HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: 2382041896v1 238/3/958    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Stadlbauer, A. Articles by Ganslandt, O. Search for Related Content PubMed PubMed Citation Articles by Stadlbauer, A. Articles by Ganslandt, O.

Preoperative Grading of Gliomas by Using Metabolite Quantification with High-Spatial-Resolution Proton MR Spectroscopic Imaging¹

¹ From the Department of Neurosurgery, Neurocenter, University of Erlangen-Nuremberg, Erlangen, Germany. Received November 8, 2004; revision requested January 5, 2005; revision received February 11; accepted March 7; final version accepted April 5. A.S. supported by the German Research Society (DFG Ga 638/2-1). S.G. supported by the Austrian Science Fund (FWF P14715-PSY). Address correspondence to E.M., MR Center of Excellence, Medical University of Vienna, Lazarettgasse 14, A-1090 Vienna, Austria (e-mail: ewald.moser{at}meduniwien.ac.at).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
Purpose: To evaluate proton magnetic resonance (MR) spectroscopic imaging with high spatial resolution for preoperative grading of suspected World Health Organization grades II and III gliomas.

Materials and Methods: Institutional ethics committee approval and informed consent were obtained for control subjects but were not required for the retrospective component involving patients. Twenty-six patients (10 women, 16 men; mean age, 37.5 years) suspected of having gliomas and 26 age- and sex-matched control subjects underwent proton MR spectroscopy. Absolute metabolite concentrations for choline-containing compounds (Cho), creatine (Cr), and N-acetylaspartate (NAA)–N-acetylaspartylglutamate (total NAA [tNAA]) were calculated by using a user-independent spectral fit program. Metabolic maps of Cho/tNAA ratios were calculated, segmented, and used for MR spectroszpcopy–guided stereotactic brain biopsy. Two-sided paired Student t tests were used to test for statistical significance.

Results: Significantly lower Cho levels (P = .002) and higher tNAA levels (P = .010) were found in grade II tumors (n = 9) compared with grade III tumors (n = 17). The average Cho/tNAA ratio over the voxels in the tumor center showed a distinct difference (P < .001) between grade II and III gliomas at a threshold of 0.8 (with ratios <0.8 for grade II). The maximum Cr concentration in the tumor showed a clear-cut threshold between grade III oligodendrogliomas and oligoastrocytomas (Cr level, <7 mmol/L) and grade III astrocytomas (Cr level, >7 mmol/L; P = .020). Comparison between the histopathologic findings from the MR spectroscopy–guided biopsy samples (76 biopsies from 26 patients) and molar metabolite values in corresponding voxels located at the biopsy sampling points showed a negative linear correlation for tNAA (r = –0.905) and a positive exponential correlation for Cho (r = 0.769) and Cho/tNAA (r = 0.885).

Conclusion: Proton MR spectroscopic imaging with high spatial resolution allows preoperative grading of gliomas.

© RSNA, 2006

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
In patients suspected of having brain tumors, magnetic resonance (MR) imaging is considered to be the reference standard for preoperative diagnostic evaluation and for providing optimal information for clinical decision making. However, even the current methods of choice—that is, T2-weighted MR imaging and pre- and postcontrast T1-weighted MR imaging (as a technique for visualizing regions where the blood-brain barrier is damaged)—are not specific for tumors and can result in ambiguous or misleading results (1,2). These clinically routine techniques show a diagnostic accuracy of 30%–90% (3,4) depending on the type of lesion. Stereotactic brain biopsy is often used for histopathologic diagnosis. This invasive technique has a morbidity of up to 3.6%, a hemorrhage rate of up to 8%, and a mortality of up to 1.7%, as assessed over a large number of studies (5–9). The diagnostic accuracy is rated as 91% (for low-grade astrocytoma), 83% (for anaplastic astrocytoma), and 88% (for glioblastoma multiforme). The histologic grade of malignancy, however, is predictable with an accuracy of only 57%–61% (10).

Proton MR spectroscopic imaging is a noninvasive tool for investigating the spatial distribution of metabolic changes in brain lesions. Unfortunately, there is no tumor-specific metabolite that is detectable with in vivo MR spectroscopy. It is possible, however, to detect specific patterns in the changes of metabolite concentrations compared with those in the normal brain. Authors of several studies have reported increased levels of choline-containing compounds (Cho) and a reduction in the signal intensity of the N-acetylaspartate (NAA) and creatine (Cr) in brain tumors (11–16). Cho are composed of choline, phosphocholine, and glycerophosphocholine and are thought to be markers for increased membrane turnover or higher cellular density (17,18). NAA is regarded as a neuronal marker mainly contained within neurons (19). The Cr peak is the signal from both Cr and phosphocreatine and plays a role in tissue energy metabolism (20).

The range of Cho increase and NAA decrease is compatible with the range of tumor infiltration (1,21). For pathologic conditions that appear similar to brain tumors at conventional MR imaging, variations in the changes of these three metabolites and others (inositol, lactate, lipids, glutamine and/or glutamate, alanine) can be used for differential diagnosis (22–31). It has recently been shown in several studies that it is possible to differentiate the degree of malignancy of brain tumors (32–34), but none of these studies presented a distinct difference in metabolic changes between the degrees of malignancy. Thus, the purpose of our study was to evaluate proton MR spectroscopic imaging with high spatial resolution for preoperative grading of suspected World Health Organization grades II and III gliomas.

Details about our MR spectroscopy–guided biopsy procedure have been described recently (35,36).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
Patients and Control Subjects
Our institutional ethic committee did not require its approval or informed consent for the retrospective component of our study, which involved patients. We did have ethics committee approval and informed consent for the prospective component of our study, which involved healthy volunteers. We examined 26 patients (age range, 18–63 years; mean age ± standard deviation, 37.5 years ± 11.7) with untreated supratentorial gliomas (suspected of being grade II or III). There were 10 women (20–50 years of age; mean age, 34.1 years ± 9.7) and 16 men (18–63 years of age; mean age, 39.6 years ± 12.6) in the patient group. The T1-weighted MR images in all patients showed hypointense lesions with no or only minor contrast material enhancement. On the T2-weighted MR images, all lesions were hyperintense. All images were examined by a neuroradiologist (B.T., 7 years of experience with brain MR imaging) and were determined to show gliomas of either grade II or grade III. All lesions were confirmed histopathologically by a neuropathologist (R.B., with 6 years of experience). Information about the types and locations of tumors is summarized in Table 1.

