| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Department of Clinical Radiology, University of Muenster, Albert-Schweitzer-Str 33, 48129 Muenster, Germany (T.M.B., U.R.B., H.L., S.D., K.L., W.H.); Department of Biometrics and Medical Informatics, Otto-von-Guericke University, Magdeburg, Germany (F.W.R.); and Department of Radiology and Neuroradiology, Klinikum Duisburg, Germany (K.P.). From the 2002 RSNA scientific assembly. Received February 4, 2003; revision requested April 23; final revision received January 14, 2004; accepted January 21. Address correspondence to T.M.B. (e-mail: bernhart@uni-muenster.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Simulated catheters, nodules, and ground-glass, linear, miliary, and reticular patterns were superimposed over an anthropomorphic chest phantom. Digital selenium radiography was performed with different tube voltages (70, 90, and 150 kVp). Hard-copy images were generated. Detection performance of five radiologists was compared by using receiver operating characteristic (ROC) analysis involving 54,000 observations.
RESULTS: The detection of ground-glass, linear, miliary, and reticular patterns over lucent lung and of nodules equal to, smaller than, and larger than 10 mm increased when 70 kVp and/or 90 kVp was used. However, only the reticular pattern was significantly better detected at lower peak voltage (P < .05). Simulated catheters and nodules over the mediastinum showed smaller areas under the ROC curve at lower peak voltage. These results were not statistically significant (P > .05).
CONCLUSION: The diagnostic performance of digital selenium radiography at lower peak voltage is at least as good as that at higher peak voltage for interstitial lung disease over lucent lung. Performance is equivalent for nodules and catheters over obscured chest regions at lower peak voltages compared with that at 150 kVp. Our results implicate that the use of high-voltage technique in digital selenium radiography should be reassessed.
© RSNA, 2004
Index terms: Experimental study • Phantoms • Radiography, comparative studies • Radiography, digital • Radiography, selenium detector
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Several aspects have to be considered with regard to tube voltage, however: The detective quantum efficiency of digital selenium radiography has been demonstrated to be significantly higher than that of other image detectors, including both film-based and storage phosphor systems (3). Scatter fraction of digital selenium radiography is decreased when a technique with lower tube voltage is used. The modulation transfer function of digital selenium radiography increases with lower tube voltage (4).
Like other digital detectors (eg, storage phosphor radiography), digital selenium radiography has a wide dynamic sensitivity range (2). A contrast-detail phantom study showed superior detail detectability on images obtained with lower peak voltage for an amorphous silicon flat-panel detector compared with storage phosphor radiography (5). The purpose of our study was to compare three tube voltages in digital selenium radiography for the detection of simulated interstitial lung disease, catheters, and nodules.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Simulated Patterns
The following patterns of simulated interstitial lung disease, catheters, and nodules were used: (a) ground-glass pattern, simulated with layers of cellulose soaked with contrast medium (iopamidol [Solutrast 300] and ioxithalamic acid [Telebrix]; Bracco-Byk Gulden, Konstanz, Germany); (b) linear pattern, simulated with silk thread soaked with contrast medium (Solutrast 300); (c) miliary pattern, simulated with clusters of birdseed (approximate size of each grain, 2 mm); (d) reticular pattern, simulated with gauze soaked with contrast medium (Solutrast 300); (e) catheters—venous catheters had a length of 30–50 mm and an 18-gauge lumen (B. Braun Melsungen, Germany); and (f) nodules, which were simulated with beeswax 5–15 mm in diameter—50% of the nodules were 10 mm or smaller.
Image Acquisition
Fifty posteroanterior chest radiographs were obtained with a digital selenium radiography system (Thoravision; Philips Medical Systems) with the following fixed parameters: Two-meter focus-film distance, 0.1 mm of copper, and 0.1 mm of aluminum filtration at 150 kVp. The phantom was positioned for standard posteroanterior chest exposure. The entrance surface dose was measured by using an ionization chamber (DALi type 77217 dosimeter and type 77334 ionization chamber; PTW, Freiburg, Germany). With the same parameters, 100 additional radiographs were obtained with the same surface dose measured at 150 kVp (recommended by the manufacturer) but with different tube voltages (50 radiographs each at 70 and 90 kVp). Tube voltages below 100 kVp were chosen, since our own experiments prior to this study showed that this lower peak voltage might be advantageous compared with high tube voltages for the detection of simulated interstitial lung disease. The surface dose was adjusted manually by using the ionization chamber to change the tube current.
