in glucose, NAA, aspartate and Cr; how-
51
MAGALHAES ET AL
ever, alanine, hypotaurine, phosphoethanolamine, and Cho all rise (4). The question is if 1H-MRS is able to add new information to the high-quality anatomic data provided by conventional magnetic resonance imaging (MRI) techniques and whether this information could be used to replace or complement biopsy for the diagnosis and clinical management of patients.
Academic Radiology, Vol 12, No 1, January 2005
Twenty patients presented with a brain tumor; 16 patients of these 20 patients were diagnosed with astrocytoma that was graded according to WHO. Glioblastoma multiforme and gliosarcoma were Grade 4. Anaplastic oligoastrocytoma and anaplastic oligodendroglioma were Grade 3. Oligodendroglioma was Grade 2. Pilocytic astrocytoma was Grade 1. The 1H-MRS data of Cho/NAA and Cho/Cr were analyzed and compared with each tumor grade of the 15 patients with gliomas. Normal 1H-MRS spectra were obtained in two patients; no biopsy was performed on these patients.
MATERIALS AND METHODS Subjects This study consisted of 27 patients:, 10 females and 17 males ages 22– 83 years (average 43.8). All 27 patients presented with central nervous system symptoms and were assessed with brain MRI that demonstrated a suspicious brain tumor. Because of the brain lesion, the 27 patients were evaluated with 1H-MRS to confirm tumor diagnosis; in the case of astrocytomas, an attempt was made to grade the tumor. Lesions suspected of being tumors in 25 patients were then stereotactically biopsied. Twenty patients were diagnosed with a brain tumor and 16 of these with astrocytomas were graded according to the World Health Organization (WHO) classification. Five other patients presented with brain lesions, but had biopsies negative for malignancy. The remaining two patients with normal 1H-MRS results were not biopsied. Methods The 1H -MRS was done at a 1.5-T whole-body Siemens (Erlangen, Germany) imager using standard clinical singlevoxel or multivoxel 1H-MRS pulse sequences and postprocess protocol. The pulse sequence used was point-resolvedsingle-volume-spectroscopy (PRESS) with repetition time (TR) ⫽ 1,500 milliseconds and echo time (TE) ⫽ 135 milliseconds. Voxel size in most cases was 1.5 ⫻ 1.5 ⫻ 1.5 cm3 for single voxel and was 1 ⫻ 1 ⫻ 1.5 cm3 for multivoxel. Matrix size for chemical shift imaging (CSI) was 16 ⫻ 16 ⫻ 1 and field of view was 160 ⫻ 160 ⫻ 15 mm3. Multivoxel technique in 16 patients and single-voxel technique was done in 9 patients. Metabolite resonance intensities were determined from fitting peak areas using Siemens online software. Metabolite signals were expressed as rations. The data were recorded for three peaks: NAA, Cho, and Cr. These values were then placed into ratios Cho/NAA and Cho/Cr. In all 25 cases with abnormal 1H-MRS results, ratios Cho/NAA and Cho/Cr were analyzed and compared with histopathologic diagnoses.
52
RESULTS MRI and 1H-MRS was obtained in 27 patients and pathology examination was obtained in 25 patients. Seventeen of the 25 patients biopsied with astrocytoma were classified by WHO criteria: 9 patients with Grade 4 tumors, 3 patients with Grade 3 tumors, 3 patients with Grade 2 tumors, and 1 patient with Grade 1 tumor. The remaining nine patients with lesions were: one patient with medulloblastoma, two patients with adenocarcinomas, two patients with abnormal spectra but without malignancy, one patient with fibrous histiocytoma, one patient with hematoma, one patient with granuloma, and one patient with astrogliosis and inflammation. Two patients with normal 1H-MRS spectra did not have samples submitted for biopsy. H-MRS spectra of a normal brain are shown in Figure 1. The histopathologic grade was obtained after serial stereotactic and open surgery. The histopathologic diagnoses were made according to WHO criteria. The histopathologic classification was then compared with 1H-MRS data. NAA, Cho, and Cr peaks were obtained from 1H-MRS studies and placed in the following metabolic ratios: Cho/NAA and Cho/Cr. 1H-MRS and pathology data with tumor grade were correlated. Nine patients had Grade 4 astrocytoma; eight with glioblastoma multiforme and one with gliosarcoma. The average metabolic ratios of Cho/NAA and Cho/Cr were 6.53 and 3.35, respectively. 1H-MRS spectra of a Grade 4 glioma is shown in Fig. 2 Three patients had Grade 3 tumors: two with an anaplastic oligoastrocytoma and one with anaplastic oligodendroglioma. Their average metabolic ratios of Cho/NAA and Cho/Cr were 1.85 and 1.62, respectively. Three patients had oligodendroglioma tumors Grade 2; their metabolic ratios of Cho/NAA and Cho/Cr were 2.21 and 1.50, respectively. 1H-MRS spectra of a Grade 2 astrocytoma are shown in Fig. 3. One patient had an astrocytoma Grade 1 with metabolic ratios of
Academic Radiology, Vol 12, No 1, January 2005
PROTON MR SPECTROSCOPY OF BRAIN TUMORS
Figure 1. A 41-year-old female with no brain tumor. The proton magnetic resonance spectroscopy spectra appeared normal with a high N-acetyl aspartate (NAA) and creatine (Cre) peak and low choline (Cho) peak (a). Magnetic resonance imaging found an abnormal signal in her right thalamus (b). No biopsy was performed.
