| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Experimental Studies |
1 From the Dept of Radiology, Samsung Med Ctr, Sungkyunkwan Univ School of Med, Seoul, Korea (T.S.K.); Dept of Radiology, Seoul National Univ Coll of Med and Institute of Radiation Med, Seoul National University Med Research Ctr, 28 Yongon-dong, Chongro-gu, Seoul 110-744, Korea (J.G.I., J.M.G., K.H.L., Y.J.L., S.H.K.); and Samsung Biomedical Research Institute, Samsung Med Ctr, Seoul, Korea (S.K.). Received Jun 20, 2001; revision requested Aug 13; revision received Sep 19; accepted Oct 16. Supported in part by grant HMP-98-G-1-014 of the Highly Advanced National Project, Ministry of Health and Welfare, Republic of Korea. Address correspondence to J.G.I. (e-mail: imjg@radcom.snu.ac.kr).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Oleic acid was injected intraatrially into three pigs (weight, 20–25 kg) at doses of 0.04, 0.05, and 0.06 mL/kg to induce pulmonary edema. Thirty-seven sets of computed radiographic, digital radiographic, and thin-section computed tomographic (CT) scans were obtained every 20–30 minutes in three pigs over 4–6 hours. Images were masked for identity, randomly sorted, and displayed on a monitor. Four radiologists rated each image for presence of parenchymal opacities by using a dichotomous scoring system in two sessions. Presence of pulmonary edema was determined with thin-section CT and a severity scale. Intra- and interobserver variations were determined with the 
statistic and the Z test and with the Cochran Q test and the McNemar test, respectively. True-positive, true-negative, false-positive, and false-negative rates were determined. McNemar test was used to determine statistical significance of differences in detection between computed and digital radiographic images.
RESULTS: There was no significant intra- or interobserver variation, except for one pair of observers during the first interpretative session with computed radiographic images (P = .016, McNemar test). Overall sensitivity (92.1%) and diagnostic accuracy (90.2%) of digital radiography were significantly higher than those of computed radiography (79.6% and 83.4%, respectively) (P < .001 for sensitivity, P = .01 for diagnostic accuracy, McNemar test). In detection of minimal and mild pulmonary edema, sensitivity of digital radiography (84%) was significantly higher than that of computed radiography (58%) (P < .001).
CONCLUSION: Soft-copy digital radiographic images are superior to soft-copy computed radiographic images obtained for detection of mild pulmonary edema in pigs.
© RSNA, 2002
Index terms: Animals • Computed tomography (CT), high-resolution, 60.12118 • Experimental study • Lung, edema, 60.71 • Radiography, comparative studies • Radiography, digital, 60.1215 • Radiography, selenium detector, 60.1215 • Radiography, storage phosphor, 60.1215
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Computed radiography with a storage phosphor plate has made digital chest radiography possible with image qualities that are comparable with those of conventional screen-film systems (4–7). Recently, with the rapid development of electronic and computer technology, a new generation of digital x-ray detectors (large-area, flat-panel detectors with integrated, thin-film transistor readout mechanisms) have been investigated and developed (8–12).
The purpose of our study was to compare the diagnostic accuracy of soft-copy digital radiographic (amorphous selenium-based flat-panel system) and soft-copy computed radiographic (storage phosphor system) images obtained for the detection of pulmonary edema in a pig model.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Animal Experimentation
Anesthesia was induced with the intramuscular injection of a mixture of 7 mg per kilogram of body weight of ketamine hydrochloride (Ketara; Yuhan Yanghang, Seoul, South Korea) and 2.3 mg/kg of xylazine hydrochloride (Rompun; Bayer Korea, Seoul, South Korea) and was maintained with the intravenous injection of 1.3 mg/kg of zolazepam hydrochloride (Zoletil; Virbac, Carros, France). Animals were not intubated. A 5-F catheter was introduced through the right external jugular vein for the intraatrial injection of oleic acid to induce permeability pulmonary edema. Prior to the injection of the oleic acid, baseline computed radiographic, selenium-based digital radiographic, and thin-section computed tomographic (CT) scans were obtained.
