Published online before print June 21, 2005, 10.1148/radiol.2362041558
(Radiology 2005;236:677-684.)
© RSNA, 2005
CT Depiction of Pulmonary Emboli: Display Window Settings1
Kyongtae T. Bae, MD, PhD,
Gita N. Mody, BS,
Dennis M. Balfe, MD,
Sanjeev Bhalla, MD,
David S. Gierada, MD,
Fernando R. Gutierrez, MD,
Christine O. Menias, MD,
Pamela K. Woodard, MD,
Jin Mo Goo, MD and
Charles F. Hildebolt, DDS, PhD
1 From the Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 S Kingshighway Blvd, St Louis, MO 63110. Received September 8, 2004; revision requested November 15; revision received November 24; accepted December 21.
Address correspondence to K.T.B. (e-mail: baet{at}mir.wustl.edu).
 |
ABSTRACT
|
|---|
PURPOSE: To compare computed tomographic (CT) window settings selected by radiologists with those determined by using two alternative approaches for depiction of pulmonary emboli (PE).
MATERIALS AND METHODS: Institutional review board approval was obtained; informed consent was not required. This study was compliant with the Health Insurance Portability and Accountability Act. Twenty-five clinical chest CT studies were obtained with a standardized PE protocol and retrospectively evaluated by five chest and two body CT radiologists. Of these studies, 13 were positive for PE, and 12 were negative. At the main pulmonary artery (PA), mean attenuations (MPA) and standard deviations (SDPA) were measured. Initially, images were displayed with a standard mediastinal window setting (window width, W = 400 HU; window center, C = 30 HU), and each observer adjusted the setting to a personally preferred setting (eg, "personal") for PE detection. Images displayed at this setting were compared in a side-by-side fashion with the "modified" (W = MPA + 2 · SDPA, C = W/2) and "double-half" (W = 2 · MPA, C = MPA/2) window setting. Each observer rated images from 1 (ie, most preferred) to 3 (ie, least preferred). For quantitative analysis, window width and center value of each setting were divided by corresponding MPA to compute a width ratio and a center ratio. Window settings and ratings were compared with repeated-measures analysis of variance, paired t tests, and Wilcoxon signed-rank tests.
RESULTS: Ratings for all three types of window settings were significantly different (P < .001). Observers preferred their personal settings the most and the modified settings the least. Mean ratios for the seven observers were 1.68 ± 0.20 for window width and 0.47 ± 0.08 for window center. Window width ratios for all settings were significantly different from each other (P < .001). Window center ratios were significantly higher for the modified setting than for the double-half setting (P = .013). Values for mean PA attenuation were correlated with window width ratios for six (86%) observers (mean r2 value = 0.29 ± 0.19, P 
.03) and with window center ratios for four (57%) observers (mean r2 value = 0.16 ± 0.14, P .02), thus indicating a trend of setting window width and window center higher when contrast enhancement is lower and vice versa.
CONCLUSION: On average, observers selected CT window settings for PE detection at a window width of slightly less than twice the mean PA attenuation and at a window center of about half the mean PA attenuation. Observers tended to use larger window widths and centers as the degree of PA enhancement was lower.
© RSNA, 2005
|
INTRODUCTION
|
|---|
Computed tomography (CT) is gaining increased acceptance as a first-line modality in the diagnosis of acute pulmonary emboli (PE). Advances in multi–detector row CT have led to improved delineation of peripheral pulmonary arteries (PAs) and detection of small PE, thereby increasing the sensitivity and specificity of this technique (1,2). High-spatial-resolution multi–detector row CT data lend themselves to multiplanar visualization and allow us to evaluate small branches of pulmonary vessels. Despite these advances in image acquisition techniques and image quality, for diagnosis of PE image interpretation is still based on cine review of individual transverse sections that are displayed at selected window settings at digital workstations (3).