Cho/NAA of 1.45 and Cho/Cr of 1.49. Two patients with adenocarcinoma had average metabolic ratios of Cho/ NAA 1.39 and Cho/Cr 1.7. One patient with a fibrous histiocytoma had metabolic ratios of Cho/NAA 2.21 and Cho/Cr 2.21. One patient with a granuloma had metabolic ratios of Cho/NAA 2.40 and Cho/Cr 5.05. Two patients with lesions negative for malignancy— one with an astrocytosis and one with hematoma with inflammation— had average metabolic ratios of Cho/NAA 0.91 and Cho/Cr 1.47. In two patients with normal 1H-MRS, the average Cho/NAA was 0.57 and Cho/Cr was 0.94; a biopsy was not performed on these patients. The summary of 1HMRS findings and pathology data with tumor grade is presented in Table 1.
The mean 1H-MRS data ratios of Cho/NAA and Cho/Cr were compared between study groups using a parametric analysis of variance test. Normality and homogeneity of variance were checked for violations using Levene’s test. Statistically significant mean differences were considered achieved at a P value ⱕ0.05, two-tailed (Table 2).
DISCUSSION 1H-MRS
is a noninvasive technique for detecting the presence of cancerous tissue in the brain through its metabolic activity. 1H-MRS has an important role in clinical
53
MAGALHAES ET AL
Academic Radiology, Vol 12, No 1, January 2005
Figure 2. A 37-year-old female with glioblastoma multiforme grade 4. The patient’s proton magnetic resonance spectroscopy study shows a significant increase in choline (Cho) peak and decrease in N-acetyl aspartate (NAA) and creatine (Cre) peaks, consistent with malignant tumor (a). Magnetic resonance imaging shows a resection cavity in the right temporal lobe with questionable radiation necrosis or recurrent or residual tumor (b). The pathology confirmed the diagnosis of a glioblastoma multiforme.
neuro-oncologic management. It has the potential for accurate diagnosis without surgical tissue sampling, to be an adjunct to imaging for planning neurosurgical procedures to enhance probability of representative biopsy and maximizing cytoreduction, to monitor treatment success, and to differentiate radioactive injury from progressive tumor. 1H-MRS also serves a role in neurosurgical planning. 1H-MRS is a means of identifying the extent of tumoral infiltration and of providing the patient with a more optimal and safe resection. There is strong evidence that 1H-MRS signal patterns can serve to identify tumor type and grade. Grading of a brain tumor and histopathologic classification of brain lesions has important implications for clinical management. Grading and
54
classification is done histopathologically, requiring a biopsy from an open or stereotactic neurosurgical procedure, even recognizing the risk of sampling error— especially in the cases of large brain tumors—may show different grades within the mass. In tumors, Cho increases and corresponds to free Cho and phosphorylcholine. These levels are thought to rise because they are precursors of the phospholipids used for membrane construction. It is thought that their levels rise in tumors because of increased membrane synthesis in proliferating cells and enhanced membrane turnover. However, a substantial necrotic component of the tumor may attenuate the Cho signal elevation. Most
Academic Radiology, Vol 12, No 1, January 2005
PROTON MR SPECTROSCOPY OF BRAIN TUMORS
Figure 3. A 22-year-old male with an oligodendroglioma, grade 2. The proton magnetic resonance spectroscopy spectra shows a significant fall in N-acetyl aspartate (NAA) and creatine (Cre) with a small rise in choline (Cho) (a). The magnetic resonance imaging report shows T2 signal abnormality around the resection cavity in the left frontal lobe (b). The pathology confirmed the diagnosis of an oligodendroglioma.
studies agree that NAA levels fall in tumors because neurons and axons are impaired and destroyed. It is also widely believed that tumors and specially gliomas show elevated Cho levels because of cell membrane proliferation (1–13). Less supported by the data is the fall in Cr level because of energy exhaustion and rise in lipids from necrosis. Lipid levels also rise in tumors unpredictably (1,3–9). In our study, all gliomas had a Cr peak slightly less than in healthy brain tissue. The lowest NAA and Cr and highest Cho were found in necrotic portions of high-grade tumors (1–13).
Herminghaus et al. had a 95% correct prospective classification between low- and high-grade tumors. This study used metabolite signal intensities of the major resonances lipids, NAA, Cr, and Cho. These levels were quantified by calculating their ratio-to-signal intensity of Cr obtained from control normal brain tissue (used as control). Their values were then analyzed by the spectra pattern analysis method to determine the group membership of a spectrum of unknown origin based on data derived from spectra with confirmed diagnoses (3). Anaplastic and undifferentiated regions of the tumor can be differentiated by a higher intensity of Cho, and the
55
MAGALHAES ET AL
Academic Radiology, Vol 12, No 1, January 2005
Table 1 Proton Magnetic Resonance Spectroscopy Data in 27 Patients With Pathologic Correlation
Cho/NAA
Cho/Cr
Tumor Grade
37/F 60/M 47/M 42/M 37/M 38/M 59/M 24/M 68/M 24/M
5.97 2.51 8.41 3.90 3.02 3.35 16.92 5.85 8.90 2.5
2.28 1.90 2.11 3.91 1.63 12.38 2.31 1.73 1.88 2.38
4 4 4 4 4 4 4 4 4 3
38/M
1.81
1.30
3
45/M
1.25
1.19
3
57/F 51/M 22/M 42/M 24/M 47/M 46/F 24/F 63/F
2.61 2.26 1.76 1.45 1.71 1.03 1.74 2.21 2.40
1.30 1.08 2.13 1.49 1.53 0.83 2.57 2.21 5.05
2 2 2 1 NA NA NA NA NA
39/F
0.18
1.53
NA
46/F 83/F
0.98 1.36
1.14 1.82
NA NA
36/M
1.12
1.39
NA
41/F 43/F
0.81 0.33
1.14 0.73
NA NA
Age/Sex
Pathology GBM GBM GBM GBM GBM GBM GBM GBM Gliosarcoma Anaplastic oligoastrocytoma Anaplastic oligoastrocytoma Anaplastic oligodendroglioma Oligodendroglioma Oligodendroglioma Oligodendroglioma Astrocytoma Medulloblastoma Adenocarcinoma Adenocarcinoma Fibrous histiocytoma Tuberculous granuloma Hematoma with inflammation Reactive astrocytosis Negative for malignancy Negative for malignancy Biopsy not performed Biopsy not performed
Cho: Choline; NAA: N-acetyl aspartate; Cr: creatine; GBM: glioblastoma multiforme; NA: not applicable; *: biopsy not performed based on 1H-MRS results.
