| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the Nuclear Magnetic Resonance Center (P.L.L., R.G.G.) and Departments of Radiology (J.D.R., R.G.G.), Pathology (D.N.L.), Neurosurgery (F.G.B., G.R.C., E.A.C.), Radiation Therapy (A.F.T., J.S.L.), and Neurology (J.W.H.), Massachusetts General Hospital and Harvard Medical School, 55 Fruit St, Gray 2, Boston, MA 02114; and Department of Neurosurgery, Stanford Medical Center, Calif (G.R.H.). Received June 4, 2001; revision requested July 12; final revision received May 7, 2002; accepted May 29. Supported in part by National Institutes of Health grants 1R21CA80113 and 1R01CA83159. Address correspondence to J.D.R. (e-mail: jrabinov@partners.org).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: This blinded prospective study included 14 patients who underwent in vivo 3-T MR spectroscopy prior to stereotactic biopsy. All patients received a previous diagnosis of glioma (grade II or higher) and high-dose radiation therapy (>54 Gy). Prior to MR spectroscopy, conventional MR imaging was performed at 1.5 T to identify a gadolinium-enhanced region within the irradiated volume. Diagnosis was assigned by means of histopathologic analysis of the biopsy samples.
RESULTS: Sixteen of 17 biopsy locations could be classified as predominantly tumor or predominantly radiation effect on the basis of the ratio of choline at the biopsy site to normal creatine level by using a value greater than 1.3 as the criterion for tumor. The remaining case, classified as recurrent tumor on the basis of MR spectroscopy results, was diagnosed as predominantly radiation effect on the basis of histopathologic findings. Disease in this patient progressed to biopsy-proven recurrence within 3 months. Overall, the ratio of choline at the biopsy site to normal creatine level was significantly elevated (unpaired two-tailed Student t test, P < .002) in those biopsy samples composed predominantly of tumor (n = 9) compared with those containing predominantly radiation effects (n = 8). The ratio was not significantly different between the two histopathologic groups.
CONCLUSION: In vivo 3-T MR spectroscopy has sufficient spatial resolution and chemical specificity to allow distinction of recurrent tumor from radiation effects in patients with treated gliomas.
© RSNA, 2002
Index terms: Brain, biopsy, 13.1261 • Brain, effects of irradiation on, 13.47 • Brain neoplasms, diagnosis, 13.363 • Brain neoplasms, MR, 13.12145 • Brain neoplasms, MR spectroscopy, 13.12145 • Radiations, injurious effects, 13.47
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Of these methods, MR spectroscopy has advantages over PET and SPECT in that high-energy radiation is not used, and radio-labeled tracers are not required. More important, MR spectroscopy is intrinsically multiparameter, yielding simultaneous information about a variety of metabolites. Results of in vitro experiments have demonstrated unequivocally that the MR spectroscopic chemical profile is different for different tissues and even for different cell types (5–8). With sufficient spatial and spectral resolution, all tissue types can be localized spatially and identified unambiguously.
Although in vivo MR spectroscopy provides diverse metabolic information about brain tumors (9–12), it has yet to achieve widespread clinical application. This is partially due to the limited spatial and spectral resolution obtained at the standard magnetic field of 1.5 T. Because both spectral resolution and signal-to-noise ratio (SNR) depend linearly on the magnetic field, these problems are overcome at a higher field strength. An increase in spectral resolution provides better separation of metabolite signals, thereby improving the ability to identify and quantify chemical species. Additionally, the increased SNR allows for better spatial resolution.
Until recently, MR imagers with higher field strengths have been limited to research studies. In 1999, however, the U.S. Food and Drug Administration approved the use of magnetic field strengths of up to 4 T for clinical brain studies. Although 1.5 T remains the industry standard, as MR imagers with higher field strengths become available for patient care, it will be important to determine which applications can provide additional diagnostic information when performed at these stronger field strengths. Thus, the purpose of our study was to determine if 3-T MR spectroscopy allows accurate distinction of recurrent tumor from radiation effects in patients with grade II or higher gliomas.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
In addition to padding, chin and forehead straps were used to reduce patient motion. Conventional MR images, consisting of sagittal T1-weighted (16/650 [repetition time msec/echo time msec], 24-cm field of view, 256 x 196 matrix, 5-mm sections) and transverse T2-weighted (72/5,000, 24-cm field of view, 256 x 256 matrix, 3-mm sections) images, were obtained for anatomic registration of the spectroscopic data. Image acquisition time was approximately 10 minutes.
