| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University of North Carolina at Chapel Hill, Chapel
Hill, NC 27599.
2 Present address: Department of Radiology, Medical University of South
Carolina, 169 Ashley Ave., PO Box 250322, Charleston, SC 29425.
3 Division of Neurosurgery, University of North Carolina at Chapel Hill, Chapel
Hill, NC 27599.
Received July 8, 2005;
accepted after revision January 19, 2006.
J. K. Smith is a recipient of a Philips Medical Systems/RSNA seed grant.
250322, Charleston, SC 29425.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. PWI and MRS studies of 22 patients with histologically proven oligodendroglioma or oligoastrocytoma (13 low-grade and nine anaplastic tumors) were retrospectively reviewed. PWI of 14 subjects was performed with a dynamic contrast-enhanced susceptibility-weighted echo-planar technique. Intratumoral relative cerebral blood volume ratio was calculated and normalized to the same value in contralateral normal-appearing white matter. Multivoxel MRS was performed with a point-resolved spectroscopy sequence at a TE of 135 milliseconds in 20 patients and with the addition of a TE of 30 seconds in 17 patients. MRS data were expressed as intratumoral metabolite ratios (choline to creatine [Cho/Cr], choline to N-acetyl aspartate, N-acetyl aspartate to creatine, and myoinositol to creatine).
RESULTS. Relative cerebral blood volume ratios were significantly different (p = 0.004) between low-grade (1.61 ± 1.20) and high-grade tumors (5.45 ± 1.96). The optimal relative cerebral blood volume ratio cutoff value in identification of anaplastic oligodendroglial tumors was 2.14. Analysis of MRS data showed significantly higher Cho/Cr ratios (p = 0.002) in high-grade than in low-grade tumors. A Cho/Cr ratio cutoff value of 2.33 had the highest accuracy in identification of high-grade tumors.
CONCLUSION. Relative cerebral blood volume measurement and MRS are helpful in differentiating low-grade from anaplastic oligodendroglial tumors.
Keywords: brain • MR technique • MRI • oncologic imaging • perfusion-weighted MRI
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Advanced MRI techniques, such as perfusion-weighted imaging (PWI), may be helpful in preoperative determination of tumor grade, which may affect surgical planning. Relative cerebral blood volume (rCBV) measurements may improve tumor grading [12]. In addition, proton MRS may be helpful in predicting tumor grade [13]. Perfusion measurements (particularly rCBV) and MRS used together improve characterization of brain tumors [14-17]. The role of rCBV and MRS in the evaluation of oligodendroglial tumors has been investigated [18-20], but we believe further evaluation is needed. Our purpose was to determine whether rCBV and MRS help differentiate low- from high-grade oligodendroglial tumors.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Conventional MRI
All patients underwent preoperative MRI on 1.5-T units (Vision or Sonata,
Siemens Medical Solutions). The conventional MRI protocol consisted of a
three-plane localizer sequence, sagittal T1-weighted images, axial 3D
T1-weighted spoiled gradient-recalled acquisition in the steady state images,
axial FLAIR and fast spin-echo T2-weighted sequences, and after acquisition of
dynamic susceptibility contrast-enhanced MR images, gadolinium-enhanced axial
3D spoiled gradient-recalled acquisition in the steady state and sagittal and
coronal T1-weighted images.
Measurement of rCBV
A dynamic susceptibility contrast-enhanced technique was used to obtain PWI
data on 14 subjects (seven with low-grade and seven with anaplastic tumors).
Single-shot GRE echo-planar images were acquired before, during, and after
rapid administration of a contrast bolus (TR/TE, 2,000/54; flip angle,
60°; matrix size, 128 x 128; field of view, 230 mm). The study
consisted of sequential acquisition of 40 single-shot echo-planar images of 20
sections each, for a total of 800 images. A single standard dose of
gadopentetate dimeglumine contrast agent (0.1 mmol/kg) was administered IV
through an 18-gauge IV line at a rate of 5 mL/s, starting at the end of the
fifth acquisition and followed by a bolus injection of saline solution.