absence of or depressed NAA and tumor may be differentiated from normal brain tissue and necrosis (6). 1H-MRS may suggest a diagnosis different from that of MRI neoplasm. Contrary to MRI, which characterizes lesions by their morphology and contrast kinetics, 1H-MRS reflects lesion metabolism, thus providing completely independent biochemical information. In our study, 27 patients with suspicious brain MRIs underwent 1H-MRS that showed abnormal spectra in 25 patients and normal spectra in the remaining 2. Of these two patients with normal 1H-MRS, one presented with headache; a brain MRI showed a T2 hypersignal lesion in the right thalamus without any 1HMRS abnormality. A magnetic resonance angiogram (MRA) performed after 1H-MRS showed a small aneurysm in the right middle cerebral artery. The other patient had dizziness with minor white matter abnormality on MRI exam; no abnormality was seen in the 1H-MRS. All 25 patients with abnormal spectra had a biopsy or open surgery of the brain lesion, and the histopathology results were correlated with 1H-MRS. Sixteen patients had pathology diagnosis of brain astrocytoma of various grades, and the correlation with 1H-MRS spectra show that Grade 4 astrocytoma have Cho/NAA and Cho/Cr ratios two or three times higher than astrocytomas Grades 2 or 3; the lowest Cho/NAA ratio was shown by astrocytoma Grade 1. This agrees with previous studies, because Grade 1 tumors have the lowest elevation of Cho and have the smallest fall in NAA and Cr (3,5–7). In our series, brain lesions negative for malignancy, astrocytosis, hematoma, or inflammation had a lower ratio of Cho/NAA. A high Cho/NAA ratio mimicking Grade 2 astrocytoma was seen in tuberculous granuloma and fibrous histiocytoma, both of which were benign brain lesions, but with a Cho/NAA ratio as high as astrocytoma Grade 2. We compared our data in tumor grade glioma with other studies. In Grade 4 astrocytoma, the NAA/Cho ratios for their study was 17.10; in our study, it was 6.53. In Grade 3 astrocytoma, the NAA/Cho ratios for their study was 4.74;
Table 2 Comparison of Mean of Cho/NAA and Cho/Cr Ratios in 15 Patients With Astrocytoma Grades 2– 4 With 10 Patients With Other Brain Lesions
Cho/NAA Cho/Cr
Grade 2 (n ⫽ 3)
Grade 3 (n ⫽ 3)
Grade 4 (n ⫽ 9)
Others (n ⫽ 10)
2.21 ⫾ 0.24 1.50 ⫾ 0.32
1.85 ⫾ 0.36 1.62 ⫾ 0.38
6.53 ⫾ 1.50 3.34 ⫾ 1.15
1.21 ⫾ 0.23 1.84 ⫾ 0.40
NAA: N-acetyl aspartate; Cho: Choline; Cr: creatine; ⫾: standard error.