The biopsy target was determined by a neurosurgeon (F.G.B., G.R.H., G.R.C., E.A.C.) on the basis of assessment of 1.5-T gadolinium-enhanced MR images obtained no more than 3 weeks prior to 3-T MR spectroscopy. The rectangular spectroscopic region of interest was localized by using the transverse T2-weighted 3-T MR images. The location was chosen by the neuroradiologist (J.D.R.) to include the previously identified biopsy target and as much of the lesion as possible, while excluding the scalp and including an area of uninvolved tissue. The size ranged from 50–100 cm3, with a mean value of 74 cm3.
Optimization of field homogeneity (shimming) was performed over the region of interest to minimize the spatial variations in the magnetic field. These variations, which degrade spectral quality, increase with magnetic field strength and are especially problematic in regions lower in the brain, such as the thalamus and temporal lobe. Although only one patient had a temporal lobe lesion, all 3-T MR examinations required a combination of automated shimming, by using an algorithm developed at our institution (14), and manual field adjustments. Total shimming time was approximately 10 minutes.
MR spectroscopic data were acquired from a 10-mm section by using a point-resolved spatially localized spectroscopy, or PRESS, pulse sequence, preceded by three chemical shift–selective pulses for water suppression. In-plane spatial resolution was 1.25 cm (16 x 16 matrix over a 20 x 20-cm field of view) for the initial six examinations (patients 1, 2, 4 [first image], 7, 10, and 14). The in-plane spatial resolution was subsequently increased to 0.75 cm (32 x 32 matrix over a 24 x 24-cm field of view) for the remaining examinations. Examination time was kept constant by reducing the number of acquisitions for each phase-encoding step from four to one. An echo time of 144 msec was used to achieve full inversion of any lactate present. Published metabolite T1s at 3 and 4 T (15,16) differ from those at 1.5 T (17) by no more than 200 msec. Therefore, to avoid extending the examination time, the repetition time was not increased beyond the 1.5 seconds that is often used at 1.5 T. Acquisition time for the MR spectroscopic data was 25 minutes. Total examination time, including setup, was approximately 1 hour.
The in vivo MR spectroscopic data were analyzed on a Sun Ultra 1 workstation (Sun Microsystems, Mountain View, Calif) by using Sage IDL processing software (GE Medical Systems). Raw data were zero filled and apodized spectrally with a 3-Hz exponential filter prior to Fourier transform. The resulting spectra were phase corrected. The spectroscopic data were overlaid onto transverse MR images for anatomic registration of the metabolite distribution. Correlation of the MR spectroscopic grid and the biopsy site was performed qualitatively by the neuroradiologist (J.D.R.) and the neurosurgeon (F.G.B., G.R.H., G.R.C., E.A.C.) by using visual landmarks. Although this method is not adequately precise at high spatial resolutions, the smallest voxel cross-sectional area used in this study (56.25 mm2) was 18 times larger than that of the core biopsy sample (3.14 mm2). When necessary, the in-plane position of the MR spectroscopic grid was adjusted so that the biopsy site was contained within a single voxel. An additional single voxel containing normal tissue was also chosen for use as a reference.
Two criteria were used in selecting the reference area. First, it had to lie outside the areas of T2 abnormality and contrast material enhancement seen on the conventional 1.5-T and 3-T MR images. Second, because radiation therapy can alter biochemistry, even in the absence of anatomic changes (18,19), we also chose reference voxels that had the largest N-acetylaspartate (NAA) signal and least elevated choline (Cho) level relative to that of creatine (Cr). Whenever possible, a reference voxel contralateral to the lesion was used.