Perfusion data were postprocessed with Perfusion Task Card software
(Massachusetts General Hospital, A. A. Martinos Center for Biomedical Imaging)
[21-23]
on a free-standing workstation, and rCBV maps were calculated offline. Passage
of contrast material through the vascular compartment causes local magnetic
field inhomogeneity that is proportional to blood volume. This inhomogeneity
in the magnetic field causes a decrease in signal intensity received from
tissues. With a singular value decomposition deconvolution approach,
quantitative measures of tissue perfusion are obtained
[21,
22]. We obtained the time
course of arterial input function by manually choosing voxels on the
ipsilateral middle cerebral artery in each patient. For each voxel the time
versus intensity curves for the dynamic images were converted into a curve of
change in T2, or 
R2(1/T2), and then tissue concentration
versus time. Maps of rCBV were determined by numerical integration of the area
under the tissue concentration versus time curve from the arrival of the
contrast bolus to tracer recirculation
[24].
Review of rCBV maps was conducted at the workstation by one of the authors, blinded at the time of analysis to clinical and MRS data. Color rCBV maps were available for identification of areas of abnormal increased cerebral blood volume. Unenhanced and contrast-enhanced T1-weighted and T2-weighted MR images were accessible during analysis of rCBV maps. Unprocessed perfusion images were used to ensure that regions of interest (ROIs) were not placed over blood vessels. Within the tumors, four to six ROIs with the highest rCBV values were selected, and the highest rCBV value was recorded. Intratumoral ROIs measured approximately 4-6 mm2. Selection of very small ROIs allowed acquisition of measurements from tumor sites with the highest rCBV signal intensities, the lowest SD, and the least partial volume with neighboring vascular structures and CSF-containing spaces [25]. Mean maximal rCBV values were calculated for the contralateral normal-appearing white matter and used as an internal standard. Normalized rCBV ratios were computed as ratios between maximal rCBV within the tumor and contralateral normal-appearing white matter because this measurement has been proved to have the best interobserver and intraobserver reproducibility [25].
MR Spectroscopy
Multivoxel spectroscopic imaging was performed after administration of
gadopentetate dimeglumine as described in the literature
[26]. The spectroscopic volume
of interest was based on the T1-weighted contrast-enhanced images. A
point-resolved spectroscopy sequence with a TE of 135 milliseconds was used
for MRS studies of 20 patients (12 with low-grade and eight with anaplastic
tumors). Imaging of 17 patients (nine with low-grade and eight with anaplastic
tumors) also included a study with a TE of 30 milliseconds. The nominal volume
size in these spectroscopic imaging studies varied between 0.5 and 1.0 cc. The
acquisition time for the spectroscopic imaging studies varied between 4 and 8
minutes. Spectroscopic data were processed with an offline workstation
(Leonardo, Siemens Medical Solutions) with software provided by the MRI vendor
(Syngo 2004A) as follows: The data were zero-filled to 2,048 points from
1,024; an exponential multiplier of 4 Hz was applied; a Fourier transform of
the time domain signal obtained was performed for acquisition of the frequency
domain signal; a polynomial baseline correct function was applied; the
baseline corrected spectrum was phase corrected, in most cases with only
zero-order phase correction; and the final spectrum was curve fitted with a
curve-fitting program supplied by the vendor for acquisition of the peak area
under the fitted resonance peak.
After baseline correction, metabolite peaks were assigned as follows: myoinositol (Mi), 3.56 ppm; choline (Cho), 3.2 ppm; creatine (Cr), 3.03 ppm; and N-acetyl aspartate (NAA), 2 ppm. Lactate was identified by its characteristic doublet at 1.32 ppm, inverted at a TE of 135 milliseconds. Long and short TE data had been acquired concomitantly for 21 patients (eight with anaplastic and 13 with low-grade tumors). Long TE MRS data were not available in one case (low-grade tumor), and short TE MRS data were not available in four cases (low-grade tumors).
