| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Experimental Studies |
1 From the Department of Clinical Radiology (K.L., C.S., S.D., D.W., H.L., T.M.B., W.H.) and Institute of Experimental Biomechanics (P.B.), University of Munster, Albert-Schweitzer-Strasse 33, D-48129 Munster, Germany. Received April 1, 2002; revision requested June 11; revision received July 9; accepted August 16. Address correspondence to K.L. (e-mail: lud@uni-muenster.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Subtle fractures were created artificially in 100 of 200 porcine rib specimens. Specimens were enclosed in containers of water to generate absorption and scatter radiation conditions similar to those of a human chest wall. Imaging was performed with flat-panel, screen-film, and storage-phosphor systems with conditions that were exactly matched. Different exposure levels equivalent to speed classes (S) of 400, 800, 1,600, and 6,400 were used. All images were independently assessed for the presence of fracture by three radiologists with a five-level confidence scale. Receiver operating characteristic (ROC) analysis was performed for a total of 4,200 observations (600 for each imaging system and exposure level). Diagnostic performance was estimated with area under the ROC curve (Az). Significance of differences in diagnostic performance was tested with analysis of variance.
RESULTS: ROC analysis yielded mean Az values for the flat-panel system of 0.879 (S = 400), 0.833 (S = 800), 0.765 (S = 1,600), and 0.576 (S = 6,400). Az values were 0.834 (S = 400) for the screen-film system and 0.789 (S = 400) and 0.729 (S = 800) for the storage-phosphor system. Analysis of variance revealed significant differences in diagnostic performance between various combinations of imaging system and exposure levels (P < .05).
CONCLUSION: The flat-panel system is superior to the screen-film and storage-phosphor systems for detection of subtle undisplaced rib fractures at clinical exposure settings (eg, S = 400). With the flat-panel system, radiation dose can be reduced by 50% to achieve diagnostic performance comparable to that of a speed class 400 screen-film system.
© RSNA, 2003
Index terms: Experimental study • Radiations, measurement • Radiography, digital • Radiography, flat panel • Radiography, comparative studies • Ribs, fractures, 471.41
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The spatial resolution of flat-panel systems, on the other hand, because it is limited by pixel size, is considerably lower than that of screen-film systems (2–4). Therefore, special attention must be focused on the performance of flat-panel systems in the depiction of fine lesions, for which spatial resolution can be expected to be an important factor.
Results of various experimental (6–9) and clinical studies (10–13) have already revealed that, for demonstrating lesions that are not critical to depict in terms of spatial resolution, flat-panel systems yield superior image quality and have a lower radiation dose requirement as compared with screen-film systems.
The aim of our study was to compare a large-area direct read-out flat-panel system with a screen-film system and a storage-phosphor system with regard to the detection of subtle undisplaced rib fractures and to assess the diagnostic performance of this system with decreasing exposure levels.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
Imaging Techniques
Images were obtained with three different imaging systems:
1. A large-area direct read-out flat-panel detector system (Digital Diagnost; Philips Medical Systems, Hamburg, Germany) with a 500-µm layer of cesium iodide for the conversion of x rays to light and an amorphous silicone matrix for the conversion of light to electric charge. This detector provides a pixel size of 143 x 143 µm (Nyquist limit, 3.5 line pairs per millimeter [lp/mm]) in a 3,000 x 3,000 pixel matrix, resulting in a 43 x 43-cm field of view.
2. A storage-phosphor system (ADC compact; Agfa, Leverkusen, Germany), which provides a pixel size of 118 x 118 µm (Nyquist limit, 4.2 lp/mm) when used with a film size of 18 x 24 cm.
3. A speed class 400 screen-film system (Insight Skeletal Regular; Kodak, Rochester, NY), which provides a spatial resolution of 6.5 lp/mm with a modulation transfer function of 10% or higher.
Special care was taken to match exposure conditions for the three imaging systems as precisely as possible: All imaging was performed with a standard x-ray tube and generator (Philips SRO 33 100/SCP 80; Philips Medical Systems) and with a 0.6-mm focal spot size. Two sets of this tube and generator combination had to be used because screen-film and storage-phosphor systems are cassette-based systems, while the flat-panel detector is integrated into the x-ray system and cannot be removed. Identical imaging conditions with both sets were ensured because values for output and mean kilovolt peak, as measured by the Federal Institute of Radiation Safety of Germany according to federal law on the use of ionizing radiation at the time of installation of both systems (independent of this study), were identical for both tube-and-generator sets. (The set for the flat-panel system had been installed 5 months, and the set for the screen-film and storage-phosphor systems had been installed 1 month prior to our investigation.)