Spectral data from the voxels corresponding to biopsy locations and normal tissue were exported to the commercial software program Igor (Wavemetrics, Lake Oswego, Ore), where peak areas for Cho, Cr, and NAA were determined by using an iterative fit assuming Lorentzian line shapes. When necessary, a baseline correction was applied by using a polynomial fit. For each biopsy location, two peak area ratios for Cho were calculated: Cho from the biopsy site relative to Cr from the biopsy site (Chob/Crb, where "b" means "biopsy") and Cho from the biopsy site relative to Cr from the normal voxel (Chob/Crn, where "n" means "normal"). The Cho/Cr peak area ratio for the normal voxel (Chon/Crn) was also measured. When NAA was observed at the biopsy location (NAAb), NAA peak areas relative to Crb and Crn were determined. The MR spectroscopic data were analyzed by a single investigator (P.L.L.), who was blinded to the diagnosis assigned on the basis of histopathologic findings.
Stereotactic Biopsy
In all patients, stereotactic biopsy was performed within 3 weeks of in vivo 3-T MR spectroscopy. Core biopsy samples 10 mm in length and 2 mm in diameter were obtained. In some instances, biopsy was performed at more than one site. Prior to biopsy, a neuroradiologist reviewed the position of the MR spectroscopic region of interest with the neurosurgeon to ensure correct correlation of the biopsy site with results of the previous MR imaging examination.
Histopathologic Findings
After diagnosis was assigned by analyzing frozen sections, cores were fixed in 10% formalin and routinely processed for paraffin embedding. The tissue was cut into 6-µm sections and stained with hematoxylin-eosin. These sections were examined by means of light microscopy by the neuropathologist (D.N.L.), who was blinded to the MR imaging and spectroscopy findings.
Histopathologic characteristics of radiation effects include necrosis with punctate mineralization, relatively low cellularity (see below), reactive glial cells, and vascular changes that include vessel wall necrosis, hyalinization, and endothelial activation. The glial cells within regions of radiation damage include both reactive astrocytes and residual neoplastic cells. The residual neoplastic cells are characterized by large, irregular, sometimes multiple nuclei, often with copious cytoplasm—microscopic features typical of irradiated human cells in many organs. These cells usually cluster loosely in perivascular islands of viable tissue that are found in necrotic zones. Importantly, mitotic activity is not present. Tumor recurrence is identified on the basis of hypercellularity, often with cells closely apposed to one another, with mitotically active anaplastic tumor cells that have hyperchromatic, irregular, "naked" nuclei. Some biopsy samples from recurrent tumors also had regions of radiation damage. Specimens were categorized by the neuropathologist as predominantly radiation effect (<50% tumor) or predominantly tumor (>50% tumor).
Statistical Analysis
Statistical analysis was performed with four metabolite ratios determined by means of in vivo MR spectroscopy: Chob/Crb, Chob/Crn, NAAb/Crb, and NAAb/Crn. For each of the four ratios, values from biopsy samples found to be predominantly tumor were compared with those found to be predominantly radiation effect to determine if significant spectral differences existed between the two histopathologic groups. An unpaired two-tailed Student t test was used to calculate P values. In addition, for both histopathologic groups, a paired two-tailed Student t test was performed to determine if Chob was significantly elevated in relation to Chon.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
In contrast, for biopsy samples determined to be predominantly radiation effect, Chob/Crn was not significantly elevated in relation to Chon/Crn by using the same statistical test (P < .5). A representative case is shown in Figure 2 (patient 13). The Chob/Crb peak area ratio is 2.0, while that of Chob/Crn is 1.1. Histopathologic findings showed sparse cellularity with reactive astrocytes and vascular fibrosis, which is consistent with radiation effects. A small area of necrosis was also present at the margin. No residual tumor was observed.
|
|
|
|
For one case deemed to be only 25% residual tumor on the basis of histopathologic findings, Chob/Crn was somewhat higher at 1.5 T. This exception corresponds to data from the first MR spectroscopic examination of patient 4, who underwent a second 3-T MR spectroscopic examination 2 months later. Chob/Crn at that time was found to be 2.5, while Chob/Crb decreased from 5.1 to 2.9. For the second biopsy, the diagnosis assigned on the basis of histopathologic findings was 75% tumor.