Some PE are depicted poorly with standard display window settings, particularly when intense contrast medium enhancement in the PA obscures subtle PE. Thus, radiologists often adjust the window settings to enhance visualization of PE (1,4). Brink et al (5) proposed a modified window setting method that is based on measurement of the mean and standard deviation of the main PA attenuation and was developed with porcine experimental data. This method, however, tends to produce narrow window widths; therefore, it is not commonly used in clinical practice. Raptopoulos and Boiselle (1) used a fixed window setting (window width, 1000 HU; window level, –100 HU) and reported that this setting, compared with standard mediastinal or lung window settings, significantly improved vessel visualization in the peripheral zone on clinical CT images. No explanation as to how this window setting was established was provided.
Determination of optimal window settings applicable to all patients undergoing CT for detection of PE is confounded by many factors (5). The attenuations of enhanced PAs that contain PE are variable; they depend on contrast medium injection protocols, patient body habitus and hemodynamic status, and pulmonary vessel sizes and positions. We think that personal preferences and criteria used to determine optimal window settings may be highly variable among radiologists who interpret PE CT images. After preliminary attempts, we realized that the task of formulating an objective method that would allow us to determine acceptable window settings is extremely difficult, if not impossible. The purpose of this study, therefore, was to compare CT window settings selected by radiologists with window settings determined by using two alternative approaches for depiction of PE.
|
MATERIALS AND METHODS
|
|---|
Institutional review board approval was obtained. Informed consent was not required because our study involved retrospective review of previously obtained images. The study was compliant with the Health Insurance Portability and Accountability Act.
Subject Selection
The logbook for a 16-detector row CT scanner (Sensation 16; Siemens Medical System, Iselin, NJ) was retrospectively reviewed (G.N.M.) for the months of June and July 2003 to obtain a list of patients who underwent chest CT to enable detection of PE. We reviewed the diagnostic reports and CT images of the patients in the logbook and selected a convenience sample (6) of 13 patients with PE and 12 patients without PE (18 women and seven men; age range, 21–86 years; mean age, 58 years). Consideration was given to the inclusion of subjects with CT images that had a wide range of PA contrast enhancement, which would allow us to evaluate the effect of contrast enhancement on the selection of window settings.
CT Scanning
Our standardized protocol consisted of 120-kVp tube voltage, 120-mAs (effective) tube current–time, 16 x 0.75-mm detector configuration, and 1-mm reconstructed section thickness. Images were reconstructed with a standard soft-tissue kernel algorithm. Contrast enhancement was achieved by injecting 120 mL of ioversol (350 mg of iodine per milliliter, Optiray 350; Tyco Health/Mallinckrodt, St Louis, Mo) with a power injector at a rate of 3.0–4.0 mL/sec. The scanning delay was determined by using a bolus tracking method, in which a region of interest was placed on the main PA, and diagnostic CT scanning was triggered at the contrast enhancement threshold of 100 HU.
Image Evaluation
CT images were retrieved from the institutional picture archiving and communication system and transferred to a clinical workstation (Siemens). Prior to the start of the image evaluation session, CT images were displayed, and 1-cm circular regions of interest were placed over the main PAs on CT images (G.N.M.). The means and standard deviations of the attenuations of pixels within the regions of interest were measured and recorded.
For the image evaluation session, we used a clinical workstation with two identical screens displayed side by side. Five chest radiologists (K.T.B., S.B., D.S.G., F.R.G., P.K.W.) and two body radiologists (D.M.B., C.O.M.) independently participated in each image evaluation session. These radiologists had 5–25 years (mean, 12.9 years) of experience in the interpretation of body CT images, and they had been interpreting CT images for PE since CT became a widely accepted clinical application for PE (ie, approximately 6 years).
At the evaluation session, each of the 25 CT studies was randomly loaded and reviewed. Images were displayed side by side on two viewing screens. Initially, images were displayed with a standard mediastinal window setting (window width, 400 HU; window center, 30 HU). On the second viewing screen, images were displayed with the lung window setting (window width, 1500 HU; window center, –800 HU). For PE detection, each radiologist adjusted the mediastinal window to a personally preferred setting (eg, "personal") while inspecting the contrast enhancement of the PA, typically at the level of the main PA, before reviewing the entire image set. The window width and window center values of this setting were subsequently recorded (G.N.M.). Six of the seven radiologists adjusted the window settings, solely relying on their visual assessments of image quality and without measuring PA attenuation. One observer had a personal approach of measuring the mean PA attenuation and setting the window center at half of the mean PA attenuation. This observer, however, visually adjusted the window width and did not use the mean PA attenuation.