All imaging was performed with a moving antiscatter grid (grid ratio, 12:1; 40 lp/mm). Geometric image parameters were identical for all imaging systems, with a focal-spot–to–object-plane distance of 115 cm and an object-plane–to–detector-plane distance of 8 cm. Collimation was kept constant (at 4 x 20 cm) to avoid systematic errors caused by differences in scatter radiation.
A voltage of 70 kVp with a total filtration of 3 mm aluminum was used, corresponding to the voltage used in routine clinical imaging of the human chest wall. In a series of images of one of the specimens obtained with different exposure levels, the optimal clinical exposure for the speed class 400 screen-film system was determined in consensus by three authors (K.L., W.H., H.L.). This exposure, achieved with a current-time product of 12 mAs, was used for the speed class 400 screen-film system, the flat-panel system, and the storage-phosphor system and is subsequently referred to as being equivalent to speed class 400.
With the flat-panel system, additional imaging was performed at 
, 
, and 1/16 of the current-time product used with the screen-film system. With the storage-phosphor system, additional images were obtained at of the current-time product used with the screen-film system. Lower radiation doses were not possible with the storage-phosphor system used in this investigation. The additional series of images obtained with the flat-panel and the storage-phosphor systems are subsequently referred to as having been obtained at exposures equivalent to speed classes 400, 800, 1,600, and 6,400. Any use of automatic exposure control was intentionally avoided because it is a potential cause of unsteady exposure with different imaging systems.
For the flat-panel and storage-phosphor systems, the image processing methods recommended by the manufacturer were used. Image processing for the flat-panel system is based on an unsharp mask type of filtering with two differently sized kernels. A filter with a small (3 x 3) kernel serves to enhance image sharpness (constituting modulation transfer function [MTF] restoration), and a large kernel (201 x 201) allows dynamic range reduction and simultaneous detail enhancement. Both filters are nonlinear; that is, they are dependent on pixel intensity to avoid noise enhancement in the brighter (low-exposure) areas.
Image processing for the storage-phosphor system is based on a complex process in which the original image is separated into 12 subimages with different spatial frequencies by means of downsampling, averaging with a 5 x 5 kernel, and subtracting the resulting image from the original image. Each of the 12 subimages undergoes contrast amplitude equalization before subimages are put together for a final image.
The following image processing parameters, which had been set up during system installation by the radiologists and medical physicists of our institution in cooperation with agents of the manufacturers and which have proved to be valuable in daily clinical practice since that time, were used: for the flat-panel system, contrast density of 1.2, contrast enhancement of 1.6, and noise reduction of 0.0; for the storage-phosphor system, musi contrast of 2, edge contrast of 1, latitude reduction of 3, and noise reduction of 0.
The flat-panel and storage-phosphor images were printed on film by using a laser printer (Imation film/Imation DryView 8700; Kodak). For both systems, the display lookup table had a sigmoid shape similar to that of a film density curve. Lookup table gradients for both of the digital imaging modalities were chosen so that the optical densities measured with a densitometer (Unilight D; Wellhoefer, Schwarzenbruck, Germany) at three predefined locations (rib cortex, central part of rib spongiosa, and area lateral to the diaphyseal cortex) were identical to those in the speed class 400 screen-film images to avoid any bias caused by differences in brightness or contrast. All imaging was performed by one of the authors who was not involved in the image evaluation process (C.S.).
Image Evaluation
All images were assessed independently by three radiologists (S.D., T.B., D.W.), who recorded the presence or absence of a rib fracture in each image according to the following five-point confidence scale: 1, definitely positive for fracture; 2, probably positive for fracture; 3, uncertain; 4, probably negative for fracture; and 5, definitely negative for fracture. This resulted in a total of 4,200 observations (200 rib specimens times three radiologists times seven combinations of imaging system and exposure level). Readers were blinded to exposure settings and imaging systems. To prevent learning bias, all images were shown in a random order, and a training phase with specimens not included in the study preceded the image evaluation process. No time constraints were applied. None of the readers were involved in the preparation or imaging of the specimens. All images were viewed on the same light box with adjustable shutters under subdued ambient light.