NAA levels at the biopsy site were decreased in relation to those of normal tissue for all patients, although NAA/Cr ratios were not sufficient to distinguish tumors from radiation effects (P < .2 for NAAb/Crb; P < .1 for NAAb/Crn). For eight of 17 biopsy sites, no NAA was detected in vivo. The diagnoses assigned on the basis of histopathologic findings for these cases were divided equally between recurrent tumor (n = 4) and predominantly radiation effect (n = 4). Measurable levels of NAA were seen more often when the larger voxel size was 1.6 cm3 (six of eight cases) rather than 0.6 cm3 (three of nine cases).
In the nine cases in which NAA was detected, five cases were assessed as tumor and four as predominantly radiation effect on the basis of histopathologic analysis. The amount of lipid observed in the area of the lesion varied from patient to patient. In some cases, it had the largest signal in the spectrum, whereas in others the signal was almost undetectable. No correlation was found between the intensity of the lipid signal and the final histopathologic diagnosis. The presence of lipids can also mask the lactate resonance at 1.3 ppm. A lactate doublet was observed clearly in only two patients. In both patients, these regions were associated with tumor recurrence. For all patients, no metabolite signals were identified other than Cho, Cr, NAA, lipid, and lactate.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The diagnostic utility of in vivo MR spectroscopy depends on how well the technique can match the standard of histopathologic findings. One question to be addressed is whether changes in Cho/Cr have adequate specificity for diagnosis. A large number of MR spectroscopic examinations have been performed in an effort to determine how spectral patterns are related to tumor grade and progression (9–12,20–24). In virtually all untreated tumors, an increase of Cho and a decrease of NAA are observed. There is less agreement in the literature as to what occurs after radiation therapy. Decrease, increase, and no change in Cho levels have all been reported after radiation treatment (10,18,19,25–31). Some groups have reported that NAA and lactate can allow reliable distinction between recurrent disease and radiation effect (26,29,30). We believe that the different conclusions drawn from these studies can be explained by taking into account two factors, which are addressed in this article: spatial resolution and choice of reference signal.
In many articles on in vivo MR spectroscopy, the voxel size is 8 cm3 or larger (11,12,21,22,24,27,29–32). The 3-T data obtained in the present study suggest that this degree of spatial resolution may not be adequate for characterizing heterogeneous irradiated tumor beds. This is because after treatment, lesions most often contain a mixture of both tumor and radiation effects. This is clearly demonstrated in Figure 3. The presence of a small focus of tumor in a 1.6-cm3 voxel containing mostly radiation effects is not sufficient to allow an unequivocal identification of the tumor on the basis of the in vivo MR spectroscopy spectrum.
A further problem with inadequate spatial resolution is the inclusion of normal tissue. Normal levels of NAA, Cho, and Cr tend to increase NAA/Cr levels and decrease Cho/Cr levels observed at the biopsy site. As voxel size and heterogeneity increase, the spectral pattern arising from the mixture of the three tissue types becomes difficult to interpret. This issue has been raised before (33), and a four-compartment model has been proposed by Ott et al (34) to try to address the problem of heterogeneity in larger voxels.
In the data presented here, the reduction or absence of NAA signal was not diagnostic of tumor recurrence. This is because disease and irradiation can both cause neuronal death and damage, resulting in a nonspecific decrease in NAA. Metabolites giving rise to the Cho signal are associated with the cell membrane and involved in both its synthesis and degradation (35). Thus, an increase in Cho may indicate either rapid cell division in a growing tumor or cell destruction from radiation and an influx of inflammatory cells. While both processes result in an increase in Cho, the mechanisms are different, and thus, the degree of Cho elevation is different. High-grade tumors can show as much as a fourfold increase in Cho signal (36).