The image on the second screen was adjusted (G.N.M.) to the "modified" window setting. Window width was calculated with the following equation: MPA + 2 · SDPA, where MPA is the mean attenuation of the PAs, and SDPA is the standard deviation, as proposed by Brink et al (5). Window center was calculated by dividing window width by two. The observers scrolled through the images on both screens and compared image quality. Although our study was not designed as a test for PE diagnosis, the observers were instructed to scroll through the entire set of images as if they were in a clinical setting and were assessing PE. The observers were blinded to the diagnosis report of each case. The image on the second screen was then adjusted (G.N.M.) with the "double-half" window setting that we devised, and it was reviewed and compared with the image on the first screen. The double-half window width was calculated by multiplying the mean PA attenuation by two, and window center was calculated by dividing the mean PA attenuation by two. This window setting was formulated because it provides a wider window width than the modified setting and is easy to calculate. The observers were allowed to switch back and forth between different window settings for evaluation and comparison.
For each case, the observers were asked to rate window settings (ie, personal, modified, and double-half settings) from 1 to 3 (ie, from most to least preferred, respectively). When no difference in preference was perceived between the three settings, the settings were rated the same (ie, 1, 1, 1). When no difference in preference was perceived between two settings, the three settings were rated either as 1, 1, and 2 if the third setting was preferred less or as 2, 2, and 1 if the third setting was preferred more.
Data and Statistical Analysis
For rating images on the basis of the three types of window settings, mean rating scores for the seven observers were calculated. Distributions of these mean rating scores were tested for normality with the Shapiro-Wilk W test. Differences among the mean scores for the seven observers were tested with repeated-measures analysis of variance followed by paired t tests.
To address the trend of the personal window settings in relation to contrast enhancement of PAs, we divided the personal window center and window width values by the corresponding mean PA attenuations to generate normalized window center and window width ratios. We also calculated the window center and window width ratios for the modified and double-half window settings. For the double-half window setting, the window center ratio was 0.5, and the window width ratio was 2.0 for all cases. Distributions of window center and window width ratios for the seven observers were tested for normality with the Shapiro-Wilk W test. For observers whose personal window center and window width settings were not normally distributed, the differences among observers were tested for statistical significance with the Wilcoxon signed-rank t test. For the 25 cases, the means of the window center and window width ratios for the seven observers were calculated and tested for significant differences with the Wilcoxon signed-rank t test.
This study was exploratory in nature. Because no pilot data existed, a power analysis was not performed. We reasoned, however, that 25 subjects (13 with findings that were positive for PE and 12 with findings that were negative for PE) would be sufficient to determine if a larger prospective study would be warranted.
Regression analysis was used to examine the relationship between mean attenuation of the PAs and mean window center and window width ratios. An 
value of less than .05 was considered to indicate statistical significance. All statistical analysis was performed with JMP Statistical Software (version 5; SAS Institute, Cary, NC).
|
RESULTS
|
|---|
Display window settings affected the conspicuity of PE (Figs 1, 2). A subtle nonocclusive embolus in the right upper lobe PA in Figure 1 is not visible at the standard mediastinal window setting, but it is apparent at other window settings. A focal embolus in the right lower lobe PA in Figure 2 is well demarcated from the enhanced blood seen at all four window settings. Both the embolus and the lung parenchyma, however, were equally dark and were not separable at the modified window setting, thereby potentially influencing the conspicuity and detection of PE.
View larger version (85K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a. Patient 3. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 416 HU; window center, 208 HU), and (c) double-half (window width, 748 HU; window center, 187 HU) window settings. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 628 HU; window center, 175 HU). A subtle nonocclusive PE (arrow) in the right upper lobe PA is obscured by apparent intensely enhanced blood, and it becomes invisible at the standard mediastinal window setting. PE is apparent at other window settings.
|
|
View larger version (49K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b. Patient 3. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 416 HU; window center, 208 HU), and (c) double-half (window width, 748 HU; window center, 187 HU) window settings. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 628 HU; window center, 175 HU). A subtle nonocclusive PE (arrow) in the right upper lobe PA is obscured by apparent intensely enhanced blood, and it becomes invisible at the standard mediastinal window setting. PE is apparent at other window settings.
