| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Thoracic Imaging |
1 From the Department of Diagnostic Radiology, University Hospital Kiel, Arnold-Heller-Strasse 9, 24105 Kiel 1, Germany (J.B., M.B., C.L., M.R., M.H.); Department of Physics, University of Würzburg, Germany (P.J.); and Siemens Aktiengesellschaft, Hamburg, Germany (J.G.). Received December 3, 2001; revision requested January 16, 2002; revision received March 28; accepted May 13. Address correspondence to J.B. (e-mail: juergen.biederer@rad.uni-kiel.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty-five patients with different lung abnormalities were examined with 3D GRE MR imaging. The small pulmonary nodules in seven, TNM stage of large intrapulmonary tumors in eight, and benign bronchial disease in five patients were evaluated. MR imaging–based diagnoses were compared with diagnoses made at CT and at discharge from the hospital. Contingency tables and the McNemar test were used to evaluate the significance of differences between MR imaging– and CT-based diagnoses.
RESULTS: The MR imaging– and CT-based diagnoses were identical in 24 of 25 patients. In the remaining patient, clinical findings confirmed the accuracy of the MR imaging finding of pleural empyema. Ten of 15 solid pulmonary nodules smaller than 10 mm in diameter were detected at MR imaging (P > .1). Tumor stages at MR imaging and CT were identical, but lymph node stages at the two examinations differed in two of eight patients owing to overestimation of lymph node size at MR imaging (P > .2). In the five patients with bronchiectasis, MR imaging depicted 26 of 33 affected lung segments; differences between MR imaging and CT findings of bronchial dilatation (P > .05) and bronchial wall thickening (P > .2) were not significant. Peribronchial fibrosis was overestimated at MR imaging owing to image artifacts (P < .05).
CONCLUSION: Study results confirmed the feasibility of fast breath-hold 3D GRE MR imaging of the lung.
© RSNA, 2002
Index terms: Computed tomography (CT), comparative studies • Lung, abnormalities, 60.26, 60.28, 60.76 • Lung neoplasms, CT, 60.31, 60.32 • Lung neoplasms, MR, 60.121412, 60.121415, 60.121416, 60.12143 • Magnetic resonance (MR), comparative studies, 60.121412, 60.121415, 60.121416, 60.12143
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Suitable techniques for the detection of small lung lesions with high spatial resolution—for example, for early lung cancer screening—are still not available. Recently developed three-dimensional (3D) gradient-recalled-echo (GRE) sequences for volumetric interpolated breath-hold (VIB) imaging of the lung might introduce new capabilities for MR imaging of lung morphology with high spatial resolution, but these sequences have not yet been applied to clinical imaging (11,12). Our aim in this study was to prospectively evaluate the clinical feasibility of fast 3D GRE VIB MR imaging and to compare this examination with standard CT in patients with different lung abnormalities.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MR and CT Imaging
All MR imaging examinations were carried out by using a commercially available 1.5-T MR imaging unit (Magnetom Vision; Siemens, Erlangen, Germany) with a maximum gradient strength of 25 mT/m and a rise time of 600-µsec and by using a dedicated phased-array body coil. The T1-weighted 3D GRE VIB MR imaging sequence involved a repetition time of 4.5 msec, an echo time of 1.9 msec, and a flip angle of 12° (11). The technical features of this sequence have been described in detail by Rofsky et al (13). Briefly, 3D GRE VIB sequences are similar to the 3D radio-frequency–spoiled GRE sequences used to perform 3D MR angiography, but a lower flip angle is used to reduce the signal intensity of flowing blood and enhance the signal intensity of stationary tissue. The matrix is interpolated in both directions to acquire MR images with a maximum in-plane spatial resolution of 512 x 512 pixels. The acquisition time is shortened by reducing the number of phase-encoding steps along the partition direction with asymmetric echo sampling and sinc interpolation (ie, zero filling) (13).
