| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Medical Physics |
1 From the Department of Radiology, E2-A, Mayo Clinic and Foundation, 200 First St SW, Rochester, MN 55905. Received April 7, 2000; revision requested June 1; revision received May 2, 2001; accepted May 15. Address correspondence to C.H.M. (e-mail: mccollough.cynthia@mayo.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: An image quality phantom underwent imaging with moving and stationary grids with both a clinical S-F combination and CR plate. CT SPR was performed with six CT scanners at various milliampere second and kilovolt peak settings. Entrance skin exposure (ESE); spatial, contrast, and temporal resolutions; geometric accuracy; and artifacts were assessed.
RESULTS: S-F or CR images, with either grid, provided image quality equivalent to that with the clinical standard, S-F with a moving grid. ESE values for both S-F and CR were 435 mR (112.2 µC/kg [1 mR = 0.258 µC/kg]) with a moving grid and 226 mR (58.3 µC/kg) with a stationary grid. All CT SPR images provided inferior spatial resolution compared with S-F or CR images. High-contrast objects generated substantial artifacts on CT SPR images. Compared with S-F, CR and CT SPR provided improved resolution of small low-contrast objects. The contrast between iodine and soft-tissue–mimicking structures on CT SPR images acquired at 80 kVp was twice that at 120 kVp. CT SPR images with acceptable noise levels required a midline ESE value of approximately 300 mR (77.4 µC/kg) at 80 kVp.
CONCLUSION: S-F and CR provided better spatial resolution than did CT SPR. However, CT SPR provided improved low-contrast resolution compared with S-F, at exposures comparable to those used for S-F or CR.
Index terms: Computed tomography (CT), technology, 80.12115 • Radiography, technology, 80.1215 • Screens and films, 80.1215 • Urography, technology, 80.12115, 80.1215
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
4 line pairs per millimeter for screen-film radiography [S-F]) compared with that of CT (0.7 line pairs per millimeter for abdominal CT). Specifically, intravenous urography is considered superior for the delineation of calyceal and papillary anatomy, collecting ducts, mucosal detail, and small filling defects. In our experience, we have found that 10% of urinary tract abnormalities were depicted more clearly or appreciated only on S-F urographic images when compared with CT cross-sectional images. Hence, by combining the strengths of each study into one CT urographic examination, the urinary tract can be evaluated more thoroughly in a single session. Many different approaches are used in CT urography (6). Perlman et al (7) advocate the acquisition of intravenous urographic images in a urography suite, followed by patient transfer to a CT suite for the CT portion of the examination. In other centers, the CT cross-sectional images are acquired first and then the patient walks to the urography suite to complete the radiographic portion of the study. Movement of the patient between procedure rooms requires additional time, can cause scheduling complications, and may affect the level of pyelocalyceal distention in the second portion of the examination.
An alternative to this approach, which was recently implemented at our center, is the acquisition of abdominal radiographs with an overhead x-ray tube while the patient is lying on the CT table (8). This method allows high-spatial-resolution intravenous urography to be performed at various times during the CT examination without the need for the patient to move. The technique requires the use of an auxiliary CT tabletop that can accommodate an S-F or computed radiography (CR) cassette under the patient without introducing artifacts on the CT image (9). Because such a tabletop cannot include the metal components needed to operate a moving grid (ie, a table bucky), a stationary grid with a high line density must be slipped over the cassette.
As another alternative, other authors (10) performed the entire CT urographic study with the patient lying on the CT table, with use of the CT scanned projection radiographic (CT SPR) image (localization image) for the abdominal radiographs. This method also optimizes examination timing and pyelocalyceal distention, but it may be limited by the relatively poor limiting spatial resolution of CT SPR images (<1 line pair per millimeter). Another alternative in the literature, which was not addressed in this study, is reconstruction of coronal maximum or minimum intensity projection images or curved planar reformation images of the collecting system from the CT data.