MR Imaging and Proton MR Spectroscopic Imaging
MR imaging and proton MR spectroscopic imaging were performed in separate sessions by using a 1.5-T clinical whole-body imager (Magnetom Sonata; Siemens, Erlangen, Germany) equipped with the standard head coil. Conventional MR imaging for diagnosis of glioma grade II or III consisted of (a) a transverse T2-weighted turbo spin-echo sequence (repetition time msec/echo time msec, 5600–6490/98; 5-mm section thickness), (b) a transverse fluid-attenuated inversion recovery sequence (10 000/103, 5-mm section thickness), and (c) coronal T1-weighted unenhanced and gadolinium-enhanced gradient-echo sequences (430/12 and 525/17, respectively; 5-mm section thickness).

In each proton MR spectroscopic imaging session, a localization image and a transverse T1-weighted spin-echo image (500/15, 256 x 256 matrix size, 16 x 16-cm field of view, and 20 sections with no intersection gap and a 2-mm section thickness) were acquired and used for planning the MR spectroscopic imaging experiment and for integration of spectroscopic data into a stereotactic system (35), respectively. The volume of interest of the MR spectroscopic imaging experiment with point-resolved spectroscopy volume preselection was aligned parallel to the transverse localizer sections and positioned to exclude lipids of the skull and subcutaneous fat. Water suppression was achieved by using three chemical shift–selective pulses prior to the point-resolved spectroscopy excitation. The MR spectroscopic imaging parameters were as follows: 1600/135, 24 x 24 circular phase-encoding scheme across a 16 x 16-cm field of view, 10-mm section thickness, 50% Hamming filter and two signals acquired, spectral width of 1000 Hz, and acquisition size of 1024 complex points. The total spectroscopic data acquisition time was less than 13 minutes. The nominal voxel size was 0.67 x 0.67 x 1.0 cm (approximately 0.45 cm³ resolution). Taking into account the effect of the k-space filter (applied 50% Hamming filter) on the full width at half maximum (FWHM) (37), and after zero-filling to a 32 x 32 matrix size, the volume of a voxel relevant for absolute quantification with a linear combination of model spectra (LCModel version 6.0; Stephen Provencher, Oakville, Ontario, Canada) was 0.52 cm³.

For registration to the frameless stereotactic system, six to eight adhesive skin fiducial marks were placed in a scattered pattern on the head surface. To obtain a neuronavigation MR data set, a three-dimensional anatomic magnetization-prepared rapid acquisition gradient-echo sequence was performed in a single session, 1 day before surgery, with the following parameters: 2020/4.38, 25 x 25-cm field of view, 1-mm isotropic voxels, and 160 sections.

Proton MR Spectroscopic Data Analysis
MR spectroscopic imaging data were exponentially filtered with a line broadening factor of 3 Hz, zero filled to 2048 data points and Fourier transformed with respect to the spectral dimension by using the freely available reconstruction program CSX (version for Linux operating system; Peter B. Barker, Baltimore, Md). To remove the residual water peak, we used a high-pass convolution filter (50-Hz stop band) (38). The magnitude of spectra was calculated, the position of NAA was set to 2.02 ppm, and a susceptibility correction was applied. The peak areas for Cho, Cr, and total NAA (tNAA) (ie, the total of NAA plus N-acetylaspartylglutamate) were calculated by means of integration over the frequency range of 3.34–3.14 ppm, 3.14–2.94 ppm, and 2.22–1.82 ppm, respectively. Smooth linear interpolation to a 256 x 256 matrix resulted in the metabolic maps. Cho and tNAA images were used to calculate a map of Cho/tNAA ratios. Segmentation of the tumor owing to the metabolic changes related to the lesion on the Cho/tNAA ratio map was performed by one of the authors (A.S.) as described previously (36).

The user-independent spectral fit program LCModel (39) enables absolute metabolite quantification of MR spectroscopic imaging data. The spectra were analyzed as a linear combination of a set of reference basis spectra. For all spectroscopic data, we used the reference basis set for point-resolved spectroscopy and echo time of 135 msec, which were not acquired with the same MR unit. Hence, LCModel concentrations are over- or underestimated by a constant, local imager-dependent factor. We used four calibration solutions with 5, 10, 20, and 50 mmol/L of NAA in 300-mL spherical glass phantoms at room temperature to obtain a calibration curve. The solvent used was a buffer solution to meet in vivo conditions for coil loading (40). Calibration measurements were obtained by prescribing a point-resolved spectroscopy voxel (2 x 2 x 2 cm) in the center of the four spheres. Parameters used were 2000/135, 256 signals acquired, and an imaging time of 8.5 minutes. Unscaled LCModel concentrations for NAA were plotted against known phantom concentrations, and a linear fit was applied to the data. The LCModel correction factor was calculated (by A.S.) from the slope of the regression model. We applied no reference pulse (eg, 90° nonselective pulse) because in our opinion this is inaccurate when using different transmitter and receiver coils. The correction factor was not corrected for temperature effects and different repetition times. Our method for quantification was similar to that of McLean et al (41).

Metabolite concentrations from the in vivo MR spectroscopic imaging data were calculated on a workstation (SGI, Mountain View, Calif) by using the LCModel program and the correction factor. Corrections for relaxation time effects by using average T1 (1452 msec) and T2 (280 msec) times from the literature (42) resulted in the absolute metabolite values. Because of the fact that N-acetylaspartylglutamate is quite difficult to resolve from NAA, it is recommended to calculate the concentration of NAA plus N-acetylaspartylglutamate as tNAA.

For all subjects (ie, all patients and control subjects), at least 40–60 voxels of predominantly white matter in normal brain tissue were selected and evaluated by two authors (A.S. and S.G., with 4 and 7 years of experience, respectively, with brain MR imaging and MR spectroscopic imaging) in consensus. Additionally, for patients, we used the spectroscopic image of the segmented pathologic metabolite changes to select the voxels in the tumor. At least 95% of the volume of every single voxel had to be located in the tumor volume, determined as the area with pathologic metabolite changes by means of the method described elsewhere (36), to be included (by A.S. and S.G.) in the statistical analysis. The sum of these selected voxels was defined as whole tumor and was differentiated in a further step between tumor border and tumor center on the basis of the segmented Cho/tNAA ratio maps by two authors (A.S. and S.G.).

All spectral fits were performed in an analysis window from 1.0 to 3.85 ppm. Spectra with a signal-to-noise ratio of less than 2 and an FWHM greater than 0.075 ppm were not included in the statistical analysis. Metabolites with a standard deviation of 20% or greater, as given by the LCModel program, were also not included in the statistical analysis. In tumors, values for tNAA with a standard deviation of less than 50% were accepted (40). The metabolite values for normal brain were compared with the summarized findings from five publications (43–47) by using the phantom replacement technique for quantification.

