| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Thoracic Imaging |
1 From the Department of Radiology, University of Vienna Medical School, Allgemeines Krankenhaus Wien, Waehringer Guertel 18–20, A-1090 Vienna, Austria (M.U., N.K., M.W., C.J.H., C.S.P.); Philips Medical Systems, Hamburg, Germany (U.N.); and Department of Radiology, University Medical Center Utrecht, the Netherlands (M.P.). From the 2002 RSNA Annual Meeting. Received October 26, 2003; revision requested January 13, 2004; final revision received June 21; accepted July 2. Address correspondence to M.U. (e-mail: martin.uffmann@meduniwien.ac.at).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: The study protocol was approved by the institutional internal review board, and written informed consent was obtained from all patients. Posteroanterior chest radiographs of 48 consecutively selected patients were obtained at 90, 121, and 150 kVp by using a flat-panel–detector unit that was based on cesium iodide technology and automated exposure control. Monte Carlo simulations were used to verify that the effective dose for all kilovoltage settings was equal. Five radiologists subjectively and independently rated the delineation of anatomic structures on hard-copy images by using a five-point scale. They also ranked image quality in a blinded side-by-side comparison. Average ranking scores were compared by using one-way analysis of variance with repeated measures. Data were analyzed for the entire patient group and for two patient subgroups that were formed according to body mass index (BMI).
RESULTS: The visibility scores of most anatomic structures were significantly superior with the 90-kVp images (mean score, 3.11), followed by the 121-kVp (mean score, 2.95) and 150-kVp images (mean score, 2.80). Differences did not reach significance (P > .05) only for the delineation of the peripheral vessels, the heart contours, and the carina. This was also true for the subgroup of patients (n = 24) with a BMI greater than and the subgroup of patients (n = 24) with a BMI less than the mean BMI (26.9 kg/m2). At side-by-side comparison, the readers rated 90-kVp images as having superior image quality in the majority of image triplets; the percentage of 90-kVp images rated as "first choice" ranged from 60% (29 of 48 patients) to 90% (43 of 48 patients), with a median of 88% (42 of 48 patients), among the readers.
CONCLUSION: Delineation of most anatomic structures and overall image quality were ranked superior in digital radiographs acquired with lower kilovoltage at a constant effective patient dose.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
In digital chest radiography, examination parameters such as tube voltage and filtration have frequently been adopted from screen-film technology. Digital systems, however, are characterized by their ability to use image processing to optimize image contrast and density. Underexposure no longer influences image density but results only in increased image noise. The increased dynamic range of the detector ensures sufficient visualization of both the lungs and the mediastinum, even at low kilovoltage settings (4). An increased image latitude, as seen with low kilovoltage settings, can be reduced by proper processing (5). The high sensitivity of digital image detectors allows for sufficiently short exposure times, also at lower kilovoltages.
Thus, lowering the kilovoltage settings appears to be technically feasible with digital radiography systems. The motivations for reducing the kilovoltage settings for digital chest radiography include the fact that low kilovoltage settings improve structural contrast owing to a general increase in x-ray attenuation at low x-ray energies. Furthermore, the detective quantum efficiency (DQE) of digital detector media increases with lower voltage (6). This gain in DQE might be translated into improved image quality.
Moreover, the estimation of radiation burden to the patient should be based on effective dose rather than incident or skin exposure because effective dose describes best the probability of stochastic radiation damage, which is the only type of radiation effect of importance at dose levels typical for chest radiography (7). Because the relationship between skin exposure and effective dose changes with radiation quality (ie, with kilovoltage), an accurate estimation of effective dose is necessary for each kilovoltage considered. The purpose of our study, therefore, was to compare the visibility of anatomic structures in direct-detector chest radiographs acquired with different tube voltages at equal effective doses to the patient.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Image Acquisition
Regular lateral and posteroanterior chest radiographs were obtained in all patients at 125 and 121 kVp, respectively. These images served for diagnostic purposes and were clinically indicated. In addition, we obtained posteroanterior radiographs at 90 and 150 kVp. Automatic exposure control adjusted to a nominal system sensitivity of 400 at 121 kVp was used for all radiographs. All exposures were performed by using an integrated 12:1 grid and a 1.8-m detector-to-tube distance. All radiographs were collimated to a constant area of 35 x 43 cm.