The field of view ranged from 350 to 450 mm for the acquisition of coronal MR images. Slabs of 80-mm thickness with 16 partitions and an initial matrix of 251 x 256 were obtained within 20 seconds or less. If a patient could not hold his or her breath for 20 seconds, the slab thickness was reduced to 40 mm in eight partitions, which allowed acquisition times of 10–11 seconds. Each acquisition included two additional partitions for 14% oversampling to eliminate mirroring artifacts. The effective section thickness of 5.0 mm was reduced to 2.5 mm by means of volume interpolation. All images were displayed with an interpolated in-plane matrix of 502 x 512 pixels. These parameters provided interpolated voxel sizes of 2.50 x 0.70 x 0.65 mm3 (350-mm field of view) to 2.50 x 0.90 x 0.88 mm3 (450-mm field of view).
The 3D GRE VIB MR imaging sequence includes an optional quick fat saturation scheme with chemically selective fat saturation pulses. For the acquisition of MR images with the fat signal suppressed, initial and interpolated matrices were reduced to 90% to stay within the time frame of a breath hold. This modification was used in patients with a prevalence of substantial heart pulsation artifacts that resulted from a thick epicardial fat layer.
The overall imaging time for the acquisition of three to five slabs was about 10 minutes. All MR images were obtained during a deep breath hold. Three coronal slabs were acquired to cover the lung completely. Five to six acquisitions were needed if 40-mm slabs were applied to reduce the breath-hold time frame. The first slab was centered at the middle of the mediastinum and the tracheal bifurcation. Additional slabs were positioned anteriorly or posteriorly to this slab. A 10-mm overlap was applied to avoid sampling errors at the slab limits. Numeric slab positioning in 2.5-mm steps permitted secondary reformations of the entire volume.
The described 3D GRE VIB sequence was added to or integrated into the MR imaging examinations, which were performed for accepted clinical indications, and the MR imaging reports were part of the clinical work-up. Contrast material–enhanced MR imaging examinations were performed only when the administration of contrast medium was needed to complete the examination for which the patient was originally referred to our department (in six of 25 examinations). In such cases, we used a power injector (Spectris; Medrad, Pittsburgh, Pa) to inject a 20-mL bolus of gadopentetate dimeglumine (Magnevist; Schering, Berlin, Germany), followed by a 20-mL saline flush (0.9% sodium chloride), at a rate of 2 mL/sec. Specific bolus timing, such as that used for MR angiography, was not applied. In the case of pleural empyema in one patient, delayed MR imaging was performed 5-10 minutes after contrast material administration.
CT scans were obtained by using a single-section spiral scanner (Somatom Plus-S; Siemens) with a beam collimation of 10 mm and a table speed of 15 mm/sec (rotation time, 1 second; pitch, 1.5; 210 mAs; 137 kV). CT scans with a section thickness of 8 mm and an increment of 7 mm (overlapping reconstruction) were reconstructed by using a slim reconstruction profile and an edge-enhancing algorithm with a matrix of 512 x 512 pixels. All CT scans were obtained after the injection of 60 mL of a contrast material containing 300 mg of iodine per milliliter (Ultravist 300; Schering) at a flow rate of 2.5 mL/sec and with a scanning delay of 25 seconds.
For MR image review, we used a commercially available workstation (MagicView; Siemens). The three coronal slabs were added to one 3D GRE VIB image series and reviewed together. To exclude sampling errors due to missed lesions at the slab limits after cutting off the overlap, we reviewed the complete middle slab separately. In addition to performing the obligatory review of the source 3D GRE VIB images in the coronal orientation, we reviewed secondary reformations of the coronal slabs in transverse and sagittal planes by using the multiplanar reformation mode at the workstation.
Image Evaluation
The 3D GRE VIB MR images were evaluated by two observers (M.R. and J.B.) who were blinded to the CT-based diagnoses. Both observers are experienced radiologists who are certified by the medical board of the state of Schleswig-Holstein and familiar with conventional thoracic imaging, CT, and MR imaging of the chest in particular. The diagnosis based on 3D GRE VIB MR imaging (including that with administration of contrast material, if applicable) findings was established at consensus and then compared with the diagnoses based on the helical CT findings at admission and on the clinical work-up at discharge from the hospital.