MR Spectroscopy–guided Stereotactic Biopsy
Integration of MR spectroscopic data as segmented Cho/tNAA ratio maps was achieved by using a combined data set consisting of MR imaging and MR spectroscopic imaging data, a so-called MR imaging–MR spectroscopic imaging hybrid data set (35). The hybrid data set was transferred to the planning workstation of the navigation system (VectorVision Sky; BrainLab, Heimstetten, Germany), and a semiautomated coregistration was applied (VectorVision 2 Planning 1.3; BrainLab) to the three-dimensional magnetization-prepared rapid acquisition gradient-echo data set. This fused three-dimensional MR data set was used for frameless stereotaxy and MR spectroscopy–guided biopsy sampling.

Prior to the tumor resection, biopsy specimens were sampled by one of three neurosurgeons (O.G., C.N., or R.F., with a range of experience of 13–30 years) in several regions according to the segmented metabolic map, which represented the pathologic metabolite change of tumor tissue in MR spectroscopy, by using a stereotactic needle that was tracked by the navigation system. This procedure before the tumor resection ensured a minimal interference of brain shift that would render the neuronavigation inaccurate. The coordinates of each biopsy specimen locus were labeled and documented in the fused three-dimensional MR data set by one of three neurosurgeons (O.G., C.N., R.F.). The histopathologic findings of each specimen were tracked back (A.S.) to the exact voxel positions in the fused MR data set containing the spectroscopic information.

Histopathologic Evaluation
All glioma specimens were histologically examined by a neuropathologist (R.B.) and graded according to the World Health Organization classification of tumors of the nervous system. Tumor cells were identified by using formalin-fixed and paraffin-embedded sections stained with either hematoxylin-eosin or monoclonal antibodies against p53 (Dako, Glostrup, Denmark), MAP2c (clone), or Ki-67 (Dako). MAP2c is a microtubule-associated protein that is solely expressed in neuronal cells and glial tumor cells (48). Semiquantitative assessment of tumor cells versus preexisting brain parenchyma was obtained microscopically by using analysis software (AnalySIS; Soft Imaging System, Leinenfeld-Echterdingen, Germany) at x200 magnification in five different subfields of 348 x 261 µm in size. Only cells with a distinct nucleus were considered. All data were calculated as mean tumor cell number as a percentage of the number of whole cells.

Statistical Analysis
MR spectroscopic imaging data were analyzed (A.S.) by using statistical software (Excel 2002; Microsoft, Redmond, Wash). Differences in molar concentrations of Cho, Cr, and tNAA, and the Cho/tNAA ratio between groups (patients with glioma grades II and III and control subjects) and tissue types (whole tumor, tumor center, tumor border, and contralateral normal brain) were analyzed by using the two-sided paired Student t test. Linear and exponential regression analyses were calculated for correlations of tumor infiltration and molar concentrations of Cho, Cr, tNAA, and the Cho/tNAA ratio. A P value of less than .05 was considered to indicate a statistically significant difference for all tests.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
In Vivo Absolute Metabolite Quantification
The correlation between the results of the LCModel fitting routine of the single-voxel calibration experiments and the given NAA concentrations in the four calibration solutions was found to be close to identity (r² = 0.999989, P < .001). Regression analysis and corrections for relaxation time effects revealed a scale factor for absolute metabolite quantification of the MR spectroscopic imaging data for our MR system of 5.2 x 10^–3.

High-spatial-resolution MR spectroscopic imaging data of good spectral quality were obtained in all patients and control subjects. Proton MR spectroscopic imaging data analysis, including the computing of metabolic maps, and segmentation of pathologic metabolite changes on the Cho/tNAA ratio image were successfully performed in all patients (Fig 1). Metabolite concentrations for Cho, Cr, and tNAA were calculated in contralateral normal brain tissue and lesions for all 26 patients and 26 control subjects, respectively (Table 2). Considering the exclusion criteria for MR spectroscopic imaging data with regard to the spectral quality (FWHM, percentage standard deviation, and signal-to-noise ratio), we found a lower limit for quantification of tNAA in tumors—1.4 mmol/L.

View larger version (130K): [in this window] [in a new window] [Download PPT slide]   Figure 1: Patient 2. In vivo absolute quantification of metabolic changes in an oligoastrocytoma grade III. Left: Transverse T2-weighted (6490/98) MR image overlaid with metabolic maps of, A, tNAA and, B, Cho. C, Enlarged section of same anatomic image depicts tumor area superimposed with segmented Cho/tNAA ratio image. = Voxels, and tumor voxels selected for statistical evaluation are enclosed by a white border. Voxel 1, contralateral normal brain tissue; 2, tumor border; 3, transition zone; and 4, tumor center. Right: Corresponding spectra for voxels 1–4 fitted with LCModel (red line) show calculated concentration of Cho, total Cr (tCr), and tNAA next to respective metabolite peak. FWHM in the spectra ranges from 0.038 to 0.054 ppm, and signal-to-noise ratio ranges from 2 to 5. For tNAA in tumor center (voxel 4), the percentage standard deviation (%SD) estimated with LCModel was more than 50%. However, the spectra show consecutive increase of Cho and decrease of tNAA and Cr from normal brain tissue to tumor center. Violet and blue = minimum value, red = maximum value in the image. Note the increasing (inverted) lactate doublet at about 1.3–1.4 ppm that has not been quantified.

Preoperative Grading and Tissue Differentiation
A comparison of the molar concentrations averaged over the whole tumor between the group of patients with grade II gliomas (n = 9) and those with grade III gliomas (n = 17) showed significantly higher values for tNAA (P = .010) and lower values for Cho (P = .002) and Cho/tNAA (P = .003) for grade II gliomas (Table 3). There were no significant differences for Cr in the tumor between grade II and III gliomas. Differentiation between tumor center and tumor border on the basis of the color-coded segmented Cho/tNAA ratio maps was performed for localization of the most malignant tumor region, which was assumed to be the region with the most pronounced metabolic changes. Voxels in predominantly red areas on the segmented Cho/tNAA ratio image were determined as voxels in the tumor center. All other voxel positions were determined as voxels located in the tumor border (Fig 2 ). For the quantified Cho/tNAA value averaged over the voxels located in the tumor center, there was a significant and definite difference (P < .001) between the group of patients with grade II gliomas (n = 9) and the group of patients with grade III gliomas (n = 17). All patients with a grade II glioma had a Cho/tNAA ratio of less than 0.8, whereas all patients with a grade III glioma had a Cho/tNAA ratio of greater than 0.8 (Fig 2)—that is, there was no overlap for the range of Cho/tNAA values in the tumor center between grade II and III gliomas.