|
|
View larger version (95K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1c. Patient 3. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 416 HU; window center, 208 HU), and (c) double-half (window width, 748 HU; window center, 187 HU) window settings. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 628 HU; window center, 175 HU). A subtle nonocclusive PE (arrow) in the right upper lobe PA is obscured by apparent intensely enhanced blood, and it becomes invisible at the standard mediastinal window setting. PE is apparent at other window settings.
|
|
View larger version (91K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1d. Patient 3. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 416 HU; window center, 208 HU), and (c) double-half (window width, 748 HU; window center, 187 HU) window settings. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 628 HU; window center, 175 HU). A subtle nonocclusive PE (arrow) in the right upper lobe PA is obscured by apparent intensely enhanced blood, and it becomes invisible at the standard mediastinal window setting. PE is apparent at other window settings.
|
|
View larger version (119K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a. Patient 23. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 454 HU; window center, 227 HU), and (c) double-half (window width, 820 HU; window center, 205 HU) window setting. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 688 HU; window center, 192 HU). A focal embolus (arrowhead) in the right lower lobe PA is well demarcated from the enhanced blood at all four window settings; however, both the embolus and the lung parenchyma were equally dark and were not separable at the modified window setting, thereby potentially affecting the conspicuity and detection of PE. The pulmonary valve (arrow) is obscured and invisible at the mediastinal window setting, but it is well depicted at the other window settings.
|
|
View larger version (79K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b. Patient 23. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 454 HU; window center, 227 HU), and (c) double-half (window width, 820 HU; window center, 205 HU) window setting. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 688 HU; window center, 192 HU). A focal embolus (arrowhead) in the right lower lobe PA is well demarcated from the enhanced blood at all four window settings; however, both the embolus and the lung parenchyma were equally dark and were not separable at the modified window setting, thereby potentially affecting the conspicuity and detection of PE. The pulmonary valve (arrow) is obscured and invisible at the mediastinal window setting, but it is well depicted at the other window settings.
|
|
View larger version (132K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2c. Patient 23. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 454 HU; window center, 227 HU), and (c) double-half (window width, 820 HU; window center, 205 HU) window setting. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 688 HU; window center, 192 HU). A focal embolus (arrowhead) in the right lower lobe PA is well demarcated from the enhanced blood at all four window settings; however, both the embolus and the lung parenchyma were equally dark and were not separable at the modified window setting, thereby potentially affecting the conspicuity and detection of PE. The pulmonary valve (arrow) is obscured and invisible at the mediastinal window setting, but it is well depicted at the other window settings.
|
|
View larger version (135K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2d. Patient 23. Transverse CT images obtained with the (a) standard mediastinal (window width, 400 HU; window center, 30 HU), (b) modified (window width, 454 HU; window center, 227 HU), and (c) double-half (window width, 820 HU; window center, 205 HU) window setting. (d) Transverse CT image obtained with the mean value of the personal window setting for all observers (window width, 688 HU; window center, 192 HU). A focal embolus (arrowhead) in the right lower lobe PA is well demarcated from the enhanced blood at all four window settings; however, both the embolus and the lung parenchyma were equally dark and were not separable at the modified window setting, thereby potentially affecting the conspicuity and detection of PE. The pulmonary valve (arrow) is obscured and invisible at the mediastinal window setting, but it is well depicted at the other window settings.
|
|
Contrast Enhancement and Preferred Settings
The preference ratings of the seven observers for the three window settings in all 25 studies are listed in the Table. A wide range of contrast enhancement values was observed, with mean attenuation of the main PA ranging from 234 to 562 HU. Two observers had variances of zero (ie, the same rating for all 25 cases) for a particular type of window setting. Shapiro-Wilk W tests indicated that the rating data were nonnormally distributed (P < .05). Differences among the mean rating scores for the seven observers, however, were normally distributed. For the three types of window settings (Fig 3), repeated-measures analysis of variance demonstrated a significant difference among the mean rating scores of the seven observers (P < .001). Paired t tests demonstrated that the ratings for all three types of window settings were significantly different from one another, including the difference between double-half and modified window settings (P < .001). The observers preferred their personal window settings the most and the modified window settings the least. Specifically, six of seven observers rated their personal window settings as the best setting. Four of these observers rated double-half window settings as the second best setting, and the remaining two observers rated modified window settings as the second best setting. One observer rated the modified window setting as the best setting and the personal window setting as the second best setting.