Data Analysis
Data were analyzed by using receiver operating characteristic (ROC) analysis (14). ROC curves were created with a maximum-likelihood curve-fitting algorithm. Diagnostic performance was estimated with the area under the ROC curve (Az). Az values were calculated for each independent reader for the seven combinations of imaging system and exposure level (screen-film system 400; flat-panel system 400, 800, 1,600, and 6,400; and storage-phosphor system 400 and 800). Furthermore, composite Az values for the three readers were calculated according to the method described by Metz (14) for the seven combinations of imaging system and exposure level.
The statistical significance of differences in diagnostic performance was tested by using analysis of variance according to the method described by DeLong et al (15) for the comparison of multiple Az values. A P value of less than .05 was considered to represent a statistically significant difference. Comparisons were made between or among (a) images obtained with all speed class 400 systems, (b) images obtained with the flat-panel system at each exposure level, (c) images obtained with the storage-phosphor system at each exposure level, and (d) images obtained with the flat-panel system at speed class 800 and those obtained with the screen-film and storage-phosphor systems at speed class 400. Statistical analyses were performed with ROCFIT (C. E. Metz, Department of Radiology, University of Chicago, Ill) and SAS software packages (SAS, Cary, NC).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
Az values for the individual readers for all combinations of imaging system and exposure level are listed in the Table.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
It is therefore difficult to predict the diagnostic performance of flat-panel systems in the depiction of fine details such as subtle, undisplaced fractures on the basis of the physical data described above. To assess the diagnostic performance of these systems, experimental or clinical studies are needed.
The performance of a flat-panel system in the depiction of fractures has previously been investigated by Strotzer et al (7). In our opinion, however, their study differed from ours: The depiction of fractures with more than a minimal degree of dislocation, as evaluated in that study, is not as critical for the assessment of any existing imaging system, in that the depiction of such fractures represents a high-contrast, low-spatial-resolution task. The fracture width or the amount of dislocation in that study is not quantified; the fractures are only described as "as minimally dislocated as possible" (7).
With 9,280 observations in that study, however, no significant differences were revealed between images obtained with the same flat-panel system at an exposure level equivalent to speed class 400 and those obtained with an exposure level equivalent to speed class 1,600, so we believe that depiction of the fractures produced in that study cannot be regarded as a critical imaging task (7). Therefore, the conclusion of Strotzer et al (7) that no significant difference between the screen-film system and the flat-panel system was shown at any exposure level is of limited value.
Moreover, the authors in that study state that the overall contrast of images obtained with their flat-panel system was lower than that of images obtained with their screen-film system (7). Overall contrast, however, may have a strong effect on observer performance in the detection of subtle lesions and should therefore be identical in all imaging systems compared. Furthermore, the study mentioned incorporated the use of a small (15 x 15-cm field of view) prototype detector (7). The results of such a prototype cannot automatically be considered commensurate with those of clinically applied systems because systems ready for clinical application may perform better (because of further development) or worse (because of changes made to achieve better stability or practicability of the entire imaging system).
In our study we tried to avoid these situations: The fact that the diagnostic performance of both the flat-panel and the storage-phosphor systems was significantly different at different exposure levels shows that depiction of the fractures produced in our model did in fact constitute a critical imaging task. We controlled contrast and brightness in the images we compared by measuring optical densities in certain areas, and we adapted brightness and contrast for both of the digital imaging systems so that they were comparable to those of the screen-film system. Furthermore, the imaging systems used in our study are used in daily clinical practice.
The range of Az values in our study shows that visibility of the experimentally created fractures was appropriate for the evaluation of the imaging system and exposure level combinations we compared: With exception of the Az values achieved with the flat-panel system at speed class 6,400, Az values ranged from 0.729 to 0.879. This is an appropriate range for the assessment of imaging systems, because Az values too close to 1.0 would represent fractures that are too easily detected, and Az values too close to 0.5 would represent fractures that are so difficult to detect that they should not be used for the evaluation of different imaging systems (14).