View larger version (120K): [in this window] [in a new window] [Download PPT slide]   Figure 2a: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 6 (astrocytoma grade II) and (b) patient 7 (oligodendroglioma grade III), superimposed with color-coded segmented Cho/NAA ratio image. Voxels in predominantly red areas (enclosed by black line) were determined as voxels in the tumor center; all other voxel positions were determined as voxels in the tumor border (enclosed by white line). Green lines show volume of interest of the MR spectroscopic examination. Violet and blue = minimum value, red = maximum value. (c) Box plot shows a significant and definite difference (P < .001) for quantified Cho/tNAA ratio averaged over the tumor center between patients with glioma grades II (n = 9) and III (n = 17). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (108K): [in this window] [in a new window] [Download PPT slide]   Figure 2b: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 6 (astrocytoma grade II) and (b) patient 7 (oligodendroglioma grade III), superimposed with color-coded segmented Cho/NAA ratio image. Voxels in predominantly red areas (enclosed by black line) were determined as voxels in the tumor center; all other voxel positions were determined as voxels in the tumor border (enclosed by white line). Green lines show volume of interest of the MR spectroscopic examination. Violet and blue = minimum value, red = maximum value. (c) Box plot shows a significant and definite difference (P < .001) for quantified Cho/tNAA ratio averaged over the tumor center between patients with glioma grades II (n = 9) and III (n = 17). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 2c: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 6 (astrocytoma grade II) and (b) patient 7 (oligodendroglioma grade III), superimposed with color-coded segmented Cho/NAA ratio image. Voxels in predominantly red areas (enclosed by black line) were determined as voxels in the tumor center; all other voxel positions were determined as voxels in the tumor border (enclosed by white line). Green lines show volume of interest of the MR spectroscopic examination. Violet and blue = minimum value, red = maximum value. (c) Box plot shows a significant and definite difference (P < .001) for quantified Cho/tNAA ratio averaged over the tumor center between patients with glioma grades II (n = 9) and III (n = 17). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (131K): [in this window] [in a new window] [Download PPT slide]   Figure 3a: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 22 (oligoastrocytoma grade III) and (b) patient 14 (astrocytoma grade III), superimposed with the volume of interest of the MR spectroscopic examination (point-resolved spectroscopy box, in green) and voxel position showing maximum Cr concentration (red squares) within the lesion. (c, d) LCModel fits (red line) of spectra corresponding to voxel positions in a and b, respectively. FWHM in both spectra is 0.054 ppm, and signal-to-noise ratio is 4. (e) Box plot of maximum Cr concentrations between patients with oligoastrocytoma or oligodendroglioma grade III (n = 12) and those with astrocytoma grade III (n = 5). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (141K): [in this window] [in a new window] [Download PPT slide]   Figure 3b: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 22 (oligoastrocytoma grade III) and (b) patient 14 (astrocytoma grade III), superimposed with the volume of interest of the MR spectroscopic examination (point-resolved spectroscopy box, in green) and voxel position showing maximum Cr concentration (red squares) within the lesion. (c, d) LCModel fits (red line) of spectra corresponding to voxel positions in a and b, respectively. FWHM in both spectra is 0.054 ppm, and signal-to-noise ratio is 4. (e) Box plot of maximum Cr concentrations between patients with oligoastrocytoma or oligodendroglioma grade III (n = 12) and those with astrocytoma grade III (n = 5). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (63K): [in this window] [in a new window] [Download PPT slide]   Figure 3c: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 22 (oligoastrocytoma grade III) and (b) patient 14 (astrocytoma grade III), superimposed with the volume of interest of the MR spectroscopic examination (point-resolved spectroscopy box, in green) and voxel position showing maximum Cr concentration (red squares) within the lesion. (c, d) LCModel fits (red line) of spectra corresponding to voxel positions in a and b, respectively. FWHM in both spectra is 0.054 ppm, and signal-to-noise ratio is 4. (e) Box plot of maximum Cr concentrations between patients with oligoastrocytoma or oligodendroglioma grade III (n = 12) and those with astrocytoma grade III (n = 5). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (61K): [in this window] [in a new window] [Download PPT slide]   Figure 3d: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 22 (oligoastrocytoma grade III) and (b) patient 14 (astrocytoma grade III), superimposed with the volume of interest of the MR spectroscopic examination (point-resolved spectroscopy box, in green) and voxel position showing maximum Cr concentration (red squares) within the lesion. (c, d) LCModel fits (red line) of spectra corresponding to voxel positions in a and b, respectively. FWHM in both spectra is 0.054 ppm, and signal-to-noise ratio is 4. (e) Box plot of maximum Cr concentrations between patients with oligoastrocytoma or oligodendroglioma grade III (n = 12) and those with astrocytoma grade III (n = 5). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (28K): [in this window] [in a new window] [Download PPT slide]   Figure 3e: (a, b) Transverse T2-weighted (6490/98) MR images in (a) patient 22 (oligoastrocytoma grade III) and (b) patient 14 (astrocytoma grade III), superimposed with the volume of interest of the MR spectroscopic examination (point-resolved spectroscopy box, in green) and voxel position showing maximum Cr concentration (red squares) within the lesion. (c, d) LCModel fits (red line) of spectra corresponding to voxel positions in a and b, respectively. FWHM in both spectra is 0.054 ppm, and signal-to-noise ratio is 4. (e) Box plot of maximum Cr concentrations between patients with oligoastrocytoma or oligodendroglioma grade III (n = 12) and those with astrocytoma grade III (n = 5). The central box represents values from lower to upper quartile (25–75 percentile), the middle line represents the median, and vertical bars extend from minimum to maximum value. Small black squares show individual data points.

