点击显示 收起
1 From the Department of Clinical Radiology, University of Muenster, Albert-Schweitzer-Strasse 33, D-48129 Muenster, Germany (K.L., H.L., T.M.L., S.D., D.W., W.H.); and Philips Medical Systems, Hamburg, Germany (K.F.K.). From the 2000 RSNA scientific assembly. Received January 11, 2001; revision requested February 26; final revision received August 6; accepted August 15.
| ABSTRACT |
|---|
| Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
|---|
MATERIALS AND METHODS: Artificial lesions (0.5–3.0 mm) were created in 100 of 200 predefined regions in 20 porcine femoral specimens. Specimens were enclosed in containers filled with water to create absorption and scatter radiation conditions comparable with those in a human extremity. Imaging was performed with a flat-panel detector system, a conventional screen-film system, and a storage-phosphor system. Levels of exposure equivalent to speed classes 400, 800, 1600, and 3200 were used. In all images, the presence or absence of a lesion was assessed by three radiologists using a five-point confidence scale. Receiver operating characteristic (ROC) analysis was performed for 4,800 observations (600 for each imaging modality and exposure level) and diagnostic performance estimated with the area under the ROC curve (Az). The significance of differences in diagnostic performance was tested with analysis of variance.
RESULTS: ROC analysis showed Az values of 0.820 (speed class 400), 0.780 (class 800), 0.758 (class 1600), and 0.676 (class 3200) for the flat-panel detector; 0.761 (class 400), 0.725 (class 800), and 0.662 (class 1600) for the storage-phosphor system; and 0.788 (class 400) for the conventional screen-film system. The Az value for the flat-panel detector at speed class 400 was significantly higher than that for all other systems (P < .05). Az values for the speed class 400 screen-film system and flat-panel detector system at speed class 800 were not significantly different.
CONCLUSION: The flat-panel detector has diagnostic performance superior to that of conventional screen-film and storage-phosphor radiography for detecting small artificial osseous lesions at clinical exposure settings. With the flat-panel detector, exposure dose can be reduced by 50% to obtain diagnostic performance comparable with that of a conventional speed class 400 screen-film system.
Index terms: Bones, radiography, 444.121 • Radiography, digital, 444.121 • Radiography, flat panel, 444.121
| INTRODUCTION |
|---|
| Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
|---|
Initial study (5–11) results suggested superior image quality and considerable potential for reducing radiation dose with these flat-panel detectors, compared with standard systems. Spatial resolution with flat-panel detectors, however, is lower than that with high-speed conventional screen-film systems (12). Therefore, with the introduction of this new-generation technology, special attention must be focused on its performance in the detection of osseous lesions, which requires not only contrast resolution but also spatial resolution.
The purpose of our study was to compare a large-area direct-readout flat-panel system with state-of-the-art conventional screen-film and storage-phosphor systems for detecting artificial osseous lesions simulating osteolytic disease and to assess its potential in reducing exposure dose.
| MATERIALS AND METHODS |
|---|
| Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
|---|
|
To simulate the surrounding soft tissue of a human extremity, each femoral bone was enclosed in a cylindrical polypropylene container (12-cm diameter) filled with water. To assure correct repositioning and collimation, each femoral bone was fixed within the container by using surgical k-wires (Synthes, Bochurn, Germany). The surgical k-wires were located in the proximal femur only, so that they were not visible on images.
Imaging Technique
Images were obtained with three imaging systems:
1. A large-area direct-readout flat-panel detector (Philips Medical Systems, Hamburg, Germany) that uses a 500-µm layer of cesium-iodide alloyed with a small amount of thallium for conversion of x rays to light and an amorphous silicone matrix for conversion of light to electric charge; the detector provides a pixel size of 143 x 143 µm (Nyquist limit, 3.6 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) that provides a pixel size of 118 x 118 µm (Nyquist limit, 4.2 lp/mm) when used with an 18 x 24-cm film size; and
3. A conventional speed class 400 screen-film system (Insight Skeletal Medium; Kodak, Rochester, NJ) that provides a spatial resolution of 5.5 lp/mm.