Permeability edema was induced with the intraatrial injection of commercially available oleic acid (C18H34O2; Sigma, Steinheim, Germany) through the external jugular catheter (at doses of 0.04, 0.05, and 0.06 mL/kg as a bolus or as subdivided injections). Immediately after the injection of oleic acid, computed radiographic, selenium-based digital radiographic, and thin-section CT scans were obtained. The imaging studies were performed rapidly to minimize time delays. The interval between computed radiography and selenium-based digital radiography was as short as 1 minute because the two radiographic units were located in the same room. Immediately after computed radiography and selenium-based digital radiography, the pigs were rapidly moved into the nearby CT room for CT scanning. The study sequence of computed radiography, selenium-based digital radiography, or thin-section CT was randomly selected to avoid bias. In total, each set of computed radiographic, selenium-based digital radiographic, and thin-section CT scans was obtained in 10 minutes. Subsequently, a set of computed radiographic, selenium-based digital radiographic, and thin-section CT scans was obtained every 20–30 minutes over 4–6 hours. A total of 37 sets of images were obtained in three pigs (10, 11, 16 sets of images per pig). Each set of images included one computed radiographic image, one digital radiographic image, and one thin-section CT scan obtained during each session.
Computed Radiography and Digital Radiography
Posteroanterior chest radiographs were obtained with computed radiographic and selenium-based digital radiographic systems that were located in the same room. Two Bucky stands were set up at the opposite sides of the same room for each detector system. Computed radiographic images were obtained with an imaging unit (FCR-9000; Fuji, Tokyo, Japan). A 35 x 43-cm imaging plate (ST-V; Fuji) with a matrix of 1,760 x 2,140 x 10 bit and a pixel size of 0.2 mm was used. The selenium-based digital radiographic images were obtained by using a unit (DirectRay; Direct Radiography, Newark, Del) with a 35 x 43-cm solid-state detector with a matrix of 2,560 x 3,072 x 12 bit and a pixel size of 0.139 mm. Radiography was performed in each pig with the selenium-based digital radiographic system and then immediately after with the computed radiographic system (or vice versa).
The radiographs were produced by using the same tube and generator and at the same exposure settings, which were 80 kVp and 250 mA, with an exposure time of 50 msec and a 180-cm focus-detector distance. Both imaging systems included a moving 10:1 antiscatter grid (103 lines per inch). The x-ray beam was collimated onto the pig’s chest. Immediately after the radiographs were obtained, a thin-section CT scan of the chest was obtained, or vice versa. The same technique and setting that were used to obtain the baseline thin-section CT scan were used to obtain the radiographs.
Thin-Section CT Scanning
Thin-section CT scans were obtained with a scanner (Somatom Plus 4; Siemens, Erlangen, Germany) with a field of view of 20–22 cm, a 512 x 512 matrix, an exposure of 140 kVp and 170 mA, and a 0.75-second scanning time. Pigs were scanned in the prone position from the thoracic inlet to the level of the diaphragm with a 10-mm interval and a 1-mm section thickness. After scanning, the images were reconstructed by using a high-spatial-frequency algorithm.
Image Acquisition and Display
Digital data were sent to a PACS server (Radmax; MaroTech, Seoul, Korea) and then distributed to display workstations. Images were downloaded onto the local hard disk drive of the display workstation before they were viewed by the radiologist. The size of each Digital Imaging and Communications in Medicine, or DICOM, file of the computed radiographic and selenium-based digital radiographic images was 7.18 and 15.0 Mbyte, respectively. A 21-inch video monitor (DR110; Dataray, Denver, Colo) with 2,048 x 2,560 x 8-bit pixels was used in a darkened room for image display. The monitor operated at 71 Hz in an interlaced mode and had a maximum brightness level of 100 foot-lambert. About 10% of the display area was allocated for the title and menu bars, and the remaining display area (2,048 x 2,300 pixels) was slightly large for the computed radiographic data and small for the selenium-based digital radiographic data.