View larger version (94K): [in this window] [in a new window] [Download PPT slide]   Figure 4a: Patient 6. (a, b) Screen shots from the frameless stereotactic system of (a) transverse and (b) sagittal reconstructions of fused three-dimensional MR data set consisting of the T1-weighted (2020/4.38) three-dimensional magnetization-prepared rapid acquisition gradient-echo data set (dark slices in b) and the hybrid data set (bright slices in b). Pink lines show the manually segmented tumor border plotted by a neurosurgeon (O.G. or C.N.) for surgical planning. Yellow dots denote the site of stereotactic biopsies from a transition area of less metabolic change (biopsy 1) and an area with maximal metabolic changes (biopsy 2). (c, d) Histopathologic results of (c) biopsy 1 and (d) biopsy 2. (e, f) Graphs of spectra fitted by LCModel (red line) of voxel positions corresponding to the locus of (e) biopsy 1 and (f) biopsy 2. The absolute values of Cho, Cr, and tNAA are superimposed. In this astrocytoma, increased tumor infiltration corresponds to an elevation of Cho and a decrease of tNAA and Cr. FWHM in the spectra is 0.038 and 0.054 ppm in e and f, respectively, and signal-to-noise ratio is 2.

View larger version (91K): [in this window] [in a new window] [Download PPT slide]   Figure 4b: Patient 6. (a, b) Screen shots from the frameless stereotactic system of (a) transverse and (b) sagittal reconstructions of fused three-dimensional MR data set consisting of the T1-weighted (2020/4.38) three-dimensional magnetization-prepared rapid acquisition gradient-echo data set (dark slices in b) and the hybrid data set (bright slices in b). Pink lines show the manually segmented tumor border plotted by a neurosurgeon (O.G. or C.N.) for surgical planning. Yellow dots denote the site of stereotactic biopsies from a transition area of less metabolic change (biopsy 1) and an area with maximal metabolic changes (biopsy 2). (c, d) Histopathologic results of (c) biopsy 1 and (d) biopsy 2. (e, f) Graphs of spectra fitted by LCModel (red line) of voxel positions corresponding to the locus of (e) biopsy 1 and (f) biopsy 2. The absolute values of Cho, Cr, and tNAA are superimposed. In this astrocytoma, increased tumor infiltration corresponds to an elevation of Cho and a decrease of tNAA and Cr. FWHM in the spectra is 0.038 and 0.054 ppm in e and f, respectively, and signal-to-noise ratio is 2.

View larger version (159K): [in this window] [in a new window] [Download PPT slide]   Figure 4c: Patient 6. (a, b) Screen shots from the frameless stereotactic system of (a) transverse and (b) sagittal reconstructions of fused three-dimensional MR data set consisting of the T1-weighted (2020/4.38) three-dimensional magnetization-prepared rapid acquisition gradient-echo data set (dark slices in b) and the hybrid data set (bright slices in b). Pink lines show the manually segmented tumor border plotted by a neurosurgeon (O.G. or C.N.) for surgical planning. Yellow dots denote the site of stereotactic biopsies from a transition area of less metabolic change (biopsy 1) and an area with maximal metabolic changes (biopsy 2). (c, d) Histopathologic results of (c) biopsy 1 and (d) biopsy 2. (e, f) Graphs of spectra fitted by LCModel (red line) of voxel positions corresponding to the locus of (e) biopsy 1 and (f) biopsy 2. The absolute values of Cho, Cr, and tNAA are superimposed. In this astrocytoma, increased tumor infiltration corresponds to an elevation of Cho and a decrease of tNAA and Cr. FWHM in the spectra is 0.038 and 0.054 ppm in e and f, respectively, and signal-to-noise ratio is 2.

View larger version (153K): [in this window] [in a new window] [Download PPT slide]   Figure 4d: Patient 6. (a, b) Screen shots from the frameless stereotactic system of (a) transverse and (b) sagittal reconstructions of fused three-dimensional MR data set consisting of the T1-weighted (2020/4.38) three-dimensional magnetization-prepared rapid acquisition gradient-echo data set (dark slices in b) and the hybrid data set (bright slices in b). Pink lines show the manually segmented tumor border plotted by a neurosurgeon (O.G. or C.N.) for surgical planning. Yellow dots denote the site of stereotactic biopsies from a transition area of less metabolic change (biopsy 1) and an area with maximal metabolic changes (biopsy 2). (c, d) Histopathologic results of (c) biopsy 1 and (d) biopsy 2. (e, f) Graphs of spectra fitted by LCModel (red line) of voxel positions corresponding to the locus of (e) biopsy 1 and (f) biopsy 2. The absolute values of Cho, Cr, and tNAA are superimposed. In this astrocytoma, increased tumor infiltration corresponds to an elevation of Cho and a decrease of tNAA and Cr. FWHM in the spectra is 0.038 and 0.054 ppm in e and f, respectively, and signal-to-noise ratio is 2.

View larger version (74K): [in this window] [in a new window] [Download PPT slide]   Figure 4e: Patient 6. (a, b) Screen shots from the frameless stereotactic system of (a) transverse and (b) sagittal reconstructions of fused three-dimensional MR data set consisting of the T1-weighted (2020/4.38) three-dimensional magnetization-prepared rapid acquisition gradient-echo data set (dark slices in b) and the hybrid data set (bright slices in b). Pink lines show the manually segmented tumor border plotted by a neurosurgeon (O.G. or C.N.) for surgical planning. Yellow dots denote the site of stereotactic biopsies from a transition area of less metabolic change (biopsy 1) and an area with maximal metabolic changes (biopsy 2). (c, d) Histopathologic results of (c) biopsy 1 and (d) biopsy 2. (e, f) Graphs of spectra fitted by LCModel (red line) of voxel positions corresponding to the locus of (e) biopsy 1 and (f) biopsy 2. The absolute values of Cho, Cr, and tNAA are superimposed. In this astrocytoma, increased tumor infiltration corresponds to an elevation of Cho and a decrease of tNAA and Cr. FWHM in the spectra is 0.038 and 0.054 ppm in e and f, respectively, and signal-to-noise ratio is 2.

View larger version (71K): [in this window] [in a new window] [Download PPT slide]   Figure 4f: Patient 6. (a, b) Screen shots from the frameless stereotactic system of (a) transverse and (b) sagittal reconstructions of fused three-dimensional MR data set consisting of the T1-weighted (2020/4.38) three-dimensional magnetization-prepared rapid acquisition gradient-echo data set (dark slices in b) and the hybrid data set (bright slices in b). Pink lines show the manually segmented tumor border plotted by a neurosurgeon (O.G. or C.N.) for surgical planning. Yellow dots denote the site of stereotactic biopsies from a transition area of less metabolic change (biopsy 1) and an area with maximal metabolic changes (biopsy 2). (c, d) Histopathologic results of (c) biopsy 1 and (d) biopsy 2. (e, f) Graphs of spectra fitted by LCModel (red line) of voxel positions corresponding to the locus of (e) biopsy 1 and (f) biopsy 2. The absolute values of Cho, Cr, and tNAA are superimposed. In this astrocytoma, increased tumor infiltration corresponds to an elevation of Cho and a decrease of tNAA and Cr. FWHM in the spectra is 0.038 and 0.054 ppm in e and f, respectively, and signal-to-noise ratio is 2.