The air gap of 15 cm is permanent to reduce scattered radiation. An additional stationary grid with 60 lamellae per centimeter and a ratio of 12:1 was used.
Variation of Parameters
Fifty posteroanterior chest radiographs were obtained with 150 kVp. The exit dose was 18.5 µGy and was defined as the dose at the beam exit side of the patient, normally in the center of the radiation field. The surface dose was 0.52 mGy.
Fifty posteroanterior chest radiographs were obtained with 90 kVp. The exit dose was 16.0 µGy, and the surface dose was 0.51 mGy.
Fifty posteroanterior chest radiographs were obtained with 70 kVp. The exit dose was 8.6 µGy, and the surface dose was 0.53 mGy.
Processing for the Digital Detector System
Processing of all images was performed by using the system-integrated program for chest examinations. All data were processed with one defined routine image processing technique for chest radiographs (the same technique used with patient images) to achieve the best possible detection of details: lung density, 1.80; abdomen density, 0.65; gamma lower limit, 1.89; gamma upper limit, 4.00; detail contrast enhancement, 1.00; and noise compensation, 0.75. This processing technique was chosen by means of consensus of two radiologists (T.M.B., U.R.B.) and a physicist (H.L.) to be optimal for images acquired with the three tube voltage techniques.
Hard-copy digital selenium radiographs were printed with a laser imager (Matrix LR 5200; Agfa, Leverkusen, Germany), with a film size of 35 x 43 cm. Films were used for this study because viewing of images on monitors has not been established on a routine basis at the University of Muenster.
Data Acquisition and Analysis
A total of 150 chest radiographs (50 templates times three subsets [70, 90, and 150 kVp]) were prepared in random order at five sittings. The Excel program (Microsoft, Redmond, Wash) offers a random number function to return a number generated from the uniform distribution on the interval (column 1). In a parallel column (column 2), we arranged the numbers of templates. After sorting the random numbers of column 1 in descending order, column 2 contained the random order for the templates. This was done for each reader.
Prior to each sitting, an image without any pattern was presented to the readers. Four board-certified radiologists (with 8–13 years of experience, all trained in a thoracic imaging fellowship; U.R.B., S.D., K.P., K.L.) and one fellow (7 years of experience, trained in a thoracic imaging fellowship; T.M.B.) participated in this study. For each area of the 50 different templates and for each pattern, the readers were asked to state the presence or absence of a simulated pattern and to rank their level of confidence on a five-point scale, as follows: 1 = definitely present, 2 = probably present, 3 = uncertain, 4 = probably not present, and 5 = definitely not present. A standard film viewer was used for assessment of the hard-copy images.
Data and Statistical Analysis
A total of 54,000 (50 templates, 12 areas per template, five readers, three tube voltages, and six patterns) observations were analyzed by using receiver operating characteristic (ROC) analysis. The critical patterns for daily routine work (ground-glass, linear, miliary, and reticular patterns over lucent lung [low attenuation], catheters over obscured chest regions [high attenuation], nodules over lucent lung [low attenuation], and obscured chest regions [high attenuation]) were assessed separately as follows.
Pattern detectability was estimated by means of the area under the ROC curve (Az). For the calculation of Az values, the LABMRMC program (1.0B BETA version 3; Charles E. Metz and Benjamin A. Herman, University of Chicago, Ill) was used (6). First, the program was used for calculations across the three tube voltages and the five readers. To calculate the confidence intervals of the Az values and the pairwise comparisons for two modalities, the analyses were repeated for each combination of two voltage techniques. The program was used for each pattern separately. Mean Az values and standard deviations were calculated for each pattern and voltage technique. The program was also used to calculate Az values for each reader and each tube voltage.