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. Graph shows mean ratings of the seven observers for the 25 patients. A score of 1 indicates the most preferred window setting, and a score of 3 indicates the least preferred window setting. To better illustrate the distribution of data points, they are spread horizontally to minimize overlapping. The horizontal lines across the diamonds represent the mean value of the group, and the vertical spans of the diamonds represent 95% confidence intervals. If the horizontal lines within the diamonds drawn above and below the mean lines overlap, group means are not significantly different at the 95% confidence interval. The horizontal spans of the diamonds are proportional to the sample sizes of the groups.
|
|
Window Width and Center Ratios
Figure 4 shows the window width and center ratios for the double-half and modified window settings and for the seven observers on the basis of their personal settings. Window width ratios for observer 1 versus observer 3 and for observer 2 versus observer 4 were not significantly different from each other (P 
.24), nor were differences between observers 4, 6, and 7 (P .17). All other differences in width ratios between observers were significant (P < .05). The mean window width ratio for the seven observers was 1.68 ± 0.20. The mean window width ratio for the double-half setting was 2.0 ± 0.0, and the mean window width ratio for the modified setting was 1.14 ± 0.05. Because all of these ratios were nonnormally distributed (P < .05, Shapiro-Wilk W test), the differences between ratios were tested with the Wilcoxon signed-rank test. Window width ratios for all methods were significantly different from each other (P < .001). On average, ratios for modified settings were 0.54 ± 0.17 lower than ratios for personal settings, and ratios for double-half settings were 0.32 ± 0.20 higher. These are differences of 32% and 19%, respectively. In terms of absolute values, the differences are significantly higher for modified ratios than for double-half ratios (P = .002).
View larger version (16K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a. Box plots of window (a) width and (b) center ratios for the double-half and modified window settings and for the seven observers on the basis of their personal settings. To better illustrate the distribution of the data points, they are spread horizontally to minimize overlapping. The ends of the boxes are the 25th and 75th quartiles. The lines across the middle of the boxes indicate the median values. The interquartile range is the difference between the quartiles. The lines extend from the boxes to the outermost points that fall within the interquartile range.
|
|
View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b. Box plots of window (a) width and (b) center ratios for the double-half and modified window settings and for the seven observers on the basis of their personal settings. To better illustrate the distribution of the data points, they are spread horizontally to minimize overlapping. The ends of the boxes are the 25th and 75th quartiles. The lines across the middle of the boxes indicate the median values. The interquartile range is the difference between the quartiles. The lines extend from the boxes to the outermost points that fall within the interquartile range.
|
|
Window center ratios for observers 1 and 3 were not significantly different from one another (P = .669), nor were center ratios for observers 4, 5, 6, and 7 (P .435). All other differences between observers for window center ratios, however, were significant (P .003). The mean (± standard deviation) window center ratio for the seven observers was 0.47 ± 0.05. The mean window center ratio was 0.5 ± 0.0 for the double-half setting and 0.57 ± 0.02 for the modified setting. Although the window center ratios for the personal settings were normally distributed (P = .86, Shapiro-Wilk W test), the center ratios for the other two settings were not (P < .05). Differences between settings were, therefore, tested with Wilcoxon signed-rank tests. Center ratios for both settings were significantly higher than center ratios for personal settings (P < .001), and ratios for the modified setting were significantly higher than those for the double-half setting (P = .013). The mean difference between ratios for observers was 0.10 ± 0.04 for the modified settings and 0.03 ± 0.05 for the double-half setting. These represent differences of 21% and 6%, respectively.