Spectroscopic data obtained at a TE of 135 milliseconds (12 low-grade and eight anaplastic tumors) were expressed as intratumoral metabolite ratios. In multivolume analysis, volumes with the highest Cho/Cr and Cho/NAA ratios and the lowest NAA/Cr ratio were selected to represent the tumor. Presence of lactate peaks anywhere within a tumor was recorded from the data set obtained at a TE of 135 milliseconds. In 17 patients (nine with low-grade and eight with anaplastic tumors), myoinositol level, recorded with MRS at short TE, was quantified and expressed as intratumoral maximal Mi/Cr ratio. We did not quantify peak areas for choline or N-acetyl aspartate from the data set obtained at short TE. A TE of 30 milliseconds does not allow full relaxation of these metabolites and thus height and area would be underestimated.
|
|
Statistical Analysis
Tumors pathologically characterized as low-grade oligodendroglioma or
oligoastrocytoma and anaplastic oligodendroglioma or oligoastrocytoma were
evaluated separately. Because of the small size of the sample and lack of
normal distribution of data, we used a nonparametric test. The Mann-Whitney
test was used for statistical comparison of rCBV, Cho/Cr, Cho/NAA, NAA/Cr, and
Mi/Cr between low-grade and anaplastic tumors. Results were considered
statistically significant at p < 0.05. Fisher's exact test was
used to compare the proportions of low-grade and anaplastic tumors when a
lactate peak was present or absent at a TE of 135 milliseconds. Sensitivity,
specificity, and accuracy were calculated for correct identification of
high-grade tumors. We also estimated positive and negative likelihood
ratios.
Binormal receiver operating characteristics (ROC) curves were used to calculate optimal cutoff values for differentiating low- and high-grade oligodendroglial tumors. This statistical technique allows determination of specificity and sensitivity as a function of a threshold value for identification of high-grade tumors. For each test showing potential usefulness (i.e., showing statistically significant difference between high-grade and low-grade tumor groups), we selected cutoff values that provided the lowest C1 and C2 error (C1 error = 1 - [sensitivity + specificity]/2; C2 error = fraction of misclassified tumors). The area under the ROC curve was used as a quantitative measure to compare the relative value of different tests. A larger area under the curve suggests better performance of a test in differentiating high- and low-grade tumors (null hypothesis is area = 0.5). We also assessed whether a combination of tests was more helpful than individual tests in characterization of tumor grade. Statistical analysis was performed with the SPSS statistical package. Binormal ROC curves were calculated with the RockIt program, which covers continuous data (Kurt Rossmann Laboratories for Radiologic Image Research, Department of Radiology, University of Chicago).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
|
|
|
Cho/NAA, NAA/Cr, and Mi/Cr ratios were not significantly different between low-grade and high-grade oligodendroglial tumors. Lactate was present in three anaplastic (one oligodendroglioma and two oligoastrocytomas, two previously treated and one newly diagnosed tumors) but in none of the low-grade tumors (p = 0.49). Figures 2D and 3D show the MRS spectra of low- and high-grade oligodendroglial tumors.
|
|
We also evaluated separately the distribution of rCBV and Cho/Cr ratios in the subset of pure oligodendrogliomas. The average rCBV was 1.75 ± 1.42 for low-grade and 5.49 ± 2.48 for anaplastic oligodendrogliomas (five low-grade and three anaplastic tumors). Although a trend toward higher rCBV ratio in anaplastic than in low-grade oligodendrogliomas was found in this small sample, criteria for statistical significance were not satisfied (p = 0.053). The average Cho/Cr ratios were 2.05 ± 2.54 for low-grade and 2.97 ± 0.56 for anaplastic oligodendrogliomas (eight low-grade and four anaplastic tumors). Cho/Cr ratios were significantly higher in anaplastic than in low-grade pure oligodendrogliomas (p < 0.05). Cho/NAA and NAA/Cr (four anaplastic and eight low-grade tumors) and Mi/Cr (four anaplastic and six low-grade tumors) ratios were not significantly different between pure low-grade and anaplastic oligodendrogliomas. Presence of lactate was found in one anaplastic oligodendroglioma and in none of the low-grade oligodendrogliomas.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
According to the most recent WHO tumor grading system, differentiation between anaplastic and low-grade oligodendroglial tumors relies mainly on the presence of areas of endothelial proliferation or hyperplasia [3, 10]. Unlike other solid tumors, low-grade oligodendroglioma can become very large and still be supplied by the microvasculature of the brain parenchyma. The presence of microvascular proliferation is a hallmark of anaplastic oligodendroglial tumors and is associated with more malignant behavior. Expression of vascular endothelial growth factor, one of the most powerful angiogenesis growth factors, has been proved to be elevated in anaplastic oligodendroglioma [27].