For an adequate assessment of significant differences, Bonferroni adjustment was performed. The 95% confidence intervals of mean Az values were calculated.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
10 mm and >10 mm) over lucent lung (P > .05), while there was a significant increase of diagnostic performance in the detection of reticular pattern at 90 kVp (P = .04). The detection of catheters and nodules over obscured chest regions showed lower Az values at 70 and/or 90 kVp than those at 150 kVp. These results were not statistically significant (P > .05).
|
|
|
|
|
|
Miliary Pattern
For the detection of miliary opacities, no significant differences were found between Az values at 70, 90, and 150 kVp over lucent lung (70 vs 150 kVp, P = .68; 90 vs 150 kVp, P = .99). Images acquired at 70 kVp (Az = 0.80) showed slightly decreased results. Images acquired at 90 kVp (Az = 0.83) showed equivalent results to those at 150 kVp (Az = 0.83) (Fig 3).
|
|
|
Catheters
There were no statistically significant differences between 70 and 150 kVp (P = .27) and 90 and 150 kVp (P = .41) in the detection of catheters over obscured chest regions. The Az values of catheters on digital images obtained with 70 and 90 kVp were 0.57 and 0.62, respectively. For detection of catheters on images obtained with 150 kVp, the Az value was 0.70 (Fig 4).
|
|
|
Over obscured chest regions, the detection of nodules larger than 10 mm was not significantly different at 70 and 150 kVp (P = .69) or at 90 and 150 kVp (P = .99) (Fig 5). For the detection of nodules 10 mm or smaller over obscured chest regions, there was no statistically significant difference between 70 and 150 kVp (P = .43) or between 90 and 150 kVp (P = .40).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
A nonanthropomorphic phantom study (5) with a Leeds phantom provided the first evidence that a low-voltage technique resulted in better diagnostic performance of digital selenium radiography. Leitz et al (10) used an anthropomorphic test phantom with different chest radiography systems and found no correlation between image quality and different tube voltages. However, they did not evaluate the combined effect of variable parameters on the overall contrast and signal-to-noise ratio of details. Sandborg et al (11) described schemes for the optimization of chest radiography by using a computer model of the patient and x-ray imaging system. They pointed out that at low tube voltage, image quality may be limited by a low properly exposed fraction, and this will set an upper limit of the speed that can be used. This restriction will not apply for digital chest radiography, however, where the energy imparted to the image detector can be varied over a much larger range than in film-based radiography and the contrast can be modified at the image display stage.
Verification of the correct x-ray tube voltage is an important quality assurance issue in diagnostic radiology. Therefore, we performed a phantom study by using an anthropomorphic chest phantom and well-established simulation low-contrast material for subtle interstitial lung disease, catheters, and nodules at different tube voltages for digital selenium radiography. The improved signal-to-noise performance of the selenium detector manifested as better detectability of low-contrast objects over lucent lung. These materials have already been proved to be realistic because of the confirmation of results of phantom studies when applied in patient trials (7,9). Our study showed that almost all of the simulated interstitial lung pattern resulted in higher Az values for images obtained with 70 and/or 90 kVp; however, this was only statistically significant for the reticular pattern at 90 kVp. The main reason for the improved image quality at lower peak voltage is the increase of x-ray absorption by the selenium detector and the higher object contrast available (12). The detection of catheters and nodules over obscured chest regions was equivalent or even slightly decreased at 70 and/or 90 kVp without a statistically significant difference from that on images obtained with 150 kVp. Chotas et al (13) showed decreased signal-to-noise ratio with increasing peak voltage in lung areas, whereas the signal-to-noise ratio in the mediastinum was independent of peak voltage. Their results were based on a photostimulable phosphor digital radiography system.
Our results were obtained in spite of the remarkably reduced exit dose, as shown in the Materials and Methods section, since we aimed to perform a study with low tube voltage and the same surface dose used with high tube voltage. However, both the more than twofold detective quantum efficiency when using low versus high tube voltage and the more than twofold primary x-ray absorption in the selenium drum outweighs the disadvantage of the lower exit dose.