Regression analysis of window width and window center ratios against mean attenuation of the PAs was performed for each observer. We used the polynomial model to determine if it would improve the fit over that of a straight-line model. A second-order polynomial, which is the simplest extension of the straight-line model, improved the fits; however, we could think of no biological reason for the inflections (ie, slight increases in window width and center ratios) that generally occurred in the regression curves with a higher mean attenuation of the PA values. Thus, we present the r2 and P values that resulted from our use of the straight-line model. All regression curves had negative slopes. Regression analysis of window width ratios on mean attenuation of the PA values resulted in the following values: observer 1, r2 value of 0.29 and P value of .006; observer 2, r2 value of 0.59 and P value less than .001; observer 3, r2 value of 0.18 and P value of .033; observer 4, r2 value of 0.23 and P value of .016; observer 5, r2 value of 0.51 and P value less than .001; observer 6, r2 value of 0.19 and P value of .028; and observer 7, r2 value of 0.05 and P value of .266. For the six of seven (86%) significant correlations, the mean r2 value was 0.29 ± 0.19. Regression analysis of window center ratios on mean attenuation of the PA values resulted in the following values: observer 1, r2 value of 0.02 and P value of .456; observer 2, r2 value of 0.40 and P value less than .001; observer 3, r2 value of 0.00 and P value of .865; observer 4, r2 value of 0.20 and P value of .023; observer 5, r2 value of 0.21 and P value less than .021; observer 6, r2 value of 0.24 and P value of .013; and observer 7, r2 value of 0.06 and P value of .233. For the four of seven (57%) significant correlations, the mean r2 value was 0.16 ± 0.14. These results indicate that observers tended to set window widths and centers higher when contrast enhancements were lower and vice versa. Thus, with increased enhancement of the PA, the window width and center were set higher, but the proportionality of the increases were less. To illustrate this point, Figure 5 contains the window width and center regressions that had the highest r2 values.
View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5a. Graphs show regression curves of the window (a) width and (b) center ratios for mean PA attenuation values (measured in Hounsfield units), as measured by observer 2. Of the regression curves of the seven observers, those of observer 2 had the highest r2 values for width ratios (r2 = 0.59, P < .001) and center ratios (r2 = 0.40, P < .001). Each point corresponds to one of the 25 cases. The 95% confidence intervals are included in each plot.
|
|
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5b. Graphs show regression curves of the window (a) width and (b) center ratios for mean PA attenuation values (measured in Hounsfield units), as measured by observer 2. Of the regression curves of the seven observers, those of observer 2 had the highest r2 values for width ratios (r2 = 0.59, P < .001) and center ratios (r2 = 0.40, P < .001). Each point corresponds to one of the 25 cases. The 95% confidence intervals are included in each plot.
|
|
|
DISCUSSION
|
|---|
The mediastinal window setting (window width, 400 HU; window center, 30 HU), which is traditionally used to enable visualization of soft-tissue anatomy on chest CT images, may not be adequate in the detection of PE. It may be too narrow and centered too low to appro-priately cover the range of contrast medium–enhanced PA attenuation values, which are usually higher than the soft-tissue attenuation values. The use of window settings that have higher window width and window center values than the mediastinal window would improve the ability of radiologists to visually distinguish PE from the background enhanced blood, as demonstrated by Brink et al (5). They proposed a modified window setting that is based on measuring the mean and standard deviation of the main PA attenuation and setting the window width equal to the mean attenuation plus 2 standard deviations and the window center equal to half the width. This setting tends to produce a narrow window width; therefore, to our knowledge, it is not commonly used in clinical practice. As a result, Brink et al (5) recommended that CT images be displayed and reviewed separately at the modified and standard mediastinal window settings for detection of PE. The modified window setting would be useful in delineating PE as filling defects, while the standard mediastinal window setting would be useful in identifying PA branches that are occluded with PE. Without the mediastinal window setting, the PE occluding the PA branches may not be differentiated from the pulmonary parenchyma at the modified window setting.
Despite a plethora of studies and reports on CT imaging of PE, few guidelines are available for setting a clinical display window for detection of PE. We recognize that there must be more than one way to adjust the display window setting for detection of PE; however, a general clinical guideline for setting display windows will help trainees and less experienced radiologists as they learn to display CT images appropriately and before they develop the interpretation skills required for detection of PE. To create a display window setting that would be optimal for detection of PE, we evaluated the feasibility of mathematically calculating window settings. We quickly realized that determining optimization criteria was not an easy task and that it was highly variable among experienced radiologists. Without a testable theoretical or mathematical formula, we decided to take an empirical approach to investigate how CT radiologists adjust the display window setting. We postulated that this empirical approach would provide insight for determining optimal window settings and might help improve the diagnostic skills of radiology trainees and radiologists for PE detection.