View larger version (33K): [in this window] [in a new window] [Download PPT slide]   Figure 5a: Graphs depict correlations of (a, b) the changes in molar concentration of (a) tNAA and (b) Cho and (c) changes of metabolite ratio Cho/tNAA, with corresponding histopathologic evaluation of 76 biopsy samples in 26 patients. Appropriate coefficients of correlation are superimposed. For all regressions, P < .001.

View larger version (30K): [in this window] [in a new window] [Download PPT slide]   Figure 5b: Graphs depict correlations of (a, b) the changes in molar concentration of (a) tNAA and (b) Cho and (c) changes of metabolite ratio Cho/tNAA, with corresponding histopathologic evaluation of 76 biopsy samples in 26 patients. Appropriate coefficients of correlation are superimposed. For all regressions, P < .001.

View larger version (29K): [in this window] [in a new window] [Download PPT slide]   Figure 5c: Graphs depict correlations of (a, b) the changes in molar concentration of (a) tNAA and (b) Cho and (c) changes of metabolite ratio Cho/tNAA, with corresponding histopathologic evaluation of 76 biopsy samples in 26 patients. Appropriate coefficients of correlation are superimposed. For all regressions, P < .001.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

The use of multivoxel MR spectroscopy for differential diagnosis and the correlation of quantitative tumor cell morphologic characteristics and metabolite concentration has been the subject of several studies in the past. McKnight and colleagues (49) presented results of a study of 68 patients with glioma in which three-dimensional MR spectroscopic imaging data were correlated to histopathologic specimens. These authors showed that by calculating a Cho-to-NAA index, or CNI, and establishing a threshold value of 2.5 they could distinguish between tumor tissue and a composite tissue consisting of edema and normal, necrotic, and gliomatous tissue areas, with a sensitivity of 90% and a specificity of 86%. The CNI values of tissue samples without tumor infiltration showed a significant difference in comparison with tissue samples containing tumor infiltration of World Health Organization grade II (P < .03), grade III (P < .005), and grade IV (P < .01). CNI values in each group of malignancy showed no significant difference, however. The authors did not correlate the CNI values and extent of tumor infiltration. A limiting feature of their study was a rather low spatial resolution of 1 cm³ nominal voxel size. In our opinion, most important was the possible lack of accuracy between the biopsy sampling sites and the MR spectroscopy voxel positions, which were not coregistered but were noted on hardcopy MR images and presented to the surgeon during the operation. In our study, we coregistered biopsy sampling site and high-spatial-resolution MR spectroscopy voxels in a common stereotactic space. These distinctions in methods may explain the differences in our results.

In a study by Vuori et al (31), MR spectroscopy was used for differentiation between low-grade gliomas (10 patients) and focal cortical developmental malformations (eight patients). In addition to a more pronounced increase of Cho and decrease of NAA in low-grade gliomas than in focal cortical developmental malformations, Vuori et al found changes in Cho and Cr levels helpful for differentiation between grade II astrocytomas (four patients) and grade II oligodendrogliomas or oligoastrocytomas (six patients). Grade II astrocytomas showed a modest increase in Cho (69%) and decrease in Cr (–27%), whereas grade II oligodendrogliomas and oligoastrocytomas had a pronounced Cho increase (149%) and Cr increase (58%) compared with age- and sex-matched control subjects. These differences were without overlap between the subgroups, but Vuori et al were not able to define threshold values for differentiation. Besides, they found no significant differences for the metabolite ratios NAA/Cho, NAA/Cr, and Cho/Cr between grade II astrocytomas and grade II oligodendrogliomas or oligoastrocytomas. Apart from the fact that they investigated only low-grade gliomas, the authors themselves pointed out that the limitation of the study was a nominal voxel size of 1.5 cm³. In contrast, in our study we were able to show that applying proton MR spectroscopic imaging with high spatial resolution (nominal voxel size of 0.45 cm³) can be considered a prerequisite to elicit distinct differences that are helpful for preoperative grading and tissue diagnosis of gliomas of grades II and III.

Law and coworkers (33) investigated sensitivity, specificity, and positive and negative predictive values of MR spectroscopic imaging and conventional MR imaging in a study of 160 gliomas. The processing of the MR spectroscopic imaging data was executed by means of calculation of the maxima for Cho/Cr and Cho/NAA, as well as the minima for NAA/Cr. As expected, the most significant results (P < .001) for a differentiation of low- versus high-grade gliomas were obtained from the Cho/NAA ratios. Cho/NAA values were 0.60–6.80 for low-grade tumors and 0.53–28.90 for high-grade tumors, which resulted in no significant differences between the two groups. Significant differences between the two tumor groups were shown for Cho/Cr (P < .012), Cho/NAA (P < .001), and NAA/Cr (P < .004). The authors also defined four threshold values (0.75–1.75) for Cho/NAA between the tumor groups, and the parameters mentioned above for each group were calculated. The threshold value of 0.75 for Cho/NAA resulted in a sensitivity of 96.7%, a specificity of 10.0%, a positive predictive value of 76.3%, and a negative predictive value of 50.0%. For conventional MR imaging, sensitivity was 72.5%, specificity was 65.0%, positive predictive value was 86.1%, and negative predictive value was 44.1%, demonstrating a potentially valuable contribution of MR spectroscopic imaging for preoperative tumor grading.

Yang et al (32) revealed significant differences between low-grade and high-grade gliomas by using MR spectroscopic imaging. They investigated 17 patients with gliomas of grades II–IV and calculated the maxima of Cho/NAA and Cho/Cr and the minima of NAA/Cr in these tumors. Significantly higher values for Cho/NAA (P < .008) and Cho/Cr (P < .03) and significantly lower values for NAA/Cr (P < .009) were found in the group of 13 patients with high-grade gliomas compared with the group of four patients with low-grade gliomas. The results of Law et al (33) and Yang et al (32) are in accordance with our findings, although both these studies have the shortcomings of a rather low nominal resolution of 1.5–2.0 cm³ and, thus, an insufficient spatial resolution of the MR spectroscopic imaging experiment. In contrast to our study, in both these other studies the authors did not calculate molar metabolite concentrations. They used total Cr level as an internal reference, which can produce misleading results (50).