56
Academic Radiology, Vol 12, No 1, January 2005
in our study, 1.85. In Grade 2 astrocytoma, the NAA/Cho ratios for their study was 10.61; in our study, 2.21. In Grade 1 astrocytoma, the NAA/Cho ratios for their study was 2.67; in our study, 1.45. In both studies, the Grade 4 astrocytoma had the largest Cho/NAA ratio and the Grade 1 astrocytoma had the lowest. This is consistent with the majority of current studies; however, our Grade 2 astrocytoma had a higher Cho/NAA ratio than Grade 3 astrocytoma, which is not consistent with most current studies (3,5–7). The 1H-MRS is well known for differentiating residual or tumor recurrence from radiation necrosis. This is an important indication for spectroscopy in our institution. Usually, high Cho peak indicates tumor progression and low Cho peak indicates radiation necrosis. The presence of lipid peaks indicates necrosis that may result from tumor progression or radiation injury (7). We believe that 1H-MRS could assess long-term prognosis because higher ratios correlated with Grade 4 astrocytoma and shorter survival rates. A recent study showed the estimated median survival was 30 months for 45 patients with anaplastic astrocytoma (Grade 3) and 12 months for 69 patients with glioblastoma multiforme (Grade 4) (14). The patients on this study have been followed and the survival data are in progress. Several limitations of the present study are worth discussing. First, we need to analyze more patients with astrocytoma, especially those with Grades 1, 2, and 3. Second, the mean Cho/NAA ratio shown by Grade 3 was slightly lower than that of Grade 2—most likely because of the small size of the population studied. Third, the elevated Cho/Cr ratio showed by the other lesions group was due to the high Cho/Cr ratio seen in the tuberculous granuloma lesion. Fourth, there is always a risk of biopsy sampling error, especially in large brain tumors, which may show different grades within the mass with different 1H-MRS findings. In conclusion, 1H-MRS can be used to differentiate malignant and nonmalignant lesions from normal brain tissue. In general, high-grade astrocytoma have higher
PROTON MR SPECTROSCOPY OF BRAIN TUMORS
Cho/NAA and Cho/Cr ratios compared with low-grade astrocytoma. ACKNOWLEDGMENT
The authors thank Ron Thomas, PhD, for the statistics. REFERENCES 1. Negendank WG, Sauter R, Brown TR, et al. Proton magnetic resonance spectroscopy in patients with glial tumors: a multicenter study. J Neurosurg 1996; 84:449 – 458. 2. Preul MC, Caramanos Z, Collins L, et al. Accurate, noninvasive diagnosis of human brain tumors by using proton magnetic resonance of spectroscopy. Nat Med 1996; 2:323–325. 3. Herminghuas S, Dierks T, Pilatus U, et al. Determination of histopathological tumor grade in neuroepithelial brain tumors by using spectral pattern analysis of in vivo spectroscopic data. J Neurosurg 2003; 98: 74 – 81. 4. Frund R, Geissler A, Gliese M, et al. Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (H-MRS) of intracranial lipomas. Front Biosci 1997; 2:13–16. 5. Alger JR, Cloughesy TF. Structural and functional imaging of cerebral neoplasia. In: Togo A, Mazziotta JC, eds. Brain mapping—the disorders. San Diego, CA: Academic Press, 2000; 1–11 6. Tzika AA, Vajapeyam S, Barnes P. Multivoxel proton MR spectroscopy and hemodynamic MR imaging of childhood brain tumors: preliminary observations. Am J Neuroradiol 1997; 18:203–218. 7. Schlemmer HP, Bachert P, Herfarth KK, et al. Proton MR spectroscopic evaluation of suspicious brain lesions after stereotactic radiotherapy. Am J Neuroradiol 2001; 22:1316 –1324. 8. Lehnhardt FG, Rohn G, Ernestus RI, et al. 1H-and 31-MR spectroscopy of primary and recurrent human brain tumors in vitro: malignancy-characteristic profiles of water soluble and lipophilic spectral components. NMR Biomed 2001; 14:307–317. 9. Dechent P, Wilken B, Maxton C, et al. Differential diagnosis of focal brain lesions in children by quantitative single-voxel proton magnetic resonance spectroscopy. Proc Intl Soc Magn Reson Med 2000; 8:1921–1925. 10. Kinoshita Y, Yokota A. Absolute concentrations of metabolites in human brain tumors using in vitro proton magnetic resonance spectroscopy. NMR Biomed 1997; 10:2–12. 11. Tien RD, Lai PH, Lazeyras F, et al. Single-voxel proton brain spectroscopy exam (probe/SV) in patients with primary brain tumors. Am J Roentgenol 1996; 167:201–209. 12. Tosi, R. Ricci, Bottura G, et al. In vivo and in vitro nuclear magnetic resonance spectroscopy investigation of an intracranial mass. Oncol Rep 2001; 8:1337–1339. 13. Nelson SJ. Multivoxel magnetic resonance spectroscopy of brain tumors. Molec Cancer Ther 2003; 2:497–507. 14. Lin CL, Lieu AS, Lee KS, et al. The conditional probabilities of survival in patients with anaplastic astrocytoma or glioblastoma multiforme. Surg Neurol 2003; 60:402– 406.
57
Proton Magnetic Resonance Spectroscopy of Brain Tumors Correlated With Pathology1 Alvaro Magalhaes, MD, William Godfrey, MS, Yimin Shen, PhD, Jiani Hu, PhD, Wilbur Smith, MD
Rationale and Objectives. Evaluate proton magnetic resonance spectroscopy (1H-MRS) for assessing and grading brain tumors. Materials and Methods. The research was done at Detroit Medical Center in a 1.5-T Siemens MR magnet using singlevoxel or multivoxel MRS. This study consisted of 27 patients: 10 females and 17 males ages 22– 83 years (average age 43.8). The data were recorded for three peaks—N-acetyl aspartate (NAA), choline (Cho), creatine (Cr)—which were used to calculate the ratios Cho/NAA and Cho/Cr. Results. Abnormal spectra were seen in 25 patients and normal spectra in 2. In 16 patients with brain astrocytoma of various grades, the pathology grading was correlated with Cho/NAA and Cho/Cr. These values were 6.53 and 3.35 for nine patients with Grade 4 astrocytoma; 1.85 and 1.62 for three patients with Grade 3 astrocytoma; 2.21 and 1.50 for three patients with Grade 2 astrocytoma; and 1.45 and 1.49 for one patient with Grade 1 astrocytoma. The remaining nine patients with abnormal spectra were also correlated with pathology. Conclusion. MRS ratios can be used to differentiate malignant and nonmalignant lesions from normal brain tissue. In general, high-grade astrocytoma have higher Cho/NAA and Cho/Cr ratios compared with low-grade astrocytoma. Key Words. Magnetic resonance spectroscopy; brain tumor; astrocytoma. ©
AUR, 2005
The grading of a tumor has important implication for clinical management. The grade of the tumor defines to a substantial extent survival probability and course of therapy. The gold standard of tumor grading is histopathologic diagnosis requiring a biopsy during an open or stereotactic neurosurgical procedure. There is strong evidence that proton magnetic resonance spectroscopy (1HMRS) signal patterns can serve to identify tumor type and grade (1–3).
