| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, Asan Medical Center, University of Ulsan College of
Medicine, 388-1 Poongnap-dong, Songpa-gu, Seoul, 138-736, Korea.
2
Department of Neonatology, Asan Medical Center, University of Ulsan College of
Medicine, Seoul, 138-736, Korea.
Received November 20, 2000;
accepted after revision February 20, 2001.
Address correspondence to C. H. Yoon.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Two radiologists reviewed 82 soft-copy neonatal chest radiographs before and after the application of edge enhancement on our picture archiving and communication system (PACS). The visibility of a pneumothorax (n = 22), central venous catheter (n = 32), umbilical arterial catheter (n = 36), endotracheal tube (n = 40), and normal anatomic structures (the minor fissure, anterior segmental bronchus of the right upper lobe, and aortic arch, n = 57) was evaluated. Six of 22 soft-copy images depicting a pneumothorax were excluded from the evaluation of image quality either because of the large size of the pneumothorax itself (n = 7) or because of the lack of confirmatory evidence that would have been provided by an additional lateral decubitus (n = 6) or cross-table lateral radiograph (n = 3). Image quality was evaluated by visual grading analysis.
RESULTS. The visibility of a pneumothorax (p < 0.01), vascular catheters (p < 0.001), the minor fissure (p < 0.001), and the anterior segmental bronchus of the right upper lobe (p < 0.001) improved significantly after applying edge enhancement to soft copies of neonatal chest radiographs, whereas the visibility of the aortic arch did not improve. Evaluations of the improvements in the visibility of the endotracheal tube were inconsistent.
CONCLUSION. Application of edge enhancement to soft copies of neonatal chest radiographs helps radiologists to identify small pneumothoraces, vascular catheters, and delicate normal structures, thereby improving the detection of subtle chest findings in the neonatal intensive care unit.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The full picture archiving and communication system (PACS) requires radiologists to rely on soft copies of neonatal chest radiographs to perform these tasks. Digital images may suffer from the lower spatial resolution, and therefore, a properly processed digital image is mandatory for depiction of subtle abnormalities [1]. Because edge enhancement increases the sharpness of structures containing high-spatial-frequency information, we assume that it may be more useful for the detection of subtle findings in neonates than in older children or adults. A pleural line in a pneumothorax and a narrow-lumen vascular catheter are among the more delicate structures appearing on a neonatal chest radiograph, and so detecting them may be challenging, particularly on an unprocessed soft copy. Therefore, our study was performed to evaluate whether edge enhancement could improve the visibility of subtle findings on soft-copy neonatal chest radiographs.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The suspected presence of a small pneumothorax on a anteroposterior chest radiograph was confirmed by an additional lateral decubitus radiograph of the chest (n = 6) or a cross-table lateral radiograph (n = 2) obtained later the same day. The absence of pneumothorax was confirmed by a cross-table lateral radiograph for one patient. These additional radiographs were not obtained for all patients because on seven of the soft copies, the presence of a large pneumothorax was obvious. Therefore, these 16 soft copies were included in the statistical analysis for the evaluation of the visibility of a pneumothorax. We were able to determine the presence of vascular catheters (central venous catheters and umbilical arterial catheters) because of their characteristic shapes. The locations of the tips of vascular catheters and endotracheal tubes were traced to detect malpositioned devices. For comparisons involving the visibility of normal anatomic structures, we excluded radiographs that revealed such diffuse parenchymal abnormalities as hyaline membrane disease and bronchopulmonary dysplasia.
To avoid order bias, we performed two review sessions 6 months apart. In the first review session, the default (unprocessed soft copy) image was evaluated first and then the postprocessed (edge-enhanced) image of the same structure. The order was reversed in the second review session. All computed chest radiographs were obtained with a FUJIX FCR 9000 system (Fuji Photo Film, Tokyo, Japan). The gray scale was 10 bits per pixel. The soft copies were displayed with a pixel resolution of 2560 x 2048 on a 21-inch (53-cm) PACS monitor (Barco, Kortrijk, Belgium).
Our PACS has three postprocessing modes that can be applied to a default image, but our experience in clinical practice led us to conclude that images postprocessed with either of two of the modes—smoothing and moderate edge-enhancing—are not much clearer than the default images. Therefore, we used images that had been postprocessed with the strong edge-enhancing mode for our comparisons. The strong edge-enhancing mode in our PACS is shown in Figure 1. It is the middle selection among five standard sharpening algorithms (DIMPL Library; DOME Imaging Systems, Waltham, MA).
|
In evaluating the visibility of a given structure on an image, the radiologists used a subjective visual grading analysis: grade 1, no visualization; grade 2, blurred visualization; and grade 3, clear visualization. Before beginning the subjective assessment of image quality, the two radiologists used five randomly selected test chest radiographs to standardize their visual grading system.