Despite the interest in these various approaches to CT urography, to our knowledge, few studies have been published that compare the image quality, radiation dose, or ability to depict subtle urothelial abnormalities of these various alternatives. The purpose of this study was to compare image quality and dose among three techniques: (a) S-F or CR performed with the patient in a separate radiography room with use of a moving grid (table bucky), (b) S-F or CR performed with the patient lying on the CT table with use of a stationary (slip-on) grid, and (c) CT SPR performed with the patient lying on the CT table. Future work, which is beyond the scope of this article, will address the clinical performance of these methods.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
|
|
|
|
CT Systems Studied
Six CT systems (A to F) were used to perform CT SPR (localization or scout imaging) (M.R.B., C.H.M.). The following CT scanners were used in this study: system A, multi–detector row spiral CT, model QX/i, GE Medical Systems, Milwaukee, Wis; system B, single–detector row spiral CT, model CT/i, GE Medical Systems; system C, single–detector row spiral CT, model HiSpeed/RP, GE Medical Systems; system D, nonspiral CT, model 9800, GE Medical Systems; system E, electron-beam CT, model C150-XLP-HRD, Imatron, South San Francisco, Calif; system F, single–detector row spiral CT, model PQ6000, Picker International (now Marconi Medical Systems), Highland Heights, Ohio.
CT Acquisition Techniques
To optimize iodine contrast, or visualization of calcified ureteral stones (10), x-ray tube potential levels should range from 60 to 90 kVp. Thus, CT SPR was performed at both the routine and the lowest kilovolt peak settings available on each system (systems A to D, 120 and 80 kVp, respectively; systems E and F, only 130 kVp available). For all the CT systems, exposure times in the CT SPR mode cannot be selected directly; rather, they are a function of table speed. Only system D provided a choice of table speed (slow or fast), and the fast option was used to allow scanning of the full z-axis length of the phantom (430 mm). Except for system E, in which the tube current was fixed at 620 mA, tube currents between 10 and 400 mA were used to obtain a distribution of radiation dose and image noise levels. None of the CT systems allowed the user to choose postprocessing image algorithms (eg, sharp or smooth). All CT SPR images were obtained with the phantom height centered at scanner isocenter. The resultant digital CT SPR images were displayed and hard-copy images printed with the appropriate window and level settings.
Exposure Measurements
Entrance skin exposure (ESE) was measured (M.R.B., C.H.M.) for all images with use of an ionization chamber and electrometer (model 10 x 5-6 chamber and MDH model 1015 electrometer; Radcal, Monrovia, Calif). The chamber was positioned at the upper left corner of the image quality phantom.
Iodine-Contrast Measurements
Image brightness of the iodine-mimicking tin disks and the uniform background material, which was acrylic, were measured (M.R.B., C.H.M.) with use of an optical densitometer on the S-F and CR images and by using digital region-of-interest measurements on the CT SPR images. On S-F and CR images, multiple measurements were made over the images of the tin disks to estimate an average optical density value. On CT SPR images, the diameter of the circular region of interest was about 0.6 of the total disk diameter to avoid measuring the edge-enhanced region of the disks. The area within the region of interest was 60 mm2.
Artifacts and Spatial, or High-Contrast, Resolution
Spatial, or high-contrast, resolution was determined by means of visual inspection (C.H.M., M.R.B., T.J.V., B.F.K., A.J.L.) of the lead resolution targets and bone and steel wool test objects in the phantom. The limiting spatial resolution, or the spatial frequency of the bar pattern at which the bars and space become blurred, was determined by consensus. A consensus forced-choice rating—better than, same as, or worse than the S-F image acquired with the moving grid—of the clarity of the fine structures of the steel wool test object was also performed (C.H.M., M.R.B., T.J.V., A.J.L.). The presence or absence of artifacts (streaks, dark or white banding of objects, signal intensity of any type not present on the S-F image acquired with the moving grid) was noted for all images (C.H.M., M.R.B., T.J.V., B.F.K., A.J.L.).