Acad Radiol 2005; 12:51–57 1 From the Department of Radiology, School of Medicine, 3990 John R, Wayne State University, Detroit Medical Center, Detroit, MI 48201-2097 (A.M., W.G., Y.S., J.H., W.S.). Received March 17, 2004; revised April 12, 2004; accepted October 23, 2004. W.G. received funding in the radiology summer externship program from the Department of Radiology and the School of Medicine of Wayne State University. Address correspondence to A.M. e-mail: [email protected]
© AUR, 2005 doi:10.1016/j.acra.2004.10.057
1H-MRS
uses radiofrequency excitation in the presence of a magnetic field to obtain signals from tissue chemicals other than water. The resonance frequency position of each peak is dependent on the chemical environment of that nucleus. It can be used to identify metabolites. The area under the peak may be calculated and yield relative measurements of the concentration of protons, when relaxation effects can be ignored—such as zero echo time. The width of the peak at half-height is proportional to 1/T2 when the magnetic field is perfectly homogenous. Many metabolites can be measured by 1H-MRS. Those of concern in this study are well differentiated in 1H-MRS and differ between neuronal and neoplastic brain tissue. Certain metabolites of neurons and abnormal brain tissue that most research is focused on are N-acetyl aspartate (NAA), choline (Cho), creatine (Cr), and lipids. In neoplastic tissue compared with normal neuronal tissue, there is a decrease in glucose, NAA, aspartate and Cr; how-
51
MAGALHAES ET AL
ever, alanine, hypotaurine, phosphoethanolamine, and Cho all rise (4). The question is if 1H-MRS is able to add new information to the high-quality anatomic data provided by conventional magnetic resonance imaging (MRI) techniques and whether this information could be used to replace or complement biopsy for the diagnosis and clinical management of patients.
Academic Radiology, Vol 12, No 1, January 2005
Twenty patients presented with a brain tumor; 16 patients of these 20 patients were diagnosed with astrocytoma that was graded according to WHO. Glioblastoma multiforme and gliosarcoma were Grade 4. Anaplastic oligoastrocytoma and anaplastic oligodendroglioma were Grade 3. Oligodendroglioma was Grade 2. Pilocytic astrocytoma was Grade 1. The 1H-MRS data of Cho/NAA and Cho/Cr were analyzed and compared with each tumor grade of the 15 patients with gliomas. Normal 1H-MRS spectra were obtained in two patients; no biopsy was performed on these patients.
MATERIALS AND METHODS Subjects This study consisted of 27 patients:, 10 females and 17 males ages 22– 83 years (average 43.8). All 27 patients presented with central nervous system symptoms and were assessed with brain MRI that demonstrated a suspicious brain tumor. Because of the brain lesion, the 27 patients were evaluated with 1H-MRS to confirm tumor diagnosis; in the case of astrocytomas, an attempt was made to grade the tumor. Lesions suspected of being tumors in 25 patients were then stereotactically biopsied. Twenty patients were diagnosed with a brain tumor and 16 of these with astrocytomas were graded according to the World Health Organization (WHO) classification. Five other patients presented with brain lesions, but had biopsies negative for malignancy. The remaining two patients with normal 1H-MRS results were not biopsied. Methods The 1H -MRS was done at a 1.5-T whole-body Siemens (Erlangen, Germany) imager using standard clinical singlevoxel or multivoxel 1H-MRS pulse sequences and postprocess protocol. The pulse sequence used was point-resolvedsingle-volume-spectroscopy (PRESS) with repetition time (TR) ⫽ 1,500 milliseconds and echo time (TE) ⫽ 135 milliseconds. Voxel size in most cases was 1.5 ⫻ 1.5 ⫻ 1.5 cm3 for single voxel and was 1 ⫻ 1 ⫻ 1.5 cm3 for multivoxel. Matrix size for chemical shift imaging (CSI) was 16 ⫻ 16 ⫻ 1 and field of view was 160 ⫻ 160 ⫻ 15 mm3. Multivoxel technique in 16 patients and single-voxel technique was done in 9 patients. Metabolite resonance intensities were determined from fitting peak areas using Siemens online software. Metabolite signals were expressed as rations. The data were recorded for three peaks: NAA, Cho, and Cr. These values were then placed into ratios Cho/NAA and Cho/Cr. In all 25 cases with abnormal 1H-MRS results, ratios Cho/NAA and Cho/Cr were analyzed and compared with histopathologic diagnoses.