Special care was taken to match the exposure conditions for the three imaging systems as precisely as possible. All imaging was performed with a standard x-ray tube and generator (Rotalix/Optimus; Philips Medical Systems). The focal spot–object plane distance was 135 cm, and the object plane–detector distance was 8 cm for all images. To avoid any bias caused by differences in scatter radiation, a constant collimation of 11 x 16 cm was used. All imaging was performed with a moving antiscatter grid (grid ratio, 12:1; 40 lines per centimeter).
Corresponding to imaging of the human femur, a voltage of 70 kVp with a total filtration of 3 mm Al was used. In a series of images of one of the specimens obtained with different exposure doses, the optimal clinical exposure dose (detector entrance dose of 3.9 µGy with a 25-mm-Al phantom) was determined for the conventional speed class 400 screen-film system. This exposure dose was used for all three imaging systems, hereafter referred to as "equivalent to speed class 400." Automatic exposure control was arbitrarily avoided as a potential cause of unsteady exposure dose with the different systems. With the digital flat-panel detector system and storage-phosphor system, additional imaging was performed at lower exposure doses, and . For the flat-panel system, an additional of this current-time product also was used (with the storage-phosphor system in our study, however, use of this low-exposure dose was technically not possible). The resulting images are hereafter referred to as "equivalent to speed classes 800, 1600, and 3200." All imaging was performed by one of the authors who was not involved in image evaluation (K.L.).
The flat-panel and storage-phosphor images were printed on film (Imation; Kodak) by using a laser printer (Imation DryView 8700; Kodak). To avoid any bias caused by differences in brightness or contrast between imaging modalities, look-up tables for both digital imaging modalities were chosen, such that the optic densities, measured with densitometry (Unilight D; Wellhoefer, Schwarzenbruck, Germany) for three predefined locations (ie, the diaphyseal cortex, diaphyseal spongiosa, and directly lateral to the diaphyseal cortex), were identical to those in the conventional images.
Image Evaluation
All images were assessed independently by three radiologists (T.M.L., S.D., D.W.), who recorded the presence or absence of a bone lesion. This resulted in a total of 4,800 observations (600 observations per imaging modality and exposure dose level). To prevent learning bias, all images were shown in random order. None of the readers was involved in preparing the phantoms. No time constraints were used. All images were viewed under subdued ambient light by using a light box with adjustable shutters. Exposure settings were not indicated on the images. The different appearance of digital and conventional images made it impossible, however, to blind readers to imaging modality. In all images, cortical bone was covered by an individual mask so that the readers could see only the trabecular bone. Thus, only the spongeous extent of the lesions was assessed. Readers were asked to rate the presence of a lesion according to a five-point confidence scale (a rating of 1 indicated that findings were definitely positive; 2, probably positive; 3, uncertain; 4, probably negative; or 5, definitely negative).
Data Analysis
Data were analyzed with ROC analysis (15). ROC curves were created with a maximum-likelihood curve-fitting algorithm. Lesion detectability was estimated by using values of the area under the ROC curve (Az). Because a range of lesion sizes instead of a single lesion size was used to guarantee variations in lesion visibility, lesion size was not taken into account.
By following methods described by Metz (15), Az values were calculated for the eight combinations of imaging modality and exposure dose (screen-film system 400; flat-panel system 400, 800, 1600, and 3200; and storage-phosphor system 400, 800, and 1600) for each of the three readers and as averaged Az values for all three readers. As described by DeLong et al (16), analysis of variance was performed to test the statistical significance of differences between mean Az values. ROCFIT (Metz CE, Department of Radiology, University of Chicago, Ill, 1989) and SAS (SAS, Cary, NC) software packages were used for statistical analysis.
| RESULTS |
|---|
| Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| DISCUSSION |
|---|
| Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
|---|
Although technical characteristics provide a valuable basis for comparing detector systems, it is difficult to transfer them into terms of diagnostic performance: For detecting small low-contrast lesions, both spatial resolution and contrast resolution are important factors.