Investigators in two studies of patients with nasopharyngeal carcinoma found that radiation treatment caused, at most, a 50% increase of Cho in the uninvolved temporal lobe in relation to Cho levels in nonirradiated control subjects (18,19). Nevertheless, investigators in one serial study of patients with glioma found Cho values as high as 2.0 to be associated with both recurrent tumors and stable disease (37). The two groups could, however, be distinguished by the interval change in Cho relative to the value in a baseline MR examination. Thus, there is some overlap between the smallest Cho signal that can be attributed to recurrent tumor and the largest Cho signal that arises from radiation effects. The latter may depend on both dose and time elapsed from treatment. Any inclusion of normal tissue with tumors will blur the distinction further, which again emphasizes the importance of adequate spatial resolution in examinations of this type.
Our motivation for using a 3-T MR imager to examine tumor recurrence in patients with brain tumors after radiation therapy was the gain in SNR and spectral resolution that is obtained at the higher field strength. Since no additional peaks were identified that are not detected routinely at 1.5 T, the primary benefit of using 3 T for this application was the ability to increase spatial resolution. The twofold improvement in SNR at 3 T compared with that at 1.5 T allowed the use of a 10-mm section thickness (rather than the usual 15–20 mm) and an in-plane resolution as small as 0.75 (32 x 32 phase-encoding matrix), while maintaining a SNR of 5–6 for Crn.
A further increase in the number of phase-encoding steps is limited as much by the increase in examination time beyond a 25-minute acquisition as it is by a decrease in per-voxel SNR. Furthermore, although the gain in spectral resolution was not sufficient to resolve additional peaks, the decrease in overlap between the Cho and Cr signals facilitates quantification.
In addition to concerns about spatial resolution, a second important consideration in interpreting metabolite levels is the reference value against which a given signal increase or decrease is measured. The difficulties of absolute quantification in in vivo MR spectroscopy are well documented (38,39). In most MR spectroscopy studies, the metabolite levels are reported relative to Cr signal either from the same voxel or from a voxel containing only normal tissue. On the basis of our results (Fig 3), Chob/Crn appears to be the best parameter for determining whether a lesion contains predominantly tumor (Chob/Crn > 1.3) or radiation effects (Chob/Crn < 1.3). While, in each case, Chob was elevated in relation to Crb within the voxel (Chob/Crb > 1.0), Cho can be decreased in relation to Crn when a substantial necrotic fraction is present. In serial studies, the reference signal is usually the pretreatment MR spectral pattern.
In the most comprehensive study of this type to date, Wald et al (40) followed up 12 patients with glioblastomas who had received brachytherapy. By using voxels as small as 0.34 cm3, they observed a decrease in Cho in relation to pretreatment levels in six of the 12 patients, which was associated with a positive clinical response. Even 25 weeks after implantation, however, the intravoxel Cho/Cr remained higher than that of normal tissue.
Assessment of Chob/Crb may indicate how much, if any, active tumor is present. The limitation of this parameter is clearly the SNR. There are great uncertainties associated with quantifying Chob/Crb for the cases of recurrent disease in which Chob is high and Crb is low. The same problem occurs in those instances in which all metabolite signals were low because of a substantial necrotic component. Thus, the detection of small residual foci of tumor in a lesion that is predominantly radiation effect remains a challenge.
When a patient with a treated glioma develops an enhancing lesion after treatment, the clinical question is often framed as "recurrent tumor versus treatment effects." The results of the present study underscore the fact that the answer is often both. In eight of the 14 patients studied, histopathologic findings confirmed the presence of both tumor and radiation effects within a single lesion. This intimate mixing of tissue types means that adequate spatial resolution is crucial to obtain useful in vivo MR spectroscopic data. The 1–2-cm3 resolution used in the present study is sufficient for a correlation with histopathologic findings. As we have demonstrated in at least one case, however, some partial volume effects will still be present, and the detection of small amounts of residual tumor in the presence of radiation effects may be missed at this spatial resolution.