52
RESULTS MRI and 1H-MRS was obtained in 27 patients and pathology examination was obtained in 25 patients. Seventeen of the 25 patients biopsied with astrocytoma were classified by WHO criteria: 9 patients with Grade 4 tumors, 3 patients with Grade 3 tumors, 3 patients with Grade 2 tumors, and 1 patient with Grade 1 tumor. The remaining nine patients with lesions were: one patient with medulloblastoma, two patients with adenocarcinomas, two patients with abnormal spectra but without malignancy, one patient with fibrous histiocytoma, one patient with hematoma, one patient with granuloma, and one patient with astrogliosis and inflammation. Two patients with normal 1H-MRS spectra did not have samples submitted for biopsy. H-MRS spectra of a normal brain are shown in Figure 1. The histopathologic grade was obtained after serial stereotactic and open surgery. The histopathologic diagnoses were made according to WHO criteria. The histopathologic classification was then compared with 1H-MRS data. NAA, Cho, and Cr peaks were obtained from 1H-MRS studies and placed in the following metabolic ratios: Cho/NAA and Cho/Cr. 1H-MRS and pathology data with tumor grade were correlated. Nine patients had Grade 4 astrocytoma; eight with glioblastoma multiforme and one with gliosarcoma. The average metabolic ratios of Cho/NAA and Cho/Cr were 6.53 and 3.35, respectively. 1H-MRS spectra of a Grade 4 glioma is shown in Fig. 2 Three patients had Grade 3 tumors: two with an anaplastic oligoastrocytoma and one with anaplastic oligodendroglioma. Their average metabolic ratios of Cho/NAA and Cho/Cr were 1.85 and 1.62, respectively. Three patients had oligodendroglioma tumors Grade 2; their metabolic ratios of Cho/NAA and Cho/Cr were 2.21 and 1.50, respectively. 1H-MRS spectra of a Grade 2 astrocytoma are shown in Fig. 3. One patient had an astrocytoma Grade 1 with metabolic ratios of
Academic Radiology, Vol 12, No 1, January 2005
PROTON MR SPECTROSCOPY OF BRAIN TUMORS
Figure 1. A 41-year-old female with no brain tumor. The proton magnetic resonance spectroscopy spectra appeared normal with a high N-acetyl aspartate (NAA) and creatine (Cre) peak and low choline (Cho) peak (a). Magnetic resonance imaging found an abnormal signal in her right thalamus (b). No biopsy was performed.
Cho/NAA of 1.45 and Cho/Cr of 1.49. Two patients with adenocarcinoma had average metabolic ratios of Cho/ NAA 1.39 and Cho/Cr 1.7. One patient with a fibrous histiocytoma had metabolic ratios of Cho/NAA 2.21 and Cho/Cr 2.21. One patient with a granuloma had metabolic ratios of Cho/NAA 2.40 and Cho/Cr 5.05. Two patients with lesions negative for malignancy— one with an astrocytosis and one with hematoma with inflammation— had average metabolic ratios of Cho/NAA 0.91 and Cho/Cr 1.47. In two patients with normal 1H-MRS, the average Cho/NAA was 0.57 and Cho/Cr was 0.94; a biopsy was not performed on these patients. The summary of 1HMRS findings and pathology data with tumor grade is presented in Table 1.
The mean 1H-MRS data ratios of Cho/NAA and Cho/Cr were compared between study groups using a parametric analysis of variance test. Normality and homogeneity of variance were checked for violations using Levene’s test. Statistically significant mean differences were considered achieved at a P value ⱕ0.05, two-tailed (Table 2).
DISCUSSION 1H-MRS
is a noninvasive technique for detecting the presence of cancerous tissue in the brain through its metabolic activity. 1H-MRS has an important role in clinical
53
MAGALHAES ET AL
Academic Radiology, Vol 12, No 1, January 2005
Figure 2. A 37-year-old female with glioblastoma multiforme grade 4. The patient’s proton magnetic resonance spectroscopy study shows a significant increase in choline (Cho) peak and decrease in N-acetyl aspartate (NAA) and creatine (Cre) peaks, consistent with malignant tumor (a). Magnetic resonance imaging shows a resection cavity in the right temporal lobe with questionable radiation necrosis or recurrent or residual tumor (b). The pathology confirmed the diagnosis of a glioblastoma multiforme.
neuro-oncologic management. It has the potential for accurate diagnosis without surgical tissue sampling, to be an adjunct to imaging for planning neurosurgical procedures to enhance probability of representative biopsy and maximizing cytoreduction, to monitor treatment success, and to differentiate radioactive injury from progressive tumor. 1H-MRS also serves a role in neurosurgical planning. 1H-MRS is a means of identifying the extent of tumoral infiltration and of providing the patient with a more optimal and safe resection. There is strong evidence that 1H-MRS signal patterns can serve to identify tumor type and grade. Grading of a brain tumor and histopathologic classification of brain lesions has important implications for clinical management. Grading and
54
classification is done histopathologically, requiring a biopsy from an open or stereotactic neurosurgical procedure, even recognizing the risk of sampling error— especially in the cases of large brain tumors—may show different grades within the mass. In tumors, Cho increases and corresponds to free Cho and phosphorylcholine. These levels are thought to rise because they are precursors of the phospholipids used for membrane construction. It is thought that their levels rise in tumors because of increased membrane synthesis in proliferating cells and enhanced membrane turnover. However, a substantial necrotic component of the tumor may attenuate the Cho signal elevation. Most
Academic Radiology, Vol 12, No 1, January 2005
PROTON MR SPECTROSCOPY OF BRAIN TUMORS
Figure 3. A 22-year-old male with an oligodendroglioma, grade 2. The proton magnetic resonance spectroscopy spectra shows a significant fall in N-acetyl aspartate (NAA) and creatine (Cre) with a small rise in choline (Cho) (a). The magnetic resonance imaging report shows T2 signal abnormality around the resection cavity in the left frontal lobe (b). The pathology confirmed the diagnosis of an oligodendroglioma.
studies agree that NAA levels fall in tumors because neurons and axons are impaired and destroyed. It is also widely believed that tumors and specially gliomas show elevated Cho levels because of cell membrane proliferation (1–13). Less supported by the data is the fall in Cr level because of energy exhaustion and rise in lipids from necrosis. Lipid levels also rise in tumors unpredictably (1,3–9). In our study, all gliomas had a Cr peak slightly less than in healthy brain tissue. The lowest NAA and Cr and highest Cho were found in necrotic portions of high-grade tumors (1–13).