Nafe et al (51) showed, in a study of 46 paitents with gliomas of grades II–IV, that an almost linear (r = 0.444) and significant (P = .002) correlation exists between Cho/total Cr ratio and tumor density in biopsy specimens. They did not calculate absolute concentrations, however, but used metabolite ratios of the signal intensity of total Cr from a reference voxel on the contralateral side. Also, Nafe et al did not investigate the correlation between NAA concentration and tumor cell density. The correlation coefficient (r = 0.444) was considerably lower than the value found in our study (r = 0.743). A reason for the difference may be that only single-voxel MR spectroscopy was performed with voxel sizes of 2.2–12.7 cm³ (ie, partial volume effects). Another reason may be the usage of total Cr as an internal reference.

Our results for absolute metabolite concentrations in the normal brain in patients and control subjects largely agree with published results. The differences in Cho/tNAA ratio between grades II and III gliomas in the tumor center clearly show the advantage of the increased spatial resolution of the MR spectroscopic imaging experiment and, thereby, reduced partial volume effects. The distinctly higher value of the maximum Cr concentration in grade III astrocytomas compared with grade III oligodendrogliomas and oligoastrocytomas may correlate with an increased energy metabolism caused by an elevated rate of growth of the former compared with the latter.

The findings of our study for the correlation of the changes in tNAA with histopathologic findings from biopsy samples are in accordance with the infiltrative nature of gliomas. Glioma cells lead to a displacement or destruction of neurons, and it is therefore to be expected that tNAA, as a neuronal marker, will decrease with increasing tumor infiltration. We found a significant negative linear (r = –0.905, P < .001) correlation of the tNAA concentration depending on the level of tumor infiltration across 76 biopsy samples. The exponential correlation of the Cho concentration with the percentage of tumor cells could be explained by two different mechanisms leading to an increase of Cho owing to increasing tumor infiltration. Cho is regarded as a marker for phospholoipid turnover in the cell membrane (18). As discussed by Vuori et al (31), on the one hand, the increased anabolism of malignant cells in neoplastic lesions leads to a rapid cell membrane proliferation and thus to a real increase in the Cho concentration. On the other hand, the membrane breakdown of cells of the brain parenchyma (catabolism) is associated with an increased mobility of the Cho of the breakdown products. This leads to a longer T2 relaxation time and an increased detectability for Cho in the MR spectroscopic experiment.

Our study was limited by the fact that we performed MR spectroscopic imaging as a two-dimensional experiment. To cover the whole or at least the bulk of the tumor volume, it will be necessary to use a three-dimensional MR spectroscopic imaging sequence with even higher spatial resolution (52) although this leads to the drawback of longer acquisition times, which are not always acceptable for patients. A strategy to overcome this is the so-called parallel imaging technique (sensitivity encoding or generalized autocalibrating partially parallel acquisitions) (53,54). However, due to the reduction in signal-to-noise involved with the imaging time reduction, this technique is actually more suitable at higher magnetic field strengths (3 T or higher).

In conclusion, results of our study demonstrated that proton MR spectroscopic imaging with high spatial resolution, segmentation, and absolute quantification of metabolic changes provides information for preoperative grading of gliomas.

	FOOTNOTES

 

Abbreviations: Cho = choline-containing compounds • Cr = creatine • FWHM = full width at half maximum • NAA = N-acetylaspartate • tNAA = total NAA

Authors stated no financial relationship to disclose.

Author contributions: Guarantors of integrity of entire study, E.M., O.G.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, A.S., S.G., R.F., T.H., B.T., E.M., O.G.; clinical studies, C.N., O.G.; experimental studies, A.S., R.B., O.G.; statistical analysis, A.S.; and manuscript editing, A.S., S.G., C.N., R.F., T.H., B.T., E.M., O.G.