Low-Contrast Resolution
Visual assessment of low-contrast resolution was performed by five experienced radiologists (T.J.V., B.F.K., A.J.L., J.P.Q., R.R.H.), who were blinded to the data acquisition system. Because the edge-enhanced appearance of high-contrast structures immediately identified which images were CT SPR images, all portions of the images were covered except for a triangular section that contained the small low-contrast plastic beads. A consensus forced-choice rating of the appearance of the beads and double-looped object was selected from the following options: better than, same as, or worse than the S-F image acquired with the moving grid.
Temporal Resolution
Temporal resolution for S-F and CR was determined on the basis of the exposure time used. For CT SPR, the total exposure time was measured with a stopwatch (M.R.B.). The total time was the time required to sequentially scan a length of 430 mm. The actual time required to image one region of the phantom object (eg, a 10-mm-long object along the z axis) was the total scanning time multiplied by the region length (in millimeters) divided by 430 mm. This time to scan a small section of the phantom, which was arbitrarily chosen to be 10 mm, was computed as an indicator of the amount of motion blur that might be expected over a small region of the CT SPR image.
Geometric Accuracy
Geometric accuracy was determined by measuring (M.R.B.) the vertical (z axis, craniocaudal), horizontal (left to right), and diagonal dimensions of the three targets. Measurements were made with rulers on S-F and CR images, and the values were corrected for image magnification. Digital calipers were used to measure these same distances on CT SPR images.
Clinical CT SPR Abdominal Imaging
Institutional review board approval was obtained for the comparison between CT SPR and S-F images acquired in patients who were lying on the CT table as part of the clinical CT urographic examination at our center. The study was deemed minimal risk, as the lower-dose CT SPR images were substituted for the higher dose tomographic images from the preempted intravenous urographic examination; hence, informed consent was not required. CT SPR images were acquired at various times after intravenous injection of contrast media (Isovue-300; Bracco Diagnostics, Princeton, NJ). CT SPR acquisitions were intermixed with either CT cross-sectional imaging or S-F radiography.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Radiation exposure.—The use of identical imaging technique factors ensured that the ESE levels for S-F and CR were identical for a given grid choice (435 mR [112.2 µC/kg] with the moving grid and 226 mR [58.3 µC/kg] with the stationary grid). Although CR images with acceptable optical density could have been produced with many different technique factors, we found that optimal image quality, or no increase in image noise, required the use of essentially the same factors as were used for conventional radiographs. A lower ESE value (226 mR [58.3 µC/kg]) was measured for acquisitions with the slip-on grid because of two factors: a shorter source-to-image distance and a lower grid ratio.
Iodine contrast.—As shown in Table 1, the background optical densities were well matched between the two grid types, although there was a very small (0.3 optical density units) decrease in optical density on the CR images compared with the S-F images. The iodine contrast values for each acquisition mode, based on the optical density difference between disks 1 and 5, were essentially identical.
|
|
|
|
|
|
|
Radiation exposure.—On the basis of the ESE values measured on images obtained with various kilovolt peak and milliampere settings, the average ratio of ESE to tube current (in milliroentgen per milliampere [to convert from milliroentgen to microcoulombs per kilogram, multiply by 0.258]) was determined for CT systems A–F (Table 2). The ratios of ESE to tube current for scanner A (multi–detector row CT scanner) were found to be approximately eight times higher than those for scanners B–D (single–detector row scanners), which were all from the same manufacturer. This is because CT SPR images have optimal z-axis resolution if they are acquired with the narrowest beam collimation, which is approximately 1.0–1.5 mm for single–detector row scanners and approximately 8 mm for multi–detector row scanners (11). The wider x-ray beam causes the ratio of ESE to tube current to be approximately eight times higher for the multi–detector row CT scanner. Despite the wider beam collimation with the multi–detector row scanners, the z-axis resolution is not compromised because the detector width of each of the four z-axis detectors is 1.25 mm (11).