Herminghaus et al. had a 95% correct prospective classification between low- and high-grade tumors. This study used metabolite signal intensities of the major resonances lipids, NAA, Cr, and Cho. These levels were quantified by calculating their ratio-to-signal intensity of Cr obtained from control normal brain tissue (used as control). Their values were then analyzed by the spectra pattern analysis method to determine the group membership of a spectrum of unknown origin based on data derived from spectra with confirmed diagnoses (3). Anaplastic and undifferentiated regions of the tumor can be differentiated by a higher intensity of Cho, and the
55
MAGALHAES ET AL
Academic Radiology, Vol 12, No 1, January 2005
Table 1 Proton Magnetic Resonance Spectroscopy Data in 27 Patients With Pathologic Correlation
Cho/NAA
Cho/Cr
Tumor Grade
37/F 60/M 47/M 42/M 37/M 38/M 59/M 24/M 68/M 24/M
5.97 2.51 8.41 3.90 3.02 3.35 16.92 5.85 8.90 2.5
2.28 1.90 2.11 3.91 1.63 12.38 2.31 1.73 1.88 2.38
4 4 4 4 4 4 4 4 4 3
38/M
1.81
1.30
3
45/M
1.25
1.19
3
57/F 51/M 22/M 42/M 24/M 47/M 46/F 24/F 63/F
2.61 2.26 1.76 1.45 1.71 1.03 1.74 2.21 2.40
1.30 1.08 2.13 1.49 1.53 0.83 2.57 2.21 5.05
2 2 2 1 NA NA NA NA NA
39/F
0.18
1.53
NA
46/F 83/F
0.98 1.36
1.14 1.82
NA NA
36/M
1.12
1.39
NA
41/F 43/F
0.81 0.33
1.14 0.73
NA NA
Age/Sex
Pathology GBM GBM GBM GBM GBM GBM GBM GBM Gliosarcoma Anaplastic oligoastrocytoma Anaplastic oligoastrocytoma Anaplastic oligodendroglioma Oligodendroglioma Oligodendroglioma Oligodendroglioma Astrocytoma Medulloblastoma Adenocarcinoma Adenocarcinoma Fibrous histiocytoma Tuberculous granuloma Hematoma with inflammation Reactive astrocytosis Negative for malignancy Negative for malignancy Biopsy not performed Biopsy not performed
Cho: Choline; NAA: N-acetyl aspartate; Cr: creatine; GBM: glioblastoma multiforme; NA: not applicable; *: biopsy not performed based on 1H-MRS results.
absence of or depressed NAA and tumor may be differentiated from normal brain tissue and necrosis (6). 1H-MRS may suggest a diagnosis different from that of MRI neoplasm. Contrary to MRI, which characterizes lesions by their morphology and contrast kinetics, 1H-MRS reflects lesion metabolism, thus providing completely independent biochemical information. In our study, 27 patients with suspicious brain MRIs underwent 1H-MRS that showed abnormal spectra in 25 patients and normal spectra in the remaining 2. Of these two patients with normal 1H-MRS, one presented with headache; a brain MRI showed a T2 hypersignal lesion in the right thalamus without any 1HMRS abnormality. A magnetic resonance angiogram (MRA) performed after 1H-MRS showed a small aneurysm in the right middle cerebral artery. The other patient had dizziness with minor white matter abnormality on MRI exam; no abnormality was seen in the 1H-MRS. All 25 patients with abnormal spectra had a biopsy or open surgery of the brain lesion, and the histopathology results were correlated with 1H-MRS. Sixteen patients had pathology diagnosis of brain astrocytoma of various grades, and the correlation with 1H-MRS spectra show that Grade 4 astrocytoma have Cho/NAA and Cho/Cr ratios two or three times higher than astrocytomas Grades 2 or 3; the lowest Cho/NAA ratio was shown by astrocytoma Grade 1. This agrees with previous studies, because Grade 1 tumors have the lowest elevation of Cho and have the smallest fall in NAA and Cr (3,5–7). In our series, brain lesions negative for malignancy, astrocytosis, hematoma, or inflammation had a lower ratio of Cho/NAA. A high Cho/NAA ratio mimicking Grade 2 astrocytoma was seen in tuberculous granuloma and fibrous histiocytoma, both of which were benign brain lesions, but with a Cho/NAA ratio as high as astrocytoma Grade 2. We compared our data in tumor grade glioma with other studies. In Grade 4 astrocytoma, the NAA/Cho ratios for their study was 17.10; in our study, it was 6.53. In Grade 3 astrocytoma, the NAA/Cho ratios for their study was 4.74;
Table 2 Comparison of Mean of Cho/NAA and Cho/Cr Ratios in 15 Patients With Astrocytoma Grades 2– 4 With 10 Patients With Other Brain Lesions
Cho/NAA Cho/Cr
Grade 2 (n ⫽ 3)
Grade 3 (n ⫽ 3)
Grade 4 (n ⫽ 9)
Others (n ⫽ 10)
2.21 ⫾ 0.24 1.50 ⫾ 0.32
1.85 ⫾ 0.36 1.62 ⫾ 0.38
6.53 ⫾ 1.50 3.34 ⫾ 1.15
1.21 ⫾ 0.23 1.84 ⫾ 0.40
NAA: N-acetyl aspartate; Cho: Choline; Cr: creatine; ⫾: standard error.