In this study, we examined the diagnostic performance of a large-area flat-panel detector in an experimental model of artificial lesions that mimic myeloma-like osteolytic disease. An experimental design was chosen for several reasons: Because lesions were created artificially, a well-defined reference standard was available, which is frequently difficult to obtain in clinical studies. The design allowed a large number of observations, and radiation exposure of patients was avoided. The range of lesion sizes and thus lesion detectability could be chosen to acquire good discrimination of imaging methods: Lesions detected too easily would have resulted in an Az value too close to 1.0, which would not have been useful in comparing imaging systems; lesions too small to detect would have resulted in an Az value close to 0.5, also useless for system comparisons. The fact that mean Az values in our study ranged between 0.662 and 0.820 shows that lesion sizes were chosen in an appropriate range. Because lesion sizes ranged to a minimum of 0.5 mm, which is 3–4 times the pixel length with the flat-panel system, spatial resolution can be expected to be an important contributing factor in lesion depiction. Because of lesion localization in the medullary cavity, however, the lesions in our study can be considered low-contrast lesions. Contrast resolution, therefore, is an important factor in their depiction, as well.
Exposure settings in our study corresponded exactly to those used in imaging of a human femur. Porcine femurs have diameters similar to those of human femurs. The water surrounding the femur accounts for absorption resembling that of soft tissue surrounding a human femur. Long bones, furthermore, have relatively homogenous absorption compared with that of the spine or pelvis. Therefore, the wider dynamic range of the digital imaging modalities could not compensate for their lower spatial resolution.
Our data show the diagnostic performance of the flat-panel system to be superior to that of screen-film and storage-phosphor systems in detecting the simulated lesions. The diagnostic performance is almost equal to that of conventional screen-film systems when using only 50% of exposure dose (speed class 800). Further reduction of radiation dose leads to lower diagnostic performance. A 25% radiation dose (speed class 1600) with the flat-panel detector, however, still offers the diagnostic performance of a storage-phosphor system used with a 400–800 speed class dose equivalent.
The results of our study agree with those of previous studies. Using a medical physics contrast-detail phantom, Neitzel et al (12) demonstrated the superior performance of a flat-panel system in comparison with two conventional and two digital imaging modalities in detecting low-contrast objects. Strotzer et al (8) compared a flat-panel with a conventional screen-film system in detecting artificial rheumatoid lesions. They demonstrated superior image quality with the flat-panel system, with equal exposure dose and an equal image quality, with an exposure dose reduction of 30%–50%. With a minimum lesion size of 1.5 mm, however, their study design did not address the effect of spatial resolution. Vandevenne et al (10) evaluated a selenium-based flat-panel system in an experimental study. They showed that, even with an exposure dose reduced by 56%, image quality was comparable with that obtained with a conventional screen-film system. The authors, however, used only a qualitative descriptive approach, and their study was based on analysis of a single specimen.
Limitations of our study must be taken into account: An in vitro model was used instead of patient images. However, in patient images, a true reference standard usually is not available. The shape of artificial lesions was cylindrical rather than circular in most true osteolytic lesions. Complete elimination of observer bias was not possible because of the specific appearance of digital and conventional images. Our study, however, was based on a large number of observations and a study design characterized by precisely matched exposure conditions and maximum effort to avoid systematic errors.
Practical application: A digital flat-panel detector offers diagnostic performance superior to that of conventional screen-film radiography and storage-phosphor radiography in detecting experimental osteolytic lesions of various sizes at exposure settings equivalent to those of a speed class 400 system, as well as diagnostic performance comparable with that of a speed class 400 conventional screen-film system, with a dose reduction of approximately 50%.
| ACKNOWLEDGMENTS |
|---|
| REFERENCES |
|---|
| Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
|---|