|
Maximum and suggested ESE values.—Table 2 also lists the maximum tube current available for the acquisition of CT SPR images, which, when used in conjunction with the ratio of ESE to tube current, determines the maximum patient ESE values at a given kilovolt peak setting. Evaluation of the phantom images obtained with various milliampere values showed that ESE values greater than 150 mR (38.7 µC/kg) at the phantom edge were preferred to avoid a CT SPR image with a grainy appearance. Images acquired with ESE values of approximately 100 mR (25.8 µC/kg) at the phantom edge showed acceptable low-contrast resolution but were noticeably grainy. Images acquired with ESE values less than 70 mR (18.1 µC/kg) at the phantom edge were deemed unacceptable. Thus, only system A provided sufficient ESE levels at 80 kVp. Systems B–D and F provided barely sufficient ESE levels at 120 or 130 kVp for the 21-cm-thick patient-equivalent phantom; thus, the maximum exposure values would likely be insufficient for larger patients. Furthermore, the use of higher kilovolt peak settings decreases the contrast between iodine or calcium and soft tissue.
Iodine contrast at 120 versus 80 kVp.—The difference in iodine, or tin, contrast on CT SPR images obtained with typical tube potential values (120 or 130 kVp) compared with a lower tube potential value (80 kVp, which was available on only systems A–D) is given in Table 3. At 80 kVp, the contrast for the tin disks was twice that at 120 kVp, although background noise at 80 kVp was also about twice that at 120 kVp for the same milliampere setting. Although the data shown in Table 3 are from only system B, the same results were observed for systems A and C. The 80-kVp setting was not tested on system D, because that setting was not calibrated at the time of data acquisition (80 kVp is not used clinically on that system).
|
Edge-enhancement artifacts.—The edge-enhancement image processing used by the manufacturers for CT SPR images is apparent in Figure 3. The processing created the dark banding apparent around high-contrast (white) structures, such as the lead letters that identify the mesh patterns. This ringing of areas with high contrast completely obscured the lead septa in the resolution test patterns, which apparently severely limited spatial resolution. The algorithms used with systems A–E produced similar results, but the algorithm used with system F produced more horizontal and vertical, rather than concentric, artifacts. The images obtained with system F also appeared substantially blurrier than those obtained with systems A–E.
S-F, CR, and CT SPR Findings
Spatial resolution.—Because the conventional test objects for spatial resolution (lead resolution targets) were overly attenuating with the CT systems and created substantial artifacts, we imaged acrylic resolution targets to allow quantification of the spatial resolution of the CT SPR images. Also, the steel wool and bone test objects were used as indicators of the spatial resolution of the CT SPR images. By looking at the bone and steel wool, we could see that the CT SPR images had substantially inferior spatial resolution compared with the S-F or CR images. This subjective finding is in agreement with the quantitative assessment (Table 4). However, the relatively good image quality of the clinical CT SPR abdominal images (Fig 4) demonstrated that this loss of spatial resolution did not necessarily make the CT SPR images unacceptable for clinical use.
|
|
|
|
Geometric accuracy.—Distance measurements along three orientations (z axis, left to right, and diagonal) were within 0.2 mm of the expected values for the S-F, CR, and CT SPR images when the S-F and CR magnification levels were determined at the plane of the test objects and the CT SPR images were acquired with the test objects at isocenter. Because of the divergence of the x-ray beam, all radiographic images are subject to some level of geometric distortion, which is caused by the different magnifications between structures at the anterior versus the posterior aspect of the patient. For CT SPR images, magnification effects were seen for only left-to-right distance measurements. As the object was raised above isocenter, the left-to-right distance measured on the CT SPR image underestimated the true distance. The effect was greater for CT SPR images than for conventional radiographs because of the relatively larger distance between the patient and the CT detectors. However, the height above or below the isocenter did not affect distance accuracy along the z axis because the thin x-ray beam did not diverge substantially along the z axis. Diagonal measurements were affected by the height above or below the isocenter because they contained both a left-to-right and a craniocaudal distance component. Because this magnification of objects affects both dimensions of a radiograph equally, the dimensions measured on a radiograph may be incorrect, but the shape may be represented accurately. On CT SPR images, some degree of geometric distortion may occur away from the isocenter because the craniocaudal and left-to-right dimensions of the objects have different magnification levels.