56
Academic Radiology, Vol 12, No 1, January 2005
in our study, 1.85. In Grade 2 astrocytoma, the NAA/Cho ratios for their study was 10.61; in our study, 2.21. In Grade 1 astrocytoma, the NAA/Cho ratios for their study was 2.67; in our study, 1.45. In both studies, the Grade 4 astrocytoma had the largest Cho/NAA ratio and the Grade 1 astrocytoma had the lowest. This is consistent with the majority of current studies; however, our Grade 2 astrocytoma had a higher Cho/NAA ratio than Grade 3 astrocytoma, which is not consistent with most current studies (3,5–7). The 1H-MRS is well known for differentiating residual or tumor recurrence from radiation necrosis. This is an important indication for spectroscopy in our institution. Usually, high Cho peak indicates tumor progression and low Cho peak indicates radiation necrosis. The presence of lipid peaks indicates necrosis that may result from tumor progression or radiation injury (7). We believe that 1H-MRS could assess long-term prognosis because higher ratios correlated with Grade 4 astrocytoma and shorter survival rates. A recent study showed the estimated median survival was 30 months for 45 patients with anaplastic astrocytoma (Grade 3) and 12 months for 69 patients with glioblastoma multiforme (Grade 4) (14). The patients on this study have been followed and the survival data are in progress. Several limitations of the present study are worth discussing. First, we need to analyze more patients with astrocytoma, especially those with Grades 1, 2, and 3. Second, the mean Cho/NAA ratio shown by Grade 3 was slightly lower than that of Grade 2—most likely because of the small size of the population studied. Third, the elevated Cho/Cr ratio showed by the other lesions group was due to the high Cho/Cr ratio seen in the tuberculous granuloma lesion. Fourth, there is always a risk of biopsy sampling error, especially in large brain tumors, which may show different grades within the mass with different 1H-MRS findings. In conclusion, 1H-MRS can be used to differentiate malignant and nonmalignant lesions from normal brain tissue. In general, high-grade astrocytoma have higher
PROTON MR SPECTROSCOPY OF BRAIN TUMORS
Cho/NAA and Cho/Cr ratios compared with low-grade astrocytoma. ACKNOWLEDGMENT
The authors thank Ron Thomas, PhD, for the statistics. REFERENCES 1. Negendank WG, Sauter R, Brown TR, et al. Proton magnetic resonance spectroscopy in patients with glial tumors: a multicenter study. J Neurosurg 1996; 84:449 – 458. 2. Preul MC, Caramanos Z, Collins L, et al. Accurate, noninvasive diagnosis of human brain tumors by using proton magnetic resonance of spectroscopy. Nat Med 1996; 2:323–325. 3. Herminghuas S, Dierks T, Pilatus U, et al. Determination of histopathological tumor grade in neuroepithelial brain tumors by using spectral pattern analysis of in vivo spectroscopic data. J Neurosurg 2003; 98: 74 – 81. 4. Frund R, Geissler A, Gliese M, et al. Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (H-MRS) of intracranial lipomas. Front Biosci 1997; 2:13–16. 5. Alger JR, Cloughesy TF. Structural and functional imaging of cerebral neoplasia. In: Togo A, Mazziotta JC, eds. Brain mapping—the disorders. San Diego, CA: Academic Press, 2000; 1–11 6. Tzika AA, Vajapeyam S, Barnes P. Multivoxel proton MR spectroscopy and hemodynamic MR imaging of childhood brain tumors: preliminary observations. Am J Neuroradiol 1997; 18:203–218. 7. Schlemmer HP, Bachert P, Herfarth KK, et al. Proton MR spectroscopic evaluation of suspicious brain lesions after stereotactic radiotherapy. Am J Neuroradiol 2001; 22:1316 –1324. 8. Lehnhardt FG, Rohn G, Ernestus RI, et al. 1H-and 31-MR spectroscopy of primary and recurrent human brain tumors in vitro: malignancy-characteristic profiles of water soluble and lipophilic spectral components. NMR Biomed 2001; 14:307–317. 9. Dechent P, Wilken B, Maxton C, et al. Differential diagnosis of focal brain lesions in children by quantitative single-voxel proton magnetic resonance spectroscopy. Proc Intl Soc Magn Reson Med 2000; 8:1921–1925. 10. Kinoshita Y, Yokota A. Absolute concentrations of metabolites in human brain tumors using in vitro proton magnetic resonance spectroscopy. NMR Biomed 1997; 10:2–12. 11. Tien RD, Lai PH, Lazeyras F, et al. Single-voxel proton brain spectroscopy exam (probe/SV) in patients with primary brain tumors. Am J Roentgenol 1996; 167:201–209. 12. Tosi, R. Ricci, Bottura G, et al. In vivo and in vitro nuclear magnetic resonance spectroscopy investigation of an intracranial mass. Oncol Rep 2001; 8:1337–1339. 13. Nelson SJ. Multivoxel magnetic resonance spectroscopy of brain tumors. Molec Cancer Ther 2003; 2:497–507. 14. Lin CL, Lieu AS, Lee KS, et al. The conditional probabilities of survival in patients with anaplastic astrocytoma or glioblastoma multiforme. Surg Neurol 2003; 60:402– 406.
57