Chest CT images are usually reviewed at both the mediastinal and the lung window setting. These standard CT window settings have been modified to improve the appearance of abnormalities of various organs, including the liver (7–9), cerebral artery (10), and PA (1,4,5). Although liver windows are used at relatively fixed values (window width, 150 HU; window center, 50–100 HU) (7), an adjustable window setting would be more appropriate in the detection of PE because the contrast enhancement of the PA is far more variable and wider than that of the hepatic artery. In this regard, we agree with Brink et al (5) that window settings appropriate for detection of PE should be adjusted with respect to the degree of contrast enhancement of the pulmonary vessels.
We proposed a new method (ie, the double-half window setting) for setting windows. With this method, the window width is set at two times the mean PA attenuation, and the window center is set at half the mean PA attenuation. The rationale for this approach is that the double-half window setting provides a wider window width than the modified window setting proposed by Brink et al (5), and it is easy to calculate. Our data indicate that the personal window settings chosen by the observers were closer to the double-half window settings than to the modified window settings. This result is also supported by observer comments that the modified window settings displayed the images with too much tissue contrast.
Some radiologists may prefer to adjust the display window settings by relying on visual qualitative assessments and without measuring the attenuation of the main PAs. In our study, observers noted that the pulmonary valve was frequently obscured by intensely enhanced blood and that it was not visible at the mediastinal window setting. The pulmonary valve was consistently visible at the other three window settings we tested. We think that because the valve is a thin low-attenuation structure that is immersed in the enhanced pulmonary arterial blood, a window setting that facilitates the visualization of the pulmonary valve should be adequate to display PE. Thus, if the pulmonary valve is visible, it can be used as an index to adjust the window setting from the default mediastinal window setting. This approach, however, was not systematically evaluated in the current study. Furthermore, the pulmonary valve may not always be demarcated because of cardiac motion at non-electrocardiographically gated scanning.
Our results demonstrated significant correlations between the degree of attenuation in the main PAs and the window centers and widths set by four of seven (57%) observers for window centers and six of seven (86%) observers for window widths. When CT attenuation of the main PA is low, window width tends to be set high; however, when CT attenuation is high, window width tends to be set low. This finding is in agreement with that of Brink et al (5), where the widths of the modified window settings were set lower than two times the PA attenuation. This was because contrast enhancement of PAs in a porcine model was usually at the high end (>450 HU) of the spectrum of enhancement encountered in routine clinical examinations.
There are several limitations to our study. First, the most critical limitation is that this study was not designed to test the efficacy of each window setting in demonstrating the presence or absence of PE for each reader. The effect of the different window settings on the sensitivity and specificity for diagnosis of PE was not assessed. During the image evaluation session, observers were instructed to select their preferred window settings, just as they would do clinically. A prospective outcome study is required to address and resolve this limitation.
Second, the number of observers is small, and all seven observers were from the same institution. This institution was where the observers started reading chest CT images for detection of PE. Our results may, therefore, simply reflect an institutional trend or bias. Personal preference may vary due to numerous variables (ie, level and/or conditions of training, years of experience, or interpretation environment), thus requiring a larger number of observers who represent the spectrum of variables. There has, however, been no cross-training imaging session for interpretation of chest CT images for PE detection, and observers were not aware of the window setting approaches of other observers.
Third, observers were aware that the purpose of this study was to evaluate their display window preferences; therefore, they may have been consciously setting consistent display preferences. We requested, however, that (if possible) observers select the display windows they preferred for PE viewing, without interpreting their chosen width and center values. Through these efforts, we attempted to create a simulation of the clinical setting for CT image interpretation for PE detection.