MRI studies have shown that contrast enhancement is not predictive of the presence of anaplasia within oligodendroglioma [11, 28]. Conversely, studies have shown that rCBV measurements may be helpful in improving preoperative grading of glioma [12, 14, 29]. Knopp et al. [12] found mean rCBV ratios of 1.44 for low-grade and 5.07 for high-grade astrocytic gliomas. In a different study by the same group [14], the average rCBV ratios were 2.14 for low-grade glioma and 5.18 for high-grade glioma. Lev et al. [20] used a spin-echo echo-planar technique to calculate rCBV maps and found a significant positive correlation between astrocytic tumor grade and rCBV values. They also concluded that the accuracy of rCBV measurements in predicting tumor grade decreased significantly when oligodendroglioma was included in the analysis. Unlike the investigators in the aforementioned studies, we focused our attention only on oligodendroglial tumors and did not compare them with astrocytomas.
In a 2005 study, Xu et al. [18] examined the role of rCBV in characterization of low-grade versus anaplastic oligodendroglial tumors. The mean normalized rCBV ratio in a series of nine low-grade tumors was 1.70 ± 0.60; the average rCBV ratio in the high-grade tumor group (which included only three patients) was 2.25 ± 0.45. A finding different from ours was that there was no significant difference in rCBV ratio between low- and high-grade tumors. An explanation for the discrepancy between that finding and ours may be that Xu et al. examined only three anaplastic oligodendroglial tumors versus nine low-grade tumors and used data sampling techniques different from ours. Those authors reported sampling rCBV maps with ROIs measuring 20-40 mm2, larger than our ROIs. The use of large ROIs is known to negatively affect sensitivity in the detection of focal areas of increased vascularization in anaplastic neoplasms [25].
Our findings show rCBV measurement is useful in characterizing oligodendroglial tumor grade. Low-grade oligodendroglioma is histologically characterized by the presence of a network of branching capillaries within the stroma, traditionally described as having the appearance of chicken wire, but only anaplastic tumors are characterized by endothelial hyperplasia and tumoral microvascular proliferation [3]. Our results suggest that only the appearance of newly formed tumoral vessels translates into a significant increase in rCBV values. In our series of patients, the sensitivity of rCBV in grading of oligodendroglial tumors was 100%, indicating that all high-grade tumors were correctly classified with the use of a cutoff value of 2.14. The specificity of rCBV measurement was 86%, which implies that with a cutoff value of 2.14, a few low-grade oligodendroglial tumors may be incorrectly classified as high-grade tumors. The consequence of such a classification error would be more aggressive management of some low-grade oligodendroglial tumors with a possible increase in morbidity. In our series, none of the high-grade tumors was missed. Therefore we suggest that rCBV ratios can be used more confidently to exclude the presence of anaplasia within oligodendroglial tumors than vice versa. Analyses of low-grade oligodendroglial tumors have revealed occasional elevated rCBV values [20, 30].
We found elevated maximal rCBV (rCBV ratio, 4.3) in one newly diagnosed low-grade oligodendroglioma. The same patient also had an elevated intratumoral Cho/Cr ratio of 8.3. This patient was treated with partial surgical resection, and the histopathologic diagnosis was low-grade oligodendroglioma. Whenever, as in this case, the entire tumor is not submitted for histopathologic evaluation, it can be argued that anaplastic components and focal areas of microvascular proliferation that might be present within the residual tumor have not been included in the specimen. The rCBV map for this neoplasm was very heterogeneous; the highest rCBV values within a tumoral area were adjacent to the left inferior frontal gyrus, which was not surgically removed so that speech could be preserved. After surgery this patient underwent follow-up at another institution. He received nine of 33 planned radiation treatments and died only 12 months after the initial diagnosis.