Before the evaluation of particular findings, the MR images were rated for the presence of artifacts and for the spatial resolution of vessel and airway structures; this evaluation approach was similar to that proposed by Biederer et al (11). Artifacts caused by respiratory motion and heart pulsation were rated as absent (grade 0), present but negligible (grade 2), or substantial with impaired diagnostic image quality (grade 3). If substantial breathing-related artifacts (grade 3) were identified during image acquisition, we used the reduced-breath-hold time frame that was described herein earlier. If fat suppression was used to reduce artifacts caused by cardiac motion, the image sets obtained with and without fat saturation were compared to determine whether artifacts were reduced and whether visualization of lung abnormalities could be improved.
In group 1, the seven patients with solid pulmonary nodules, we evaluated the number, size, and location of the nodules at MR imaging and CT separately to assess the sensitivity of 3D GRE VIB MR imaging for the detection of small nodules. Only those lesions smaller than 10 mm in diameter were included, and they were divided into three groups: lesions with diameters of 3 mm or less, lesions with diameters of 3–5 mm, and lesions larger than 5 mm in diameter (largest diameter at CT).
Group 2 comprised eight patients with large intrapulmonary tumors and a diagnosis at admission of suspected bronchial carcinoma. To judge the potential value of 3D GRE VIB for staging purposes in these patients, we evaluated the images by using the TNM classification system for staging of bronchial carcinoma (14). Tumor size and lymph node and metastasis stages were evaluated separately at MR imaging and CT and compared afterward. Mediastinal lymph node enlargement was diagnosed when the transverse diameter of the node was greater than 10 mm (15 mm for infracarinal lymph nodes).
Group 3 comprised five patients with benign bronchial disorders—namely, bronchiectases. We examined these patients further to estimate the diagnostic accuracy of MR imaging for depicting benign bronchial abnormalities. With both MR imaging and CT, each lung segment was evaluated for three criteria—bronchial dilatation, thickened bronchial walls, and peribronchial fibrosis.
The last group, group 4, comprised five patients with miscellaneous abnormalities—namely, a mediastinal mass or an abnormality of the pleura and chest wall. For these patients, 3D GRE VIB MR image evaluation was focused on the detection and diagnosis of mediastinal and pleural abnormalities. Contrast material was administered intravenously in all five of these patients. We compared pre- and postcontrast MR images to evaluate whether contrast material enhancement helped in establishing the final diagnosis.
Statistical Analysis
CT was considered the reference imaging modality for evaluation of the diagnostic qualities of MR imaging. Two-by-two contingency tables were used to identify the degree to which MR imaging was in accordance with CT in enabling the identification of pulmonary nodules in patient group 1, mediastinal lymph nodes (ie, N stage) in group 2, and bronchial dilatation, thickened bronchial walls, or peribronchial fibrosis in group 3. Standard definitions of sensitivity, specificity, and negative and positive predictive values were applied. The McNemar test (with two-sided P values) was used to evaluate the significance of discordance between 3D GRE VIB MR imaging and CT findings. A two-sided P value of less than or equal to .05 was considered the threshold for statistical significance.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
With phase encoding oriented in the right-to-left direction, pulsation artifacts were present next to the heart and the large vessels of the mediastinum. In 20 of 25 MR examinations, these artifacts were negligible in the diagnosis of the lung abnormality. In five examinations, the image quality was judged to be insufficient—for example, in the detection of small lesions next to the heart. In four patients with a thick epicardial fat layer, suppression of the fat signal visibly reduced the pulsation artifacts on the MR images obtained from the anterior slab. In one patient, this procedure facilitated the detection of a single pulmonary nodule. No further effects on the visibility of lung abnormalities in these four patients were observed.
The lung vessels were identified by the high signal intensity of blood flow, without contrast material administration. Fourth- to fifth-order subsegmental vessels were visible in all parts of the lung. The signal intensities of the vessels varied according to the flow velocity and the flow direction toward the slab orientation. Pulmonary arteries and veins could be discriminated by not only their different courses but also their different signal intensities on the MR images. Lobar and segmental bronchi were identified. The signal intensity and spatial resolution of the bronchial walls were sufficient to delineate the third-order segmental bronchi as tubular structures. Fourth-order subsegmental bronchi could not be discriminated.