Detector system.—All images were obtained with a cesium iodide amorphous silicon flat-panel–detector unit (Pixium 4600; Trixell, Moirans, France) integrated into a wall stand (Digital Diagnost System; Philips, Hamburg, Germany). The size of the detector area of the flat-panel unit was 43 x 43 cm. The pixel size was 143 µm.
Exposure measurements and estimation of effective dose.— The system automatically recorded the following parameters for each image: tube voltage (in peak kilovoltage), exposure time (in milliseconds), tube current (in milliamperes), and kerma area product (KAP) (in decigray–square centimeters). The KAP indication on the system was calibrated before the study according to the manufacturer’s instructions.
For all three kilovoltage levels, we calculated the correlation between the tube current and the BMI. To assess the effect of tube voltage on the correlation between tube current and BMI, we normalized the milliampere-second settings of the 90- and 150-kVp images relative to the milliampere-second settings of the 121-kVp images (by dividing the milliampere-second value at 90 or 150 kVp by that at 121 kVp) and then calculated the correlation between the normalized data and BMI. Similarly, we assessed the correlation between BMI and the normalized KAP.
The effective dose to the patient was estimated by multiplying the KAP values by kilovoltage-specific conversion factors that were calculated by using the Monte Carlo simulation software PCXMC, which was developed by STUK, the radiation and nuclear safety authority in Finland (8). The conversion factors were determined for the specific exposure conditions applied in this study (focus detector distance, 180 cm; tube filtration, 2.5-mm aluminum; field size, 35 x 43 cm) and for the lowest, the mean, and the highest BMI values in our study group.
Image processing and display.—Image processing was kept identical for all images, independent of kilovoltage settings. Images were processed with the same algorithm that is routinely used in our institution for the processing of hard-copy films of the chest. This algorithm is based on a variant of unsharp mask filtering that is used with the flat-panel system. By using this processing technique, the lungs were adjusted to an optical density of 1.8. A gamma of 2.6, a detail contrast enhancement of 0.4, and a noise-reduction factor of 0.8 were used. All images were printed with a laser printer as 35 x 43-cm hard copies (Imation Dry View 8700; 3M, St Paul, Minn).
Reading Methods
For evaluation, hard-copy images were mounted on a view box (Rotolux; Schulte, Warstein, Germany) with a minimum brightness of 2000 candelas per square meter. The hard-copy images were masked to the edges to prevent glare. Reading conditions were kept constant, with the ambient room light subdued. To ensure adaptation of the eyes to the ambient lighting conditions, each reading session was started with an initial run of five trial images that were similar to the study images but were not part of the data analysis. To minimize reading-order bias, each reader saw the images in a different random order. The observers were free to vary the viewing distance. There were no time constraints during the reading sessions.
Five readers independently evaluated the images. Two readers (C.S. and M.U.) were board-certified radiologists with 8 and 12 years of experience in chest radiology, respectively. The three others were third-year radiology residents.
The readers read all posteroanterior views twice, with a time interval of at least 2 weeks between the readings. The first time, the images were presented separately and in a random order that was different for each reader. The readers were asked to subjectively assess the visibility of predefined anatomic structures by using a five-point scale in which a score of 5 corresponded to excellent visibility; a score of 4, good visibility; a score of 3, moderate visibility; a score of 2, poor visibility; and a score of 1, an unacceptable image. Ten anatomic structures were graded: (a) the lung parenchyma without rib superimposition, (b) the lung parenchyma with rib superimposition, (c) the perihilar vessels, (d) the peripheral vessels (within a 2-cm-wide subpleural space), (e) the costophrenic recess, (f) the cardiophrenic recess, (g) the retrocardiac area, (h) the carina, (i) the heart contours, and (j) the lower thoracic spine.
For the items listed, an overall assessment that included both sides of the body was performed. For anatomic structures within the lung, only regions of normal opacity were evaluated. In cases of unilateral pleural effusion, the pleural recesses were graded for the normal side only; in cases of bilateral effusion, these items were excluded from evaluation.
For the second reading, the three posteroanterior views of each patient were presented side-by-side as image "triplets." Both the order of the patients and the order of the films in each patient were varied randomly. Readers were unaware of the tube voltage used. The readers were then asked to rank images subjectively according to their preference in terms of overall image quality. No equivalent ranking was possible.
Statistical Analysis
Differences in age between the male and female patients were analyzed with the t test. We used a level of significance of P 
.05.
The correlation between tube current and BMI was performed for all three kilovoltage levels by using the Pearson correlation. Similarly, the exposure parameters milliampere-seconds and KAP normalized to the settings at 121 kVp were correlated with BMI by using the Pearson correlation. The KAP values for the 48 patients were analyzed by using analysis of variance with repeated measures.