Single-voxel and multivoxel proton MRS have been used for assessment and grading of glial tumors [14, 31, 32]. Elevated choline levels indicate increased membrane synthesis and increased cellularity [13, 33]. High cell density is a parameter that characterizes anaplastic oligodendrogliomas [3] analogous to high-grade astrocytic tumors. MRS metabolite ratios such as Cho/Cr have shown good correlation with glial tumor grade [34]. In our patients, Cho/Cr ratio had sensitivity, specificity, and accuracy of 100%, 83.3%, and 90%. With a threshold Cho/Cr value of 2.33, none of the high-grade tumors was misclassified. Conversely, two low-grade tumors had elevated Cho/Cr values (2.7 and 8.3). There are two possible explanations for these findings. According to the WHO classification of tumors, low-grade oligodendroglioma and oligoastrocytoma are defined by the absence of endothelial proliferation and necrosis but can have moderately elevated cell density and therefore elevated choline levels [3]. An alternative explanation is that because these tumors were managed with only partial resection, areas of more cellular and anaplastic tumoral tissues may not have been submitted for histologic analysis. In the case of the first misclassified low-grade oligoastrocytoma (maximal Cho/Cr ratio, 2.7), nine intratumoral volumes were evaluated, and in only three volumes was the Cho/Cr ratio higher than 2. The other misclassified low-grade oligodendroglioma (maximal Cho/Cr ratio, 8.3) had multiple areas of elevation of Cho/Cr ratio and a significantly elevated maximal rCBV ratio of 4.3. The clinical information suggested that MRS and rCBV findings may have been accurate for prognosis in this case. No significant elevation in myoinositol level was found in low-grade oligodendroglial tumors as has been reported for other glial tumors [35].
In a short TE MRS study of the metabolite composition of high- and low-grade oligodendrogliomas, the authors [19] found significant elevation of lipid and lactate values in high-grade tumors and of glutamate and glutamine values in low-grade tumors. We did not analyze the glutamine or glutamate levels of oligodendroglial tumors because we believe that quantification of these metabolites is not accurate or reproducible with the technique we used. In another study, the investigators [18] tested the usefulness of MRS with long and short TE in differentiating low-grade and anaplastic oligodendroglial tumors. Those authors found a significantly higher normalized choline level in high-grade than in low-grade tumors. They also reported that a combination of lipids and lactate was more often present in anaplastic tumors. Lipids and lactate are markers of necrosis, and both are metabolites present in high-grade tumors. MRS studies at a TE of 135 milliseconds allow separate observation of lactate and lipids. At a TE of 30 milliseconds, lipid and lactate peaks cannot be differentiated. In addition, some lipids may be normally detected at short TE and are of uncertain significance. For these reasons, we evaluated only lactate peaks at a TE of 135 milliseconds. In our series, we found lactate peaks in three anaplastic and no low-grade tumors, a finding that was significantly different between groups. Because previously treated patients were included in our study and that by Xu et al. [18], the significance of lactate peaks should be considered with caution [36]. Presence of lactate may be a sequela of treatment and may not imply a higher degree of malignancy.
Our study showed that use of rCBV and Cho/Cr measurements may improve preoperative grading of oligodendroglial tumors. PWI and MRS have the advantage over histopathologic grading of allowing evaluation of the entire lesion in vivo. Results of ROC analysis suggest that rCBV measurement may be slightly more accurate than MRS in the identification of high-grade tumors (area under the curve, 0.96 for rCBV and 0.91 for Cho/Cr).
We recognize that our study was limited by being a retrospective and not a prospective evaluation of clinical and MR data. Given this limitation and the size of our sample, we do not intend to suggest that the sensitivity, specificity, and accuracy estimated in this study should be used for therapeutic decisions. Our goal was to evaluate the potential usefulness of noninvasive MR techniques in the characterization of oligodendroglial tumors. Only results of larger prospective studies would help determine clinically relevant negative and positive predictive values for Cho/Cr and rCBV ratios. Another limitation of the study was that rCBV and MRS data were generated by single observers.