Diagnoses Based on MR Imaging Findings
The results of separate MR image readings, as compared with the results of CT scan readings, are listed in the Table. In the majority of cases—in 24 of 25 patients—the 3D GRE VIB MR imaging–based main diagnosis did not differ from the CT-based diagnosis. In one case, contrast-enhanced MR imaging evaluation yielded important additional information: pleural empyema in an 86-year-old man with uncomplicated pleural effusion at CT. In another patient, 3D GRE VIB MR imaging with contrast material enhancement and multiplanar reformation confirmed a thrombosed posttraumatic aneurysm of 5.3 cm, which was suspected after the CT image reading. The diagnosis was confirmed by means of puncture of the empyema in the first patient and by means of angiography in the second patient. In both patients, 3D GRE VIB sequences yielded good-diagnostic-quality MR images.
|
|
|
|
|
|
|
Large Intrapulmonary Tumor Staging
In the eight patients with large intrapulmonary tumors in group 2, the stages of tumor size at MR imaging and CT did not differ. The invasive character of the three T4 tumors could be easily detected on the 3D GRE VIB MR images. The angiographic capacities of this sequence in particular enabled the acquisition of valuable information: tumor invasion into large vessels was demonstrated in two of the eight patients. The N stage of lymph nodes at MR imaging and CT differed in two cases. In both patients, the size of the lymph nodes was overestimated at MR imaging. This led to a false-positive classification of N2 stage (paratracheal lymph node estimated to be larger than 10 mm in diameter at MR imaging but 8 mm [thus, N0] at CT) in one patient and a false-positive classification of N1 stage (detection at MR imaging of a hilar lymph node that was not confirmed at CT) in another patient. In both cases, image quality was judged to be sufficient. Histopathologic analysis performed after surgery revealed no nodal involvement with malignancy. Due to the small number of patients, a discordance between the results of the two imaging modalities in the depiction of mediastinal lymph nodes was not confirmed (two-sided P > .2). In one patient, both imaging examinations revealed a mass in the right upper lobe that was interpreted as bronchial carcinoma before surgery, but histopathologic analysis revealed an inflammatory pseudotumor. In two other patients, biopsy revealed small cell bronchial carcinoma. Thus, retrospectively, TNM staging, per definition, was not applicable in three of the eight patients.
Bronchiectasis
The correct diagnosis in group 3, the five patients with bronchiectasis related to either cystic fibrosis (3) or postpneumonic changes (2), could be established with both modalities. CT of the 95 lung segments in these patients revealed 33 segments with ectatic bronchi. MR imaging depicted 26 of these 33 segments and had a sensitivity of 79% and a specificity of 98% (61 of 62 lung segments) on the basis of these criteria for bronchiectasis. The positive and negative predictive values were 96% (26 of 27 segments) and 90% (61 of 68 segments), respectively. The discordance between the two imaging modalities was not significant (McNemar two-sided P > .05).
MR imaging depicted 25 of the 28 lung segments that had thickened bronchial walls at CT, for a sensitivity of 89%, a specificity of 98% (66 of 67 segments), and positive and negative predictive values of 96% (25 of 26 segments) and 96% (66 of 69 segments), respectively. The discordance between CT and MR imaging according to this criterion was not significant (McNemar two-sided P > .2). Peribronchial fibrosis was detected in seven segments at CT. MR imaging depicted six of these seven segments, for a sensitivity of 86%, but it yielded false-positive results based on this criterion in 10 other segments. Thus, MR imaging reached a specificity of 89% (78 of 88 segments), a negative predictive value of 99% (78 of 79 segments), and a positive predictive value of only 38% (six of 16 segments). The discordance between the two modalities according to the criterion of peribronchial fibrosis was significant (McNemar two-sided P < .05). Retrospectively, the high prevalence of this false-positive finding could be related to motion artifacts and the blurred appearance of the thickened bronchial walls of the affected segments.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Three-dimensional GRE VIB MR imaging sequences combine angiographic capabilities and high spatial resolution with reasonable soft-tissue signal intensity. To our knowledge, this sequence was first applied to abdominal MR imaging, but studies of the MR imaging of lung abnormalities with this sequence in patients have not yet been published. Liver studies with transverse sections have involved the use of a 160 x 160 (interpolated to 256 x 256) matrix and included fat saturation (13). Semelka et al (12) observed that parts of the thorax also were well depicted on these transverse 3D GRE MR images. They suggested that high image quality and negligible image artifacts could be expected if the technique was applied to chest imaging.