Therefore, the computed radiographic images were enlarged by 7% by using pixel replication, and the selenium-based digital radiographic images were reduced by 33% by using pixel subsampling to fit the remaining monitor display area. Because the video monitor could only display an 8-bit gray scale, the gray scale of the digital images was modified by using a 12- to 8-bit (selenium-based digital radiographic image) or 10- to 8-bit (computed radiographic image) look-up table. The soft-copy images were displayed without unsharp masking. Only the window widths and the image levels were optimized automatically with a customized program, which produced the same density for the computed radiographic and selenium-based digital radiographic images. No other image postprocessing was performed. Observers were allowed to adjust the brightness and contrast of the images. For this study, pig identification was obscured on all images and replaced by a sequence number. Computed radiographic and selenium-based digital radiographic images were displayed in a random manner.
Image Interpretation
Four chest radiologists (T.S.K., J.M.G., Y.J.L., S.H.K.) served as observers for the study and evaluated the images independently. All 74 radiographic images (37 computed radiographic images and 37 selenium-based digital radiographic images) were masked for identity and assigned randomly to prevent selection bias, and they were viewed with a single video monitor of a PACS workstation. Observers determined whether opacity suggestive of pulmonary edema was present on each chest radiograph and gave only a dichotomous "yes" or "no" response. To evaluate intraobserver variation, a second session of interpretation was performed 1 week later in the same manner.
All 37 thin-section CT scans of the chest were evaluated by two chest radiologists (T.S.K., J.G.I.), and decisions about the findings were reached with a consensus (consensus needed for one of 37 scans). The presence of pulmonary edema was considered positive when areas of ground-glass attenuation and/or consolidation were depicted on thin-section CT scans (14). The severity of pulmonary edema on thin-section CT scans was graded according to a five-point scale by first selecting a section of the thin-section CT scan in which the area of ground-glass attenuation and/or consolidation was most prominent. The ratio of the area of parenchymal abnormality to the entire parenchymal area was visually estimated and classified according to the five-point scale as follows: grade 0, less than 1% (normal); grade 1, 1%–10% (minimal pulmonary edema); grade 2, 11%–40% (mild); grade 3, 41%–70% (moderate); and grade 4, more than 70% (severe) (Figs 1–3).
|
|
|
|
|
|
|
|
|
statistic with a Z test was used to test for intraobserver variation between the first and second sessions for each reader. Interobserver variation was also determined with the results from the first and second sessions for each modality by using the Cochran Q test (SAS version 6.12; SAS Institute, Cary, NC). When the Cochran Q test resulted in a P value that indicated a significant difference (P 
.05), each pair was tested individually by using the McNemar test. Exact P values were also calculated with the McNemar test. The presence of pulmonary edema was established with the CT scan of the chest. True-positive, true-negative, false-positive, and false-negative rates were determined for computed radiography and selenium-based digital radiography. The McNemar test was used to determine the statistical significance of differences between computed radiography and selenium-based digital radiography in terms of sensitivity, specificity, and diagnostic accuracy. The statistical significance of the sensitivity values of computed radiography and selenium-based digital radiography in thin-section CT grades 1 and 2, which corresponded to minimal and mild pulmonary edema, was also determined.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
coefficients for the two sessions for computed radiography and selenium-based digital radiography were 0.779 and 0.782, respectively. No significant intraobserver variation was found for computed radiography or selenium-based digital radiography (P < .001, Z test, both computed and selenium-based digital radiography). The radiologist disagreement rates between observers and for each imaging modality for the two interpretation sessions are summarized in Table 3. With the Cochran Q test, which was used to evaluate interobserver variation, significant variation was determined only in the first session of computed radiographic interpretation (P = .021), and the McNemar test demonstrated that this significant variation was caused by observers 3 and 4 (P = .016). No other significant interobserver variation was found for either computed radiography or selenium-based digital radiography (P > .05).