Clinical CT SPR Abdominal Images and Patient Dose
Figure 4 demonstrates the clinical potential of CT SPR for imaging of the urinary tract. The objectionable artifacts created by highly attenuating structures in the phantom (eg, the lead numbers and resolution targets and the steel mesh) are not present on the clinical images, which do not include such high-contrast structures. Although the spatial resolution was inferior to that with S-F and CR, the contrast resolution of iodinated structures on the CT SPR images was good. Evaluation of the phantom images obtained with various milliampere settings showed that edge ESE values greater than 150 mR (38.7 µC/kg) were preferred to avoid a grainy appearance on the CT SPR images. The CT SPR images acquired at 300 mA delivered a maximum ESE level and effective dose of 330 mR (85.1 µC/kg) and 54 mrem (0.54 mSv [1 mrem = 0.01 mSv]), respectively. In comparison, one abdominal radiograph obtained in a 21-cm-thick patient delivered an ESE level of 412 mR (106.3 µC/kg) and an effective dose of 50 mrem (0.5 mSv), and one abdominopelvic CT scan delivered an ESE level of approximately 2,500 mR (645 µC/kg) and an effective dose of approximately 1,100 mrem (11 mSv) (13).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
With traditional S-F images acquired with a reciprocating (bucky) grid as the reference standard, we concluded that image quality was essentially equivalent with CR or with use of a stationary high-line-density grid. Hence, these techniques appear to be acceptable options for abdominal imaging at the time of CT. However, the use of S-F or CR techniques with a moving grid (ie, table bucky) requires that the patient move from a CT suite to a radiography suite, or vice versa, during the examination.
Modifications made to the CT urography suite at our institution (8,9) eliminated this requirement by allowing the acquisition of S-F or CR images (with a high-line-density stationary slip-on grid) while the patient was lying on the CT table. This approach required the installation of a ceiling-mounted x-ray tube above the CT table and the attachment of an auxiliary CT tabletop, which had a hollow bay under the patient surface in which to place the radiographic cassette, to the manufactured tabletop (9). The clinical advantages of performing radiography at the time of CT may extend beyond CT urography to other CT applications, including trauma, skeletal, chest, and pediatric CT examinations.
CT SPR images acquired with the six different CT scanners at our institution provided spatial resolution inferior to that on S-F or CR images. Additionally, high-contrast objects generated substantial artifacts on the CT SPR images. Compared with S-F images, however, CR and CT SPR images provided improved resolution of small low-contrast objects. Hence, despite the decreased spatial resolution and objectionable artifacts, we believe CT SPR has considerable potential for urographic applications, primarily because of the improved low-contrast resolution.
Radiation doses are also important in the establishment of the optimal type and number of abdominal radiographs to be acquired at CT urography. Our results indicated that CT SPR images with acceptable levels of image contrast and noise required a midline ESE level of about 300 mR (77.4 µC/kg) and a tube voltage of 80 kVp. This exposure was slightly less than that required for S-F or CR with a table bucky grid (435 mR [112.2 µC/kg]) and slightly higher than that required for S-F or CR with a stationary tabletop grid (226 mR [58.3 µC/kg]). On the basis of findings in our phantom study, which evaluated ESE versus acceptable noise levels for six CT systems, we recommend use of a tube potential of 80 kVp and midline exposure levels of about 300 mR (77.4 µC/kg), which could be obtained with only system A and a tube current of approximately 300 mA. These settings delivered a maximal ESE level of 330 mR (85.1 µC/kg) and an effective dose of 54 mrem (0.54 mSv) for a phantom thickness equivalent to a 21-cm thick patient, which represents a relatively thin patient. As with S-F or CR for comparable image noise, the ESE and effective dose values must increase with patient thickness, approximately doubling for each additional 4–5 cm of patient thickness. The image quality parameters reported herein would not vary much with increasing patient thickness, providing the ESE level was appropriately increased to keep the image noise constant.