Neitzel et al (3) indicated an increase of detective quantum efficiency by a factor of two to three when compared with phosphor-based x-ray detectors. The reason for this better performance can be found in the different amounts of internally generated noise. The conversion process in selenium is virtually free of intrinsic noise sources. In addition, within the photoconductor selenium, the release charge of x-rays can be guided to the surface by an electric field to prevent blurring in contrast to other digital detectors, such as storage phosphor radiography (3). Therefore, within the selenium detector, there is less blurring than in other digital detectors.
Launders et al (12,14) indicated an increase of detective quantum efficiency from 0.26 to 0.57 for decreasing tube voltage from 150 to 60 kVp. As a result of higher quantum absorption and low system noise, the image noise was also significantly lower than that of other detectors, with better detection of subtle pulmonary patterns, as we were able to show over lucent lung, also at lower tube voltages with reduced exit dose. Launders et al suggested in their clinical trial that the lower peak voltage (90 kVp) has advantages in terms of perceived image quality over the generally accepted higher peak voltage (150 kVp), which is becoming the standard technique (12).
The effective surface dose decreased with increase of peak voltage. This is the argument to use higher peak voltage settings. However, Launders et al (12) showed that the calculated normalized effective dose data increased with peak voltage in contrast to the measured entrance surface dose. They stated that the effective dose reached a minimum value at around 110 kVp, and, on the basis of their interpolation, this effective dose was significantly less than it was at 150 kVp.
With low-voltage techniques, the scattered radiation is reduced, as shown by Neitzel at al (3). Baorong et al (15) indicated in their experimental study that scattered radiation plays an important role. Their Monte Carlo calculations of the contrast curve showed that the radiation that scattered back from the walls, floor, and ceiling of their experimental room had an influence of 1% or 2% on the values of the contrast curve. Boone et al (16) pointed out that the absorbed primary energy decreased with increasing energy of x-rays; the backward secondary emission has a peak at almost 20 keV, and the forward secondary emission has a peak at 35 keV, with a corresponding energy deposition of 15% and almost 2%, respectively.
The anthropomorphic chest phantom used for the present study is slim. Therefore, we used additional 40-mm acrylic plates to simulate a more obese patient. Nevertheless, as a limitation of our study, the influence of scatter fractions and varying body diameters was not assessed and may limit the actual practical value of the technique of using lower tube voltage. It has to be emphasized that thicker phantom materials are associated with a marked decrease of transmission for low tube voltages, which results in more blurring of the images for the protocol we used with the same surface dose but reduced exit dose.
Consequently, by using our protocol with the same surface dose at each tube voltage, possible image degradation may occur in obese patients as a result of the reduced exit dose for these patients. This effect might outweigh the advantage of lower tube voltage. Therefore, a further study has to be performed to examine the influence of lower tube voltage in obese patients. In our study design, there is an inherent dependency: Adjacent areas could create interpretative biases because of their adjacency, since only the 50 templates were stochastically independent.
In a previous study in which a similar study design was used, however, we modeled the dependency between two areas from the same template by a random template factor by using general linear model procedure of the SAS system to analyze the original data from the five-point confidence scale (17). In these models, the factor "template" was not statistically significant. Therefore, we concluded that it was possible to reduce the model and to use the simplification of the model with the independent areas for the ROC analysis with the LABMRMC program (Charles E. Metz and Benjamin A. Herman).
The intrinsic modulation transfer function of the selenium detector is moderate compared with other detector systems. The modulation transfer function value at 2 line pairs per millimeter is smaller (by a factor of approximately three) than that of the storage phosphor plates. The modulation transfer function describes how the contrast of image components is transmitted as a function of their spatial frequency and is expressed in line pairs per millimeter (2). However, it is important that in a digital system the detector modulation transfer function alone does not necessarily limit the image quality (3). As shown by Que and Rowlands (4), the modulation transfer function for a 500-µm-thick amorphous selenium detector increased with lower kiloelectron voltage. The measured modulation transfer function was extremely high, especially for lower x-ray photons (23 line pairs per millimeter at 100 keV and 91 line pairs per millimeter at 50 keV). Therefore, we could explain the better detectability of low-contrast patterns over lucent lung.