To our knowledge, Biederer et al (11) proposed the first 3D GRE VIB protocol for MR imaging of the lungs with the acquisition of coronal images and confirmed high image quality in a study with healthy volunteers. The protocol described herein involves the use of a similar approach with the highest reasonable in-plane resolution of the sequence and thin coronal sections to optimize the spatial resolution. Compared with the fast two-dimensional GRE sequence, 3D GRE VIB sequences are better compensated for the reduction of signal-to-noise ratio with small voxel sizes (3,11,13,16,18). Rofsky et al (13) attributed this better compensation to an inherent improvement in signal-to-noise ratio with volumetric interpolated techniques due to filtering procedures.
Results of the present study involving 25 patients demonstrate the feasibility of the described 3D GRE VIB MR imaging protocol. In this study, the 3D GRE VIB MR imaging sequence yielded good images of the lung vasculature and segmentation without administration of contrast material. Differences in the signal intensity of blood flow between pulmonary arteries and veins may be diagnostically useful, because the overlapping of pulmonary arteries and veins can impair the discrimination between the two vessel systems at postcontrast MR imaging examinations (4,8,19,20). The tracheobronchial system also was well depicted at 3D GRE VIB MR imaging. The clear delineation of the details of small lung structures with the 3D GRE VIB sequence reflects good qualities of image data processing. The reduced partial volume effects in particular may be attributed to zero filling interpolation, as described with contrast-enhanced 3D GRE MR angiography (13,21).
The overall image quality achieved with the 3D GRE VIB MR imaging sequence was at least comparable to that achieved in previously published two-dimensional GRE MR imaging studies of the lung (3,16,18). Ghosting from cardiac and vessel pulsations, which is generally present on coronal two-dimensional GRE MR images, was absent or moderate on most of the 3D GRE VIB MR images in this series (12,17). Presaturation pulses, which are sometimes used in two-dimensional GRE MR imaging protocols to suppress the ghosting of mediastinal structures, were not necessary in this study (17). Breath-hold acquisition minimized the artifacts from respiratory motion and yielded images of completely inflated lungs. The similar inflation, with a comparable arrangement of anatomic structures, depicted at MR imaging and CT was useful for comparing findings between the two modalities (15,17).
Three-dimensional GRE VIB MR imaging has high potential—for example, for screening purposes—because it can depict lesions smaller than 3 mm in diameter. The high contrast between solid tissue and the black background of inflated and intact lung parenchyma facilitates the detection of small nodules. The 12° flip angle was the same as that recommended for liver studies and appears to be applicable to imaging of lung abnormalities (13). However, our study population included only seven subjects with small pulmonary nodules, and the database was not sufficient to calculate thresholds for lesion sizes and detection rates. The potential of the described MR imaging method for tumor staging was emphasized by the agreement between MR imaging and CT findings in the patients with tumors. The T stages determined with both modalities were identical, whereas there was a tendency to overestimate the N stage at MR imaging.
The findings in the small group of patients with bronchiectases (n = 5) were encouraging. Although MR imaging of the lung with a 3D GRE VIB sequence was not considered the ideal tool to image peripheral bronchial disorders, it facilitated a correct diagnosis in all patients. The sensitivity, specificity, and positive and negative predictive values of 3D GRE VIB MR imaging for depiction of bronchiectasis and the associated diagnostic criteria ranged from 79% to 99%, with the exception of a low positive predictive value for the detection of peribronchial fibrosis. This low value resulted from an overestimation of the prevalence of this finding due to motion artifacts. The diagnostic value of VIB MR images in the detection of peripheral bronchial disorders may be considered to be inferior to that of CT and thin-section CT but superior to that of conventional radiography, because chest radiographic studies in patients with bronchiectasis depict only indirect signs (22).
Three-dimensional GRE VIB sequences have high potential for facilitating fast MR imaging of the lungs. At present, the hardware of the system limits the slab thickness to 80 mm. Thicker slabs can be obtained only with lower spatial resolution. This is a minor drawback of the current protocol. A smaller number of slabs and a sufficient overlap are desirable to exclude the possibility of sampling errors at the slab limits. The coronal section orientation contributes to the reduction of sampling errors, because the motion pattern of the lung during inspiration and expiration predominantly implies expansion along the body axis (23,24).