The rating data of the five readers were averaged for each anatomic structure and image. To assess the effect of BMI, patients were sorted into one of two groups: A group with BMI values that were higher than the mean BMI, and a group with BMI values that were lower than the mean BMI. The statistical significance of differences between the three tube voltage settings with respect to the visibility of anatomic structures was assessed by using a one-way analysis of variance with repeated measures. The dependent variables in this test included the rating scores at the three different kilovoltage settings averaged across all five readers, and the independent variable was BMI. For post-hoc testing, critical differences were calculated to determine the differences among the three tube voltage groups. SPSS version 11.5 (SPSS, Chicago, Ill) was used for statistical analysis.
Power analysis was performed (nQuery Adviser, version 5.0; Statistical Solutions, Cork, Ireland) to compare the image quality scores with respect to the different acquisition techniques. The 
error value (the probability of a type I error) was set at .05, and both the ß error value (the probability of falsely accepting the null hypothesis) and the power (ie, 1 – ß) were calculated.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Exposure Parameters and Dose Values
The mean values of the exposure parameters, the KAP, and the effective dose for the images obtained at the three tube voltages are given in Table 1. As compared with the value at 121 kVp, the mean milliampere-seconds value decreased by a factor of 0.57 at 150 kVp and increased by a factor of 2.32 at 90 kVp. Compared with that at 121 kVp, the mean KAP decreased by a factor of 0.82 at 150 kVp, while it increased by a factor of 1.35 at 90 kVp. The mean exposure time was 5.68 msec at 150 kVp, 7.75 msec at 121 kVp, and 13.10 msec at 90 kVp. For all patients except three, the exposure time at 90 kVp was less than 20 msec, which is the maximum value recommended in the European Guidelines on Quality Criteria for Diagnostic Radiologic Images (2). For the three patients for whom the exposure time exceeded this limit, the exposure times were 24, 25, and 26 msec. These patients had high BMI values of 36, 34, and 33, respectively.
|
|
|
|
|
|
|
For a defined tube voltage, the conversion factor was found to be dependent on BMI, and it generally decreased with increasing BMI (Table 2). For each BMI value, however, the ratios of the conversion factors (normalized to the value at 121 kVp) remained constant within ±3% for the different kilovoltage levels. Both the independence of the relative KAP from BMI (Fig 2b) and the constant relative conversion factor led us to the conclusion that the effective dose remains approximately constant for each BMI level when the tube voltage is changed from 121 to 90 or 150 kVp. This is shown in Figure 2c, which illustrates the fact that the effective dose values relative to that with the standard acquisition technique (121 kVp) were not dependent on BMI. In addition, the relative effective dose values did not differ substantially between the different kilovoltage settings (Fig 2c).
|
Rating Scores
Visibility of anatomic structures.—The visibility scores of most anatomic structures, as averaged across the five readers, were significantly higher with the 90-kVp images (mean score, 3.11), followed by the 121-kVp (mean score, 2.95) and 150-kVp images (mean score, 2.80) (Fig 3). The differences in rating scores did not reach statistical significance only for the peripheral vessels, the cardiac contours, and the carina. This was true for the entire patient group, as well as for the subgroup of patients with a BMI greater than and the subgroup of patients with a BMI less than the mean BMI. For all three kilovoltage settings, rating scores were higher for the low-BMI group than for the high-BMI group (Fig 4, Table 3).
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Previous investigators have suggested an increased detail detectability (10) or an improved signal-to-noise ratio (11) with digital detector systems when images were acquired with lower voltage. To date, however, most data have been obtained from phantom studies (6,12–14). There are only a few reports available that describe experience obtained in a clinical setting (6,15). Most studies were performed at a fixed entrance surface dose; only Chotas et al (11) and Tingberg and Sjöström (14) made reference to effective dose, which is a more suitable indicator of radiation risk.
In the present study, we observed an improved grading of image quality when images were acquired with 90 kVp. The differences in the grading of image quality remained statistically significant between the three kilovoltage settings, even in patients with an above-average BMI.