Interstitial lung disease, with its low contrast and high spatial frequencies, is an adequate tool for the evaluation of a new detector system and acquisition parameters. Digital selenium radiography provides both a higher detective quantum efficiency as a function of dose and a higher spatial frequency than other detectors, which overcome the disadvantage of the lower exit dose applied in our protocol. Digital selenium radiography has an almost linear relationship between dose and signal over a wide range of exposures (wide dynamic range) and a postprocessing feature that is well adapted to the chest (2). Digital selenium radiography offers an adequate visualization of lucent lung and areas of high attenuation over the mediastinum and the diaphragm in one image that is even more superior to images from asymmetric screen-film systems. This effect is even better on images obtained with low tube voltage, as we demonstrated in our study. Digital selenium radiography is suitable for the detection of various patterns of simulated interstitial lung disease, catheters, and nodules at lower peak voltage (90 kVp) and offers the potential of higher Az values because of its high detective quantum efficiency.
Radiographs obtained with lower tube voltage showed increased noise in areas of high attenuation as a result of the reduced exit dose. It has to be evaluated in a patient study whether this effect is of clinical relevance when acquiring images with lower peak voltage. Moreover, when applying our protocol with the same surface dose for the phantom, the exposure time increased from 13.5 msec at 150 kVp to 27.5 msec at 90 kVp to 50 msec at 70 kVp. Since the standards of the American College of Radiology for the performance of chest radiography in adults require an exposure time of less than 40 msec, the 70-kVp protocol exceeded this recommendation. Potential clinical problems of longer exposure times are motion artifacts that result in degradation of image quality.
When the results of all readers are analyzed, it is noteworthy that one of the readers had lower Az values for almost all patterns (except for nodules of any size over lucent lung) and all tube voltages. A possible reason might be that the research and clinical interests of this reader are mainly based on the musculoskeletal system. This reader has also never participated in a phantom study like this before. The reader with the most experience in chest radiography performed as well as or better than all the other readers. This reader has also never participated in a phantom study like this before. A reason for the worst detection of reticular pattern of this reader was not found.
Our results showed that for simulated interstitial lung disease, catheters, and nodules, a reduction of the tube voltage is possible for digital selenium radiography, with improved diagnostic performance over lucent lung and no significant loss in performance over obscured chest areas, even with a reduced exit dose. However, the diagnostic accuracy in daily routine can be influenced according to such additional properties as image processing and has to be assessed by using clinical images.
Practical application: Digital selenium radiography is suitable for the detection of various patterns of simulated interstitial lung disease, catheters, and nodules at lower peak voltage (90 kVp) and offers the potential of higher Az values as a result of its high detective quantum efficiency.
|
FOOTNOTES |
|---|
Abbreviations: Az = area under ROC curve, ROC = receiver operating characteristic
Author contributions: Guarantors of integrity of entire study, T.M.B., U.R.B.; study concepts and design, T.M.B., U.R.B.; literature research, T.M.B., U.R.B.; experimental studies, T.M.B., U.R.B., H.L., S.D., K.P., K.L.; data acquisition, T.M.B., U.R.B., H.L.; data analysis/interpretation, T.M.B., U.R.B., F.W.R.; statistical analysis, F.W.R.; manuscript preparation, T.M.B., U.R.B.; manuscript definition of intellectual content, T.M.B., W.H.; manuscript editing, T.M.B., U.R.B.; manuscript revision/review, T.M.B., W.H.; manuscript final version approval, T.M.B.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. W. Venema, M. van Straten, G. J. den Heeten, T. M. Bernhardt, H. Lenzen, and W. Heindel Digital Radiography of the Chest: Reassessment of the High-Voltage Technique? * Dr Bernhardt and colleagues respond: Radiology, April 1, 2005; 235(1): 336 - 338. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