Fourth, the display window choices were not distinguished with regard to whether PE was located centrally or peripherally in the PA. The enhancement of PAs may differ substantially between the central and peripheral arteries. The location of the PE relative to the lung background and the vessel wall may affect the adequacy of the display window setting. In practice, some radiologists may use more than one display window setting to detect PE, or they may change the setting as they scroll through the images. Images displayed at the standard mediastinal or lung window settings may provide additional cues. This multisetting approach was not addressed in this study.
Fifth, no follow-up or further clinical evaluation was performed to confirm the diagnosis of PE given in the original report of the cases.
In conclusion, display window widths and centers used by the observers in the detection of PE were consistently greater than those of conventional mediastinal window settings. With the mean PA attenuation serving as a reference index, on average the observers selected CT window settings for PE detection with a window width of slightly less than twice the mean PA attenuation and a window center of about half the mean PA attenuation; furthermore, they tended to use higher window widths and centers with lower degrees of PA enhancement. A reader who desires to adopt this trend may measure the attenuation in the main PA on CT images. The display window width is equal to 1.6–2.0 times the mean attenuation, and the display window center is equal to 0.4–0.6 times the mean attenuation. When the mean attenuation is less than 300 HU, higher proportionality is used.
|
FOOTNOTES
|
|---|
Abbreviations: PA = pulmonary artery • PE = pulmonary emboli
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, K.T.B.; study concepts, K.T.B., G.N.M.; study design, K.T.B., G.N.M., C.F.H.; literature research, K.T.B., G.N.M., J.M.G.; clinical studies, K.T.B., D.M.B., S.B., D.S.G., F.R.G., C.O.M., P.K.W.; data acquisition, K.T.B., G.N.M.; data analysis/interpretation, K.T.B., G.N.M., C.F.H.; statistical analysis, K.T.B., C.F.H.; manuscript preparation, K.T.B., G.N.M.; manuscript definition of intellectual content, all authors; manuscript editing, K.T.B., G.N.M., J.M.G., C.F.H.; manuscript revision/review and final version approval, all authors
|
References
|
|---|
- Raptopoulos V, Boiselle PM. Multi–detector row spiral CT pulmonary angiography: comparison with single–detector row spiral CT. Radiology 2001; 221:606–613.[Abstract/Free Full Text]
- Schoepf UJ, Holzknecht N, Helmberger TK, et al. Subsegmental pulmonary emboli: improved detection with thin-collimation multi–detector row spiral CT. Radiology 2002; 222:483–490.[Abstract/Free Full Text]
- Schoepf UJ, Costello P. CT angiography for diagnosis of pulmonary embolism: state of the art. Radiology 2004; 230:329–337.[Abstract/Free Full Text]
- Coche E, Verschuren F, Keyeux A, et al. Diagnosis of acute pulmonary embolism in outpatients: comparison of thin-collimation multi–detector row spiral CT and planar ventilation-perfusion scintigraphy. Radiology 2003; 229:757–765.[Abstract/Free Full Text]
- Brink JA, Woodard PK, Horesh L, et al. Depiction of pulmonary emboli with spiral CT: optimization of display window settings in a porcine model. Radiology 1997; 204:703–708.[Abstract/Free Full Text]
- Obuchowski NA. Special topics III: bias. Radiology 2003; 229:617–621.[Abstract/Free Full Text]
- Mayo-Smith WW, Gupta H, Ridlen MS, Brody JM, Clements NC, Cronan JJ. Detecting hepatic lesions: the added utility of CT liver window settings. Radiology 1999; 210:601–604.[Abstract/Free Full Text]
- Brink JA. Liver window settings at hepatic CT: added value or marginal cost? Radiology 1999; 210:593–594.[Free Full Text]
- Patten RM, Gunberg SR, Brandenburger DK, Richardson ML. CT detection of hepatic and splenic injuries: usefulness of liver window settings. AJR Am J Roentgenol 2000; 175:1107–1110.[Abstract/Free Full Text]
- Gadda D, Vannucchi L, Niccolai F, Neri AT, Carmignani L, Pacini P. CT in acute stroke: improved detection of dense intracranial arteries by varying window parameters and performing a thin-slice helical scan. Neuroradiology 2002; 44:900–906.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
C. Wittram
How I Do It: CT Pulmonary Angiography
Am. J. Roentgenol.,
May 1, 2007;
188(5):
1255 - 1261.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