Both the depiction of lung areas that were obscured and the depiction of those that were not obscured by overlying bone structures were rated superiorly at 90 kVp. On the basis of physical properties, lowering the kilovoltage increases absorption by the ribs, and one would suspect that the underlying lung parenchyma would be obscured as a consequence. This effect appears to be much less apparent in digital radiography than in conventional screen-film radiography. In fact, in 1996, Oda et al (12) reported that even though the density difference (contrast) between lung parenchyma and ribs increases on digital radiographs obtained with a low-kilovoltage technique (80 vs 140 kVp), the increase in density difference was less than 50% compared with the difference observed between 80 and 140 kVp at screen-film radiography. This is most likely explained by processing effects. Data processing in digital chest radiography is chosen to achieve a compression of the dynamic range, which results in a global decrease in large density differences. This has also been described as "image harmonization" in the literature (5).
Lowering the tube voltage necessarily leads to an increase in exposure time. At 90 kVp, exposure times in our study group were within the limit of 20 msec, as recommended in the European Guidelines on Quality Criteria for Diagnostic Radiographic Images (2), for 45 of 48 patients. Negative effects of prolonged exposure time, such as blurring of the heart border or the vessels, were not observed.
The depiction of the lower thoracic spine was also rated superiorly on the 90-kVp images. This is remarkable because the lower thoracic spine represents a structure located in the most dense part of the chest that is usually characterized by relatively low contrast and is frequently insufficiently visualized on conventional chest radiographs that are obtained at low kilovoltage settings. The superior delineation of structures on the digital radiographs, even in high-attenuation areas, is most likely due to the effect of image processing that is targeted at structural enhancement. Dobbins et al (16) suggested that the scatter within the mediastinum decreases with decreasing tube voltage, which might also be advantageous.
Several other authors have investigated the effect of tube voltage on image quality. Various study setups were used, and, most importantly, different detector types were evaluated. In a phantom study in which the acquisition of anteroposterior chest radiographs with storage-phosphor plates (bedside radiographs) was simulated, Chotas et al (11) observed higher signal-to-noise values with lower kilovoltage settings at matched effective dose equivalents. The differences in image quality were relatively small. However, it could be expected that this effect observed for anteroposterior radiographs would be even larger for posteroanterior radiographs owing to the fact that Chotas et al chose to reduce the skin entrance exposure for anteroposterior radiographs more than would be necessary for posteroanterior radiographs to keep the effective dose constant.
Tingberg and Sjöström (14) investigated the effect of tube voltage on the quality of anthropomorphic phantom images of the chest and pelvis, also by using a computed radiography system. On the basis of visual grading of anatomic structures, they observed a superior quality for both chest and skeletal applications of the images obtained with reduced tube voltage (70 kVp for radiographs of the chest) at a constant effective dose level.
In two studies, the effect of kilovoltage on images obtained with a selenium drum (Thoravision; Philips, Hamburg, Germany) was evaluated. One was a preference study whose approach was similar to ours that involved comparison of 90- and 150-kVp images (15). The results of that study also suggested a significant preference for the low-kilovoltage images.
Launders et al (6) investigated both physical parameters and subjective preference for clinical images, also by using the selenium drum detector. DQE values markedly increased with decreasing tube voltage owing to the increased absorption by the selenium layer. In the clinical part of the study, a preference for images acquired with 90 kVp was found for three of six anatomic regions; this was interpreted as an advantage in terms of perceived image quality over the use of the standard technique (150 kVp).
However, the situation for selenium-based and cesium iodide–based detectors may be different because the kilovoltage dependence of the x-ray absorption efficiency is more pronounced with selenium than with cesium iodide (17). In three relatively recent studies, the effect of kilovoltage on image quality at cesium iodide–detector radiography was evaluated. Rong et al (13) conducted a contrast-detail phantom study and reported a generally superior low-contrast performance of the cesium iodide flat-panel detector compared with the performance of screen-film or computed radiography systems. However, when they compared the performance with 80-, 100-, and 125-kVp images obtained with automated exposure control (similar detector exposure), they could not find significant effects of the kilovoltage setting on image quality. The authors explained this finding by referring to the fact that the DQE values of the cesium iodide detector system did not change markedly enough within the kilovoltage range tested to have an effect on image quality. However, the geometry of the phantom (CDRAD 2.0; Instrumentale Dienst, Nijmegen, the Netherlands) used in that study did not reflect the requirements of a human chest, which has various high- and low-absorption structures.
Bernhardt et al (4) evaluated the detectability of simulated interstitial lung lesions in a phantom study at three different tube voltages. They reported an equivalent performance with all images—irrespective of kilovoltage—except for the detectability of a reticular pattern and nodules over lucent lung, which were better seen with lower kilovoltage. However, it has to be noted that these images had been obtained with a lower effective dose compared with the standard 125-kVp images.