Exposure settings in our study corresponded to those used in clinical imaging of the human chest wall. We estimated that the water surrounding the specimens produced absorption and scatter radiation conditions approximating those of radiography of the human chest wall. This estimation was based on the thickness of the thoracic wall seen in clinical practice (eg, at computed tomography of the chest) and on the current-time product found optimal for the screen-film system, which is in the same range as typical current-time products at clinical imaging of the chest wall. Due to the fact that absorption and scatter radiation vary largely with patient constitution in any case, we believe that this rough estimation was appropriate for the purpose of our study.
As a limitation of our study, no superimposed lung structures were simulated. Also, the absorption of the water basin in our study was much more homogeneous than the absorption of different areas of the chest wall (eg, the areas superimposed over the diaphragm versus those superimposed over the lung only). Our model was therefore much less dynamic in terms of absorption range than an actual human chest wall. In contrast to the collimation used in clinical imaging, where a larger part of the chest is usually imaged and the distance between the fractured rib and the detector is more variable, the collimation used in our study was smaller, and the object-to-detector distance was constant. These differences from clinical imaging protocol, however, are likely to affect images obtained with any of the imaging systems we evaluated in the same way. Therefore, we believe that our results would not have been different if we had simulated superimposed structures or had used larger collimations or different object-to-detector distances.
It should be emphasized, furthermore, that the imaging systems in this study were evaluated strictly as complete imaging systems, including not only the detector itself but also all the steps of image generation (ie, image processing). It can be speculated that the diagnostic performance of both of the digital imaging systems we evaluated may have been improved if we had used them with soft-copy display with interactive adjustment of contrast, brightness, and contrast detail enhancement. Whether further improvement can be achieved by making specific changes in the image processing methods used with the flat-panel or the storage-phosphor system cannot be predicted on the basis of the data we obtained. Therefore, the results of our study apply only to hard-copy display under the conditions described above.
Although readers were blinded to imaging system and exposure level, the experienced readers might have been able to estimate whether a very low radiation dose was used on the basis of their subjective impression of image noise. This limitation could not be avoided and applies to similar studies as well.
Our data were based on a large number of observations, were gathered according to a study design characterized by precisely matched exposure conditions and a maximum effort to avoid systematic errors, and lead to the following conclusion: On the basis of our results with the imaging systems we studied, when an identical radiation dose is administered, a flat-panel system is superior to a speed class 400 screen-film system and to a storage-phosphor system for the detection of subtle undisplaced rib fractures. With a radiation dose reduction of 50%, the diagnostic performance of a flat-panel system is still equal to that of a screen-film and a storage-phosphor system.
Practical application: Compared with the use of screen-film and storage-phosphor systems, the use of a digital flat-panel system for fracture detection can result in improved diagnostic performance and reduced radiation exposure.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, K.L.; study concepts and design, K.L.; literature research, W.H.; experimental studies, C.S., P.B.; data acquisition, C.S., P.B.; data analysis/interpretation, S.D.; statistical analysis, H.L.; manuscript preparation and definition of intellectual content, D.W.; manuscript editing, T.M.B.; manuscript revision/review and final version approval, K.L.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. L. Kleinman, D. Zurakowski, K. J. Strauss, R. H. Cleveland, J. M. Perez-Rosello, D. P. Nichols, K. H. Zou, and P. K. Kleinman Detection of Simulated Inflicted Metaphyseal Fractures in a Fetal Pig Model: Image Optimization and Dose Reduction with Computed Radiography Radiology, May 1, 2008; 247(2): 381 - 390. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Neofotistou, V. Tsapaki, S. Kottou, A. Schreiner-Karoussou, and E. Vano Does digital imaging decrease patient dose? A pilot study and review of the literature Radiat Prot Dosimetry, December 1, 2005; 117(1-3): 204 - 210. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
U. Rapp-Bernhardt, T. M. Bernhardt, H. Lenzen, R. Esseling, F. W. Roehl, M. Schiborr, I. Theobald-Hormann, and W. Heindel Experimental Evaluation of a Portable Indirect Flat-Panel Detector for the Pediatric Chest: Comparison with Storage Phosphor Radiography at Different Exposures by Using a Chest Phantom Radiology, November 1, 2005; 237(2): 485 - 491. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. Huda The Current Concept of Speed Should Not Be Used to Describe Digital Imaging Systems Radiology, February 1, 2005; 234(2): 345 - 346. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