An additional point is that complete necrosis (defined spectroscopically as an absence of all metabolites), although sometimes observed at the center of the lesion, never occurred at the biopsy site, which was always located at the rim of the enhancing region. Even in instances in which the histopathologic findings showed only radiation effects, metabolites were still detected with MR spectroscopy. We therefore believe that "recurrence versus necrosis" is a poor description of the clinical issue, and we prefer to use the term treatment effects, rather than radiation necrosis. Although this nomenclature is vague, our current understanding of the details of the effects of radiation on tissue is vague, as well. Radiation effects can manifest in a variety of ways, including vascular damage, cytologic changes, mineralization, and necrosis. The relationship between these effects and parameters such as total dosage, fractionation, mode of delivery, time elapsed from treatment, and adjuvant chemotherapy is not yet known in detail, nor is it known how these parameters are reflected in the spectral patterns of MR spectroscopy.
Despite our incomplete understanding of the underlying biochemistry that causes changes in metabolite levels, in the present study, we were able to distinguish lesions composed of predominantly tumor recurrence from those containing predominantly radiation effect on the basis of Chob/Crn levels. In each case, histopathologic samples were obtained within 3 weeks of MR imaging. The data strongly support the hypothesis that correlations exist between in vivo MR spectroscopic findings and histopathologic findings, provided there is adequate spatial resolution in the spectroscopic data. The ability to discriminate tumor recurrence from radiation effects on the basis of changes in Cho levels shows that the technique has excellent chemical specificity. Thus, the results of the present study demonstrate that 3-T MR spectroscopy allows characterization of the complex mixture of tissue types and metabolism present in irradiated tumor beds.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, R.G.G., P.L.L.; study concepts, R.G.G., J.D.R., G.R.H., A.F.T.; study design, R.G.G., P.L.L., J.D.R.; literature research, P.L.L.; clinical studies, J.D.R., J.W.H., D.N.L., A.F.T., J.S.L.; experimental studies, P.L.L.; data acquisition, P.L.L., J.D.R., F.G.B., G.R.H., G.R.C., E.A.C., D.N.L.; data analysis/interpretation, P.L.L., J.D.R., D.N.L.; statistical analysis, P.L.L.; manuscript preparation, P.L.L., J.D.R.; manuscript definition of intellectual content, R.G.G., P.L.L., J.D.R.; manuscript editing, P.L.L., J.D.R., F.G.B., D.N.L.; manuscript revision/review and final version approval, R.G.G.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. Hourani, L.J. Brant, T. Rizk, J.D. Weingart, P.B. Barker, and A. Horska Can Proton MR Spectroscopic and Perfusion Imaging Differentiate Between Neoplastic and Nonneoplastic Brain Lesions in Adults? AJNR Am. J. Neuroradiol., February 1, 2008; 29(2): 366 - 372. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Goebell, J. Fiehler, X.-Q. Ding, S. Paustenbach, S. Nietz, O. Heese, T. Kucinski, C. Hagel, M. Westphal, and H. Zeumer Disarrangement of Fiber Tracts and Decline of Neuronal Density Correlate in Glioma Patients--A Combined Diffusion Tensor Imaging and 1H-MR Spectroscopy Study AJNR Am. J. Neuroradiol., August 1, 2006; 27(7): 1426 - 1431. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. G. Sorensen Magnetic Resonance As a Cancer Imaging Biomarker J. Clin. Oncol., July 10, 2006; 24(20): 3274 - 3281. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. A. Butowski, P. K. Sneed, and S. M. Chang Diagnosis and Treatment of Recurrent High-Grade Astrocytoma J. Clin. Oncol., March 10, 2006; 24(8): 1273 - 1280. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Vazquez, A. Castellote, J. Piqueras, P. Ortuno, J. Sanchez-Toledo, P. Nogues, and J. Lucaya Second Malignancies in Pediatric Patients: Imaging Findings and Differential Diagnosis RadioGraphics, September 1, 2003; 23(5): 1155 - 1172. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
and 
= individual cases.