	References

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 

1. Dowling C, Bollen AW, Noworolski SM, et al. Preoperative proton MR spectroscopic imaging of brain tumors: correlation with histopathologic analysis of resection specimens. AJNR Am J Neuroradiol 2001;22:604–612.[Abstract/Free Full Text]
2. Kondziolka D, Lunsford LD, Martinez AJ. Unreliability of contemporary neurodiagnostic imaging in evaluating suspected adult supratentorial (low-grade) astrocytoma. J Neurosurg 1993;79:533–536.[Medline]
3. Alesch F, Pappaterra J, Trattnig S, Koos WT. The role of stereotactic biopsy in radiosurgery. Acta Neurochir Suppl 1995;63:20–24.[Medline]
4. Mihara F, Numaguchi Y, Rothman M, Sato S, Fiandaca MS. MR imaging of adult supratentorial astrocytomas: an attempt of semi-automatic grading. Radiat Med 1995;13:5–9.[Medline]
5. Bernstein M, Parrent AG. Complications of CT-guided stereotactic biopsy of intra-axial brain lesions. J Neurosurg 1994;81:165–168.[Medline]
6. Field M, Witham TF, Flickinger JC, Kondziolka D, Lunsford LD. Comprehensive assessment of hemorrhage risks and outcomes after stereotactic brain biopsy. J Neurosurg 2001;94:545–551.[Medline]
7. Kreth FW, Muacevic A, Medele R, Bise K, Meyer T, Reulen HJ. The risk of haemorrhage after image guided stereotactic biopsy of intra-axial brain tumours: a prospective study. Acta Neurochir (Wien) 2001;143:539–545.[CrossRef][Medline]
8. Sawin PD, Hitchon PW, Follett KA, Torner JC. Computed imaging-assisted stereotactic brain biopsy: a risk analysis of 225 consecutive cases. Surg Neurol 1998;49:640–649.[CrossRef][Medline]
9. Yu X, Liu Z, Tian Z, et al. Stereotactic biopsy for intracranial space-occupying lesions: clinical analysis of 550 cases. Stereotact Funct Neurosurg 2000;75:103–108.[CrossRef][Medline]
10. Christy PS, Tervonen O, Scheithauer BW, Forbes GS. Use of a neural network and a multiple regression model to predict histologic grade of astrocytoma from MRI appearances. Neuroradiology 1995;37:89–93.[Medline]
11. Alger JR, Frank JA, Bizzi A, et al. Metabolism of human gliomas: assessment with H-1 MR spectroscopy and F-18 fluorodeoxyglucose PET. Radiology 1990;177:633–641.[Abstract/Free Full Text]
12. Frahm J, Bruhn H, Hanicke W, Merboldt KD, Mursch K, Markakis E. Localized proton NMR spectroscopy of brain tumors using short-echo time STEAM sequences. J Comput Assist Tomogr 1991;15:915–922.[Medline]
13. Majos C, Alonso J, Aguilera C, et al. Adult primitive neuroectodermal tumor: proton MR spectroscopic findings with possible application for differential diagnosis. Radiology 2002;225:556–566.[Abstract/Free Full Text]
14. 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.[Medline]
15. Peeling J, Sutherland G. High-resolution 1H NMR spectroscopy studies of extracts of human cerebral neoplasms. Magn Reson Med 1992;24:123–136.[Medline]
16. Ott D, Hennig J, Ernst T. Human brain tumors: assessment with in vivo proton MR spectroscopy. Radiology 1993;186:745–752.[Abstract/Free Full Text]
17. Michaelis T, Merboldt KD, Bruhn H, Hanicke W, Frahm J. Absolute concentrations of metabolites in the adult human brain in vivo: quantification of localized proton MR spectra. Radiology 1993;187:219–227.[Abstract/Free Full Text]
18. Miller BL. A review of chemical issues in 1H NMR spectroscopy: N-acetyl-L-aspartate, creatine and choline. NMR Biomed 1991;4:47–52.[Medline]
19. Urenjak J, Williams SR, Gadian DG, Noble M. Proton nuclear magnetic resonance spectroscopy unambiguously identifies different neural cell types. J Neurosci 1993;13:981–989.[Abstract]
20. Kemp GJ. Non-invasive methods for studying brain energy metabolism: what they show and what it means. Dev Neurosci 2000;22:418–428.[CrossRef][Medline]
21. Croteau D, Scarpace L, Hearshen D, et al. Correlation between magnetic resonance spectroscopy imaging and image-guided biopsies: semiquantitative and qualitative histopathological analyses of patients with untreated glioma. Neurosurgery 2001;49:823–829.[CrossRef][Medline]
22. Howe FA, Barton SJ, Cudlip SA, et al. Metabolic profiles of human brain tumors using quantitative in vivo 1H magnetic resonance spectroscopy. Magn Reson Med 2003;49:223–232.[CrossRef][Medline]
23. Preul MC, Caramanos Z, Collins DL, et al. Accurate, noninvasive diagnosis of human brain tumors by using proton magnetic resonance spectroscopy. Nat Med 1996;2:323–325.[CrossRef][Medline]
24. Kwock L, Smith JK, Castillo M, et al. Clinical applications of proton MR spectroscopy in oncology. Technol Cancer Res Treat 2002;1:17–28.[Medline]
25. Nelson SJ, McKnight TR, Henry RG. Characterization of untreated gliomas by magnetic resonance spectroscopic imaging. Neuroimaging Clin N Am 2002;12:599–613.[CrossRef][Medline]
26. Rock JP, Hearshen D, Scarpace L, et al. Correlations between magnetic resonance spectroscopy and image-guided histopathology, with special attention to radiation necrosis. Neurosurgery 2002;51:912–919.[CrossRef][Medline]
27. Law M, Cha S, Knopp EA, Johnson G, Arnett J, Litt AW. High-grade gliomas and solitary metastases: differentiation by using perfusion and proton spectroscopic MR imaging. Radiology 2002;222:715–721.[Abstract/Free Full Text]
28. Tate AR, Majos C, Moreno A, Howe FA, Griffiths JR, Arus C. Automated classification of short echo time in in vivo 1H brain tumor spectra: a multicenter study. Magn Reson Med 2003;49:29–36.[CrossRef][Medline]
29. Galanaud D, Chinot O, Nicoli F, et al. Use of proton magnetic resonance spectroscopy of the brain to differentiate gliomatosis cerebri from low-grade glioma. J Neurosurg 2003;98:269–276.[Medline]
30. Howe FA, Opstad KS. 1H MR spectroscopy of brain tumours and masses. NMR Biomed 2003;16:123–131.[CrossRef][Medline]
31. Vuori K, Kankaanranta L, Hakkinen AM, et al. Low-grade gliomas and focal cortical developmental malformations: differentiation with proton MR spectroscopy. Radiology 2004;230:703–708.[Abstract/Free Full Text]
32. Yang D, Korogi Y, Sugahara T, et al. Cerebral gliomas: prospective comparison of multivoxel 2D chemical-shift imaging proton MR spectroscopy, echoplanar perfusion and diffusion-weighted MRI. Neuroradiology 2002;44:656–666.[CrossRef][Medline]
33. Law M, Yang S, Wang H, et al. Glioma grading: sensitivity, specificity, and predictive values of perfusion MR imaging and proton MR spectroscopic imaging compared with conventional MR imaging. AJNR Am J Neuroradiol 2003;24:1989–1998.[Abstract/Free Full Text]
34. Herminghaus 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.[Medline]
35. Stadlbauer A, Moser E, Gruber S, Nimsky C, Fahlbusch R, Ganslandt O. Integration of biochemical images of a tumor into frameless stereotaxy achieved using a magnetic resonance imaging/magnetic resonance spectroscopy hybrid data set. J Neurosurg 2004;101:287–294.[Medline]
36. Stadlbauer A, Moser E, Gruber S, et al. Improved delineation of brain tumors: an automated method for segmentation based on pathologic changes of (1)H-MRSI metabolites in gliomas. Neuroimage 2004;23:454–461.[CrossRef][Medline]
37. Vikhoff-Baaz B, Starck G, Ljungberg M, Lagerstrand K, Forssell-Aronsson E, Ekholm S. Effects of k-space filtering and image interpolation on image fidelity in (1)H MRSI. Magn Reson Imaging 2001;19:1227–1234.[CrossRef][Medline]
38. Jacobs MA, Horska A, van Zijl PC, Barker PB. Quantitative proton MR spectroscopic imaging of normal human cerebellum and brain stem. Magn Reson Med 2001;46:699–705.[CrossRef][Medline]
39. Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993;30:672–679.[Medline]
40. Provencher SW. LCModel & LCMgui user's manual. http://s-provencher.com/pages/lcmodel.shtml. Updated July 7, 2003. Accessed July 14, 2003.
41. McLean MA, Woermann FG, Barker GJ, Duncan JS. Quantitative analysis of short echo time (1)H-MRSI of cerebral gray and white matter. Magn Reson Med 2000;44:401–411.[CrossRef][Medline]
42. Isobe T, Matsumura A, Anno I, et al. Quantification of cerebral metabolites in glioma patients with proton MR spectroscopy using T2 relaxation time correction. Magn Reson Imaging 2002;20:343–349.