Future hardware upgrades and partial parallel imaging techniques in particular are expected to enable high-spatial-resolution, single-breath-hold MR imaging of the entire lung. At present, multiplanar reformations are difficult to perform when the deepness of inspiration varies between acquisitions.
For the time being, a few recommendations to improve the image quality of 3D GRE VIB MR imaging can be made. Generally, we do not recommend fat saturation for thoracic studies. We make exceptions to this recommendation in cases of substantial cardiac ghosting due to a thick epicardial fat layer. In these cases, suppression of the fat signal can visibly reduce the signal intensity of artifacts. As a compromise, we suggest acquiring anterior images with fat signal suppression but with a slightly lower spatial resolution (ie, matrix of 90%) to keep within the time frame of a breath hold.
Cardiac triggering might be useful for reducing the ghosting of mediastinal structures, as described with 3D GRE MR angiography of the mediastinum (25). The results of our studies of 3D GRE VIB MR imaging sequences are promising, but as is typical with triggering methods, the imaging times were longer and the slab thickness had to be reduced to less than 40 mm per eight partitions (Biederer J, unpublished data, 2000). As a compromise, we suggest using a selected additional triggered series to image the hilar vasculature, if necessary.
Despite having limitations, the proposed 3D GRE protocol for MR imaging can be considered a prototype for one-shot VIB MR imaging of the lung. The possible applications for such a technique are widespread. Fast 3D data acquisition with a spatial resolution close to that achieved with standard helical CT and additional angiographic capabilities, but without the need for contrast material or radiation exposure, offers new applications for MR imaging of the lungs. Screening studies—for example, for bronchial carcinoma in patients at high risk for pulmonary malignancy, staging examinations, and pediatric examinations are feasible. The costs of such an examination would be acceptable if the time frames for the entire procedure barely exceeded those of performing chest radiography in two imaging planes. MR imaging of the lung remains time consuming, however, because a standard closed MR imaging unit is still used and patients enter the magnet while lying on a table. At this stage we may be able to speculate about the use of dedicated open MR imaging units for lung examinations. Such open units would be used to examine patients while they were standing upright and thus dramatically expedite the imaging procedure.
At the present stage of imaging technology development, short acquisition times, high spatial resolution, good visualization of the lung anatomy, and a low prevalence of artifacts are apparent advantages of 3D GRE VIB MR imaging. The 3D GRE VIB sequence has already been introduced into clinical chest MR imaging protocols at our institution and is used as part of combined sequence protocols. We recommend using the described 3D GRE VIB sequence particularly in combination with fast T2-weighted MR imaging sequences—for example, T2-weighted single-shot spin-echo train sequences—because the signal intensity of fresh pneumonic infiltrates on T1-weighted GRE images is very low. T2-weighted single-shot spin-echo train MR imaging is very fast and can depict pulmonary fluid content with very high signal intensity (26,27). Future efforts will need to be focused on the further development of such fast combined sequences and their clinical applications. Three-dimensional GRE VIB sequences will contribute greatly to future MR imaging studies of lung morphology.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, J.B., M.R.; study concepts, J.B., M.B., J.G., C.L., M.R., M.H.; study design, J.B., M.B., J.G., M.R., M.H.; literature research, J.B., M.B., M.R.; clinical studies, J.B., M.B., M.R.; experimental studies, J.B., M.B.; data acquisition, J.B., M.B., M.R.; data analysis/interpretation, J.B., C.L., P.J.; statistical analysis, J.B., C.L.; manuscript preparation, J.B., M.B., C.L.; manuscript definition of intellectual content, J.B., M.B.; manuscript editing, J.B., C.L.; manuscript revision/review and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. D. Brennan, T. Gleeson, L. E. Coate, C. Cronin, D. Carney, and S. J. Eustace A Comparison of Whole-Body MRI and CT for the Staging of Lymphoma Am. J. Roentgenol., September 1, 2005; 185(3): 711 - 716. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