Because of the design of our study, the findings do not shed light on the debate on preoperative differentiation of astrocytic from oligodendroglial tumors. According to results of one study [30], the presence of elevated rCBV values in a tumor that on conventional MRI has the appearance of a low-grade glioma should suggest the diagnosis of low-grade oligodendroglioma. Another limitation of our study was that preoperative MRS data and rCBV values were obtained in some cases of oligodendroglial tumor for which the patient had been previously treated. We decided to include all pathologically confirmed recurrent oligodendroglial tumors because of the relatively low prevalence of this subset of gliomas. We did not assess conventional MRI features in our series of oligodendroglial tumors, because this information has been previously reported [11]. Another potential limitation of our study was the inclusion of pure oligodendroglial and mixed oligoastrocytic tumors. Precise diagnostic criteria for categorization of oligodendroglioma and oligoastrocytoma have not been established [3]. The clinical and histologic distinction between oligodendroglioma and oligoastrocytoma and pure astroglial tumors is extremely important, because oligoastrocytoma, like oligodendroglioma, responds favorably to chemotherapy [37]. Studies with rodent models of brain tumors have shown that a progenitor cell of the oligodendroglial lineage can differentiate into oligodendrocytes or type 2 astrocytoma. The hypothesis that oligodendroglioma and oligoastrocytoma are derived from an analogous common precursor in humans has been debated [3]. Nevertheless, an exploratory analysis focusing only on the subset of pure oligodendrogliomas has shown a trend toward higher rCBV in anaplastic versus low-grade oligodendroglioma (three anaplastic and five low-grade tumors) and significantly higher Cho/Cr in anaplastic than low-grade oligodendroglioma (four anaplastic and eight low-grade tumors).
Studies have shown that loss of genes in chromosomes 1p and 19q is a unique genetic feature of most oligodendroglial tumors. These genetic traits have been consistently associated with a good response to chemotherapy and a better prognosis in patients with oligodendroglioma [38]. The 1p and 19q status was not tested in our series of tumors; therefore the incidence of 1p and 19q heterozygosity in low-versus high-grade tumors and oligodendroglioma versus oligoastrocytoma is not known. Mean duration of follow-up in our series of patients was not adequate for assessment of the correlation between imaging data and length of survival. Longitudinal longterm studies are needed for assessment of the correlations between rCBV and metabolite ratios and patient outcome.
We conclude that in this evaluation of the role of rCBV maps and MRS in the characterization of oligodendroglial tumors, cutoff values of 2.14 and 2.33 for rCBV and Cho/Cr ratios, respectively, yielded the highest accuracy in differentiation of highfrom low-grade oligodendroglial tumors. Cho/Cr and rCBV ratios provided information on vascularity and metabolic composition of oligodendroglial tumors that may contribute to noninvasive differentiation of low-grade from high-grade tumors.
Acknowledgments
We thank Gregory Sorensen and Thomas Benner, Massachusetts General
Hospital, for providing the perfusion software package used in this study.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Bisdas, M. Kirkpatrick, P. Giglio, C. Welsh, M.V. Spampinato, and Z. Rumboldt Cerebral Blood Volume Measurements by Perfusion-Weighted MR Imaging in Gliomas: Ready for Prime Time in Predicting Short-Term Outcome and Recurrent Disease? AJNR Am. J. Neuroradiol., April 1, 2009; 30(4): 681 - 688. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. B. Caseiras, J.S. Thornton, T. Yousry, C. Benton, J. Rees, A.D. Waldman, and H.R. Jager Inclusion or Exclusion of Intratumoral Vessels in Relative Cerebral Blood Volume Characterization in Low-Grade Gliomas: Does It Make a Difference? AJNR Am. J. Neuroradiol., June 1, 2008; 29(6): 1140 - 1141. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Chawla, S. Wang, R.L. Wolf, J.H. Woo, J. Wang, D.M. O'Rourke, K.D. Judy, M.S. Grady, E.R. Melhem, and H. Poptani Arterial Spin-Labeling and MR Spectroscopy in the Differentiation of Gliomas AJNR Am. J. Neuroradiol., October 1, 2007; 28(9): 1683 - 1689. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