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
In principle, this approach has a potential image quality as high as or higher than that of computed radiography (19). In comparison with the screen-film system and computed radiography, the selenium detector is characterized by a higher detective quantum efficiency and resolution (20). The selenium drum detector, which is one of the new digital x-ray detectors, produced images that were superior to those of conventional screen-film systems (18,20–24) and the computed radiographic system (25,26) in terms of the depiction of anatomic regions. All new digital detector systems must be verified for diagnostic accuracy and performance at a soft-copy–viewing workstation before they are used clinically in a PACS environment. Although physical measurements give an indication of the overall performance of the system, observer studies are the most conclusive way of determining the overall performance of a new system (11). Recently, Goo et al (27) showed that the soft-copy images produced with flat-panel selenium-based digital radiography were perceived as equal or superior to those produced with computed radiography in most anatomic regions of the thorax.
The results of our comparative study of soft-copy computed radiographic images and selenium-based digital radiographic images obtained for the detection of experimental pulmonary edema showed that the sensitivity and diagnostic accuracy of selenium-based digital radiography are significantly higher than those of computed radiography, which correlates closely with the results of Goo et al (27) and those in previous articles in which hard-copy digital and screen-film radiographic and/or computed radiographic images were compared (21,22,25,28). These results are attributed to the superiority of digital radiography versus computed radiography in terms of spatial resolution and gray scale. Theoretically, digital radiography, which has a matrix of 2,560 x 3,072 pixels (139 x 139 µm per pixel), can depict more fine details than can computed radiography with a matrix of 1,760 x 2,140 pixels (200 x 200 µm per pixel). Pixel size is an important parameter in digital radiography because it directly influences the spatial resolution of images, particularly in the depiction of fine detail (11).
Another explanation for the better performance of selenium-based digital radiography is related to the absence of light scattering within the detector. Even if other factors such as the matrix and pixel sizes were equal, sharper images could be obtained with selenium-based digital radiography than with screen-film radiography or computed radiography. The conversion of x-ray photons to electrical charges and electrical data is direct by means of arrays of semiconductor elements without the intervening light stage, such as in an intensifying screen or a photostimulable phosphor imaging plate. The latter are used in the screen-film system and computed radiography, respectively. In the screen-film system and computed radiography, light scattering of intermediate light fluorescence results in image blurring (27,29).
Another important factor related to diagnostic accuracy in the detection of pulmonary abnormalities on chest radiographs is the image gray scale. The number of gray levels in a digital system determines how well it reproduces subtle contrast differences. Selenium-based digital radiographic images are digitized in 12-bit gray scale (4,096 shades), whereas most of the currently used computed radiographic systems provide 10-bit images (1,024 shades). Some computed radiographic systems that can create 12-bit images are now also commercially available. Therefore, selenium-based digital radiography can, theoretically, more accurately depict the subtle variations in attenuation (28). This concept correlates with the results of our study, namely, that the soft-copy selenium-based digital radiographic images showed significantly higher sensitivity and diagnostic accuracy than did the soft-copy computed radiographic images obtained for the detection of experimental pulmonary edema. Moreover, this superiority of the selenium-based digital radiographic images was more pronounced in cases of minimal and mild pulmonary edema.
Although both the computed radiographic and selenium-based digital radiographic images were displayed on the same video monitor, which used an 8-bit gray scale display (256 shades) after contrast scaling with a look-up table, the original image depth could be readily reviewed by using the workstation mouse to adjust the window level and image width. In our study, observers rarely adjusted the contrast themselves during the examination of images. Therefore, we believe that even the downscaled display might have reflected the original image depth of selenium-based digital radiographic images and that the sensitivity of these images could be improved with this procedure. Although the spatial resolution of selenium-based digital radiographic images is better than that of computed radiographic images with a factor of 1.4 on the original images, the limitations imposed by using the monitor must have decreased the superior resolution of selenium-based radiographic images.