Conditions that could have affected the evaluation of the soft copy were maintained at constant settings during visual grading analysis. The luminance of the display was kept at the same level under continuously low ambient lighting. Image window and level settings were not manipulated, and magnification of images was not used. The statistical analysis was performed using the Wilcoxon's signed rank test, and a value of p less than 0.05 was considered statistically significant.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Either one or the other of the two radiologists saw improvement in the visibility of pneumothoraces on seven of 16 soft-copy images after edge enhancement. Furthermore, pneumothoraces on three soft copies not detected by the radiologists before edge enhancement (Fig. 2A) were confidently diagnosed by the radiologists on the basis of the same soft-copy images after edge enhancement (Fig. 2B). Later the same day, an additional lateral decubitus radiograph confirmed the presence of a pneumothorax in the patients from whom these soft copies had been obtained (Fig. 2C).
|
|
|
Malpositioned central venous catheters were visible on three of 32 soft copies that depicted the devices. Among these, two of the malpositioned catheters were not identified until the application of edge enhancement (Fig. 3A,3B). Malpositioning of the umbilical arterial catheter or of the endotracheal tube was not observed.
|
|
The second review session with reversed order showed the improved visibility on the edge-enhanced soft copies for all evaluated structures except the aortic arch (Table 2).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
There have been studies in which various image-processing algorithms to enhance lesion conspicuity have been applied to computed chest radiographs of adults, but the results of such studies have been contradictory [1, 3,4,5,6,7,8]. Some reports have indicated that the application of image processing improved the visualization of subtle findings and thereby provided an increased confidence level of the interpretation [1, 4,5,6, 8]. According to other reports, the effect of image processing did not greatly influence diagnostic performance [3, 7]. Franken et al. [9] evaluated the effect of manipulating window and level settings, magnification, gray-scale inversion, and magic window for local magnification on soft copies of neonatal chest and abdomen images. However, to the best of our knowledge, the application of edge enhancement to soft copies of neonatal chest radiographs has not been evaluated.
A smaller pixel size of the computed radiograph, ranging from 0.1 to 0.4 mm, was required to identify subtle findings—such as small pneumothoraces and septal lines—in adults [10, 11]. In our study, the pixel size of the computed radiograph was within the previously reported range required for identification of subtle findings. For neonatal computed radiographs, however, an even smaller pixel size may be required. Moreover, using edge enhancement can compensate for the limited spatial resolution of a computed radiograph.
We can exclude the possibility of interpreter order bias because the results of the second review session, which was conducted with the images in reverse order, showed the same improvements of the visibility of subtle findings after edge enhancement as those of the first review session. Both radiologists preferred the edge-enhanced soft copy as an aid in identifying subtle findings even when they saw the postprocessed image first. In the first session, the two radiologists had inconsistent results for images of the endotracheal tube. However, they consistently recorded the improved visibility of the endotracheal tube after edge enhancement in the second review session. This difference may reflect a learning curve for the use of edge-enhanced soft copies. The visibility of the aortic arch was not assessed as improved in either review session, probably because the aortic arch contains mostly low-spatial-frequency information.
Although manipulating window and level settings can compensate for the unsatisfactory quality of improperly exposed computed radiographs, such manipulation requires extra time [9, 12]. In our study, image window and level settings were not manipulated. Nevertheless, rapid and improved identification of subtle findings on soft-copy neonatal chest radiographs was possible because the response time of the edge enhancement was less than 1 sec. Therefore, we speculate that the method has appropriate clinical uses. Despite such an advantage of the edge-enhanced computed radiograph, other investigators have reported that the post-processed image was inferior to the default image in depicting other pulmonary abnormalities such as nodular, micronodular, and mediastinal structures [3, 4]. Therefore, we believe that the edge-enhanced soft copy supplements the unprocessed soft copy but is not a substitute for it.
Our study may be limited by the small number of soft copies studied. However, at the beginning of this study, we decided that at least 30 soft copies in each category were required for statistical analysis. We met this requirement in all categories except images showing pneumothorax, probably because of the relatively low incidence of this condition. Increasing the sample size used in a future study will strengthen statistical results.
Subjective visual grading analysis may be a second limitation of our study. Subjective assessment can be sometimes misleading if individual grading differences among interpreters are large enough to affect results. Objective analyses such as receiver operating characteristic analysis can be used to determine observer performance when an independent means of determining the true value of an image of a given structure is available [6]. However, we believe that our subjective method is probably sufficient to evaluate the effect of edge enhancement qualitatively.
In conclusion, we found that the application of edge enhancement to soft copies of neonatal chest radiographs helps radiologists to identify small pneumothoraces, narrow-lumen catheters, and delicate normal anatomic structures, thus improving the detection of subtle imaging findings in neonates.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  E. A. Krupinski, G. J. Becker, D. Laszakovits, A. M. Gerdeman, and M. G. Evanoff Evaluation of Off-the-Shelf Displays for Use in the American Board of Radiology Maintenance of Certification Examination Radiology, March 1, 2009; 250(3): 658 - 664. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Don, B. R. Whiting, J. S. Ellinwood, D. H. Foos, K. A. Kronemer, and R. A. Kraus Neonatal Chest Computed Radiography: Image Processing and Optimal Image Display Am. J. Roentgenol., April 1, 2007; 188(4): 1138 - 1144. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B Molander, H-G Grondahl, and A Ekestubbe Quality of film-based and digital panoramic radiography Dentomaxillofac. Radiol., January 1, 2004; 33(1): 32 - 36. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