In summary, S-F and CR abdominal imaging techniques provided better spatial resolution than did CT SPR; however, CT SPR images provided improved low-contrast resolution compared with S-F images, at exposure levels comparable to those with S-F or CR techniques. We believe that improvements in spatial resolution and artifact reduction for high-contrast objects would substantially increase the clinical usefulness of CT SPR, for both CT urography and other CT imaging applications. Toward that goal, we are working to optimize CT SPR postprocessing algorithms (14).
These technical considerations demonstrate some potential advantages and disadvantages with CT SPR as compared with S-F or CR for CT urographic examinations. However, further work is required to determine the diagnostic accuracy of CT SPR images compared with the reference standard S-F images. Ongoing work at our institution has been designed to help determine the clinical ability of CT SPR images to depict clinical disease states, as our modified CT suite allows closely spaced (in time) acquisition of both CT SPR and S-F images.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, all authors; study concepts, C.H.M., A.J.L., B.F.K., T.J.V., R.R.H.; study design, C.H.M.; literature research, C.H.M., T.J.V.; clinical studies, A.J.L., B.F.K.; experimental studies, C.H.M., M.R.B.; data acquisition and analysis, C.H.M., M.R.B.; statistical analysis, C.H.M.; manuscript preparation, C.H.M.; definition of intellectual content, C.H.M.; manuscript editing, C.H.M., B.F.K.; manuscript review and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. G. Silverman, J. R. Leyendecker, and E. S. Amis Jr What Is the Current Role of CT Urography and MR Urography in the Evaluation of the Urinary Tract? Radiology, February 1, 2009; 250(2): 309 - 323. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. C. Chow, S. W. Kwan, E. W. Olcott, and G. Sommer Split-Bolus MDCT Urography with Synchronous Nephrographic and Excretory Phase Enhancement Am. J. Roentgenol., August 1, 2007; 189(2): 314 - 322. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. S. Sudakoff, D. P. Dunn, R. S. Hellman, M. A. Laguna, C. R. Wilson, R. W. Prost, D. C. Eastwood, and H. J. Lim Opacification of the Genitourinary Collecting System During MDCT Urography with Enhanced CT Digital Radiography: Nonsaline Versus Saline Bolus Am. J. Roentgenol., January 1, 2006; 186(1): 122 - 129. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Norrman and J. Persliden A factorial experiment on image quality and radiation dose Radiat Prot Dosimetry, May 17, 2005; 114(1-3): 246 - 252. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M Noroozian, R H Cohan, E M Caoili, N C Cowan, and J H Ellis Multislice CT urography: state of the art Br. J. Radiol., December 1, 2004; 77(suppl_1): S74 - S86. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Kawashima, T. J. Vrtiska, A. J. LeRoy, R. P. Hartman, C. H. McCollough, and B. F. King Jr CT Urography RadioGraphics, October 1, 2004; 24(suppl_1): S35 - S54. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. D. Nawfel, P. F. Judy, A. R. Schleipman, and S. G. Silverman Patient Radiation Dose at CT Urography and Conventional Urography Radiology, July 1, 2004; 232(1): 126 - 132. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J.-K. Kim, S.-Y. Park, H.-j. Kim, C.-S. Kim, H.-J. Ahn, T.-Y. Ahn, and K.-S. Cho Living Donor Kidneys: Usefulness of Multi-Detector Row CT for Comprehensive Evaluation Radiology, December 1, 2003; 229(3): 869 - 876. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