In a detailed approach, Dobbins et al (16) optimized the x-ray spectrum for a cesium iodide silicon flat-panel detector by using a computer spectrum model and additional narrow beam experimental measurements. In contrast to the behavior predicted by the computer model, an unfavorable decrease in signal-to-noise ratio per exposure was observed with increasing tube voltage in the mediastinum (full-field data). The investigators regarded 120 kVp as the optimal compromise between signal-to-noise ratio per incident patient exposure (which improves with lower kilovoltage) on one hand and the ratio of tissue contrast to bone contrast (which increases—and therefore improves—with higher kilovoltage) on the other.
For practical reasons, Dobbins et al (16) had used patient entrance exposure as a surrogate measure for patient risk. However, calculation of the effective dose is regarded as a more appropriate indicator of the risk associated with different kilovoltage settings (7). Although entrance exposure inevitably increases with lower kilovoltage, the effective dose equivalent remains relatively constant (6,18). Thus, the investigators most likely overestimated the patient risk with the low kilovoltage settings or, vice versa, reduced the acquisition exposure too much, which would cause a reduction in signal-to-noise ratio and might result in an underestimation of the properties of the low-kilovoltage technique.
Our study design bears the following limitations: We did not investigate the effect of an additional prefilter, which is sometimes used in chest radiography. Previous studies revealed that use of additional 0.2-mm copper filtration further decreased incident exposure to the patient without substantially altering image quality (16). It is not clear whether this still holds true if incident exposure is substituted for constant effective patient dose. The main disadvantage of using additional copper filtration, however, is the further increase in exposure time—most likely beyond the recommended 20-msec threshold.
We chose to summarize the subjective rating scores across all five readers to minimize the effect of reader variability. 
Statistics were not applicable owing to the asymmetric distribution of the rating scores. In addition, analysis would not have reflected the matching trends for slightly different levels of individual raters.
We focused on overall image quality and delineation of anatomic structures and did not specifically address the visibility of abnormal findings. The number of abnormal findings in our study group was too small, and most of the abnormal findings were relatively obvious and too diverse for further analysis. However, we would expect that similar results would be obtained for the delineation of abnormal findings as well, because we included contours in both high- and low-attenuation areas and specifically addressed low-contrast structures, such as the tracheal contour within the mediastinum and the peripheral vasculature.
One could argue that a similar effect of increased detail detectability could also be achieved with adequate processing. Although simple edge enhancement is already able to separately optimize image latitude and local structural contrast, it suffers from a nonselective enhancement of all structures—regardless of local contrast and size—and from an inevitably associated increase in image noise. Recently introduced elaborate multifrequency processing algorithms can variably adapt enhancement to structural features such as structural size, density, location, and contour definition. Only such elaborate processing may be able to achieve effects on image quality similar to those we observed with lower kilovoltage settings. While the increased structural contrast at low kilovoltage settings is an effect of increased DQE and increased absorption, frequency processing alters the contrast for the different frequency components of an image. It is likely that differences between these two approaches can be seen only for very subtle low-contrast structures that show only marginal absorption differences. Whether differences between the two approaches have a clinical impact has yet to be seen.
In conclusion, the delineation of anatomic structures and overall image quality are ranked superiorly if digital radiographs are acquired with lower kilovoltage at constant effective dose to the patient. Our results suggest that the current choice of high kilovoltage settings for posteroanterior digital chest radiographs should be reconsidered. Notably, with the system used in our study, posteroanterior chest radiographs may be obtained with lower kilovoltage settings by using automatic exposure control without substantially altering the effective patient dose.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, C.S.P.; study concepts, M.U., M.P., C.S.P.; study design, M.U., U.N.; literature research, N.K., M.U.; clinical studies, M.U., N.K.; data acquisition, N.K., M.U., C.S.P., U.N.; data analysis/interpretation, U.N., M.U., M.W., C.J.H.; statistical analysis, M.W.; manuscript preparation, M.U., C.S.P., U.N.; manuscript definition of intellectual content, M.P., U.N., M.U.; manuscript editing, C.S.P., M.U.; manuscript revision/review, M.P., N.K., M.W., C.J.H.; manuscript final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  L J M Kroft, W J H Veldkamp, B J A Mertens, J-P A van Delft, and J Geleijns Dose reduction in digital chest radiography and perceived image quality Br. J. Radiol., December 1, 2007; 80(960): 984 - 988. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