The exposure setting used for computed radiographic and selenium-based digital radiographic images in our study, namely, 80 kVp, 250 mA, 50-msec exposure time, and a 180-cm focus-detector distance, with a moving 10:1 antiscatter grid, was identical to that used in Seoul National University Hospital, Korea, for radiography of the chest in erect children who weighed 20–30 kg.
We were able to create a range of subtle to obvious pulmonary edema by intraatrially injecting variable doses of oleic acid in a pig model. No significant intra- or interobserver variation was observed, except between observers 3 and 4 during the first session of interpretation of computed radiographic images (P = .016). These results demonstrate the high reliability of the results obtained after image interpretation.
Although we masked the images for identity and randomly assigned the whole images, there was a possibility that observers could have distinguished selenium-based digital radiographic images from computed radiographic images, because the selenium-based digital radiographic images generally seemed to have a slightly lower signal-to-noise ratio (quantum mottling), and this ability to discriminate might have resulted in some degree of bias. In computed radiographic images, the quantum mottles might have been blurred out, ironically, in the light-scattering process, working as an image filtering with an increase of signal-to-noise ratio, though this process causes a simultaneous degradation in image spatial resolution (27). However, such a scattering process does not occur in selenium-based digital radiography, and the minimal quantum mottles observed on selenium-based digital radiographic images might have influenced our results, because the specificity of computed radiography was significantly higher than that of selenium-based digital radiography, although the P value of .046 was borderline.
Before the clinical application of flat-panel selenium-based digital radiography is attempted, further work is needed to optimize the exposure techniques to improve image quality and reduce radiation dose. Findings in this study demonstrated that the soft-copy selenium-based digital radiographic images allowed significantly higher sensitivity and diagnostic accuracy than did the soft-copy computed radiographic images obtained for the detection of experimental pulmonary edema and that this superiority was more pronounced in cases of mild pulmonary edema. This digital detector system is expected to improve the detection of pulmonary lesions with subtle contrast differences.
In conclusion, results of this comparative study of soft-copy computed radiographic images and selenium-based digital radiographic images obtained for the detection of experimental pulmonary edema showed that the sensitivity and diagnostic accuracy of selenium-based digital radiography were significantly higher than those of computed radiography.
Practical application: Results of this study of detection of pulmonary edema, a common chest radiographic pattern in patients in the intensive care unit, induced in a pig model demonstrated that selenium-based digital radiography can outperform computed radiography in terms of its superior sensitivity and diagnostic accuracy. Thus, selenium-based digital radiography has potential for use in patients in the intensive care unit.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, J.G.I.; study concepts, J.G.I.; study design, T.S.K., J.G.I.; literature research, T.S.K.; experimental studies, T.S.K., K.H.L.; data acquisition, K.H.L.; data analysis/interpretation, all authors; statistical analysis, S.K.; manuscript preparation, T.S.K.; manuscript definition of intellectual content, J.G.I.; manuscript editing, T.S.K., J.G.I.; manuscript revision/review, J.M.G., J.G.I.; manuscript final version approval, J.G.I.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Korner, C. H. Weber, S. Wirth, K.-J. Pfeifer, M. F. Reiser, and M. Treitl Advances in Digital Radiography: Physical Principles and System Overview RadioGraphics, May 1, 2007; 27(3): 675 - 686. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Bacher, P. Smeets, L. Vereecken, A. De Hauwere, P. Duyck, R. De Man, K. Verstraete, and H. Thierens Image quality and radiation dose on digital chest imaging: comparison of amorphous silicon and amorphous selenium flat-panel systems. Am. J. Roentgenol., September 1, 2006; 187(3): 630 - 637. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. P. Busch and K. Faulkner Image quality and dose management in digital radiography: a new paradigm for optimisation Radiat Prot Dosimetry, December 1, 2005; 117(1-3): 143 - 147. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
