| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, University of Cologne Medical School, University of
Cologne, Kerpenerstr. 62, 50924 Cologne, Germany.
2
Philips Medical Systems, Röntgenstr. 24, 22335
Hamburg, Germany.
3
Department of Medical Statistics, Informatics and Epidemiology, University of
Cologne Medical School, 50924 Cologne, Germany.
4
Department of Trauma, Hand and Reconstructive Surgery, University of Cologne
Medical School, 50924 Cologne, Germany.
Received November 7, 2000;
accepted after revision June 22, 2001.
Supported by a grant from the research funding program
KÖLN FORTUNE of the University of Cologne.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. One hundred bones taken from domestic pigs (50 ribs and 50 femurs) were divided into two equal groups. Fissures and bone lesions were created in 50 bones and 50 served as controls. The bones were examined using film-screen systems, digital luminescence radiography, and digital selenium radiography at various doses. Digital selenium radiography exposure values were adapted to the image geometry differing from the reference methods with a detector focus distance of 2.15 m. Four radiologists independently evaluated image quality and detectability of fissures and lesions on a five-point scale of confidence. Statistical evaluation was based on receiver operating characteristic curve analysis.
RESULTS. Fissures and bone lesions were detected most reliably using the mammography film-screen system, but the difference in the results of the analog and digital reference images did not achieve statistical significance.
CONCLUSION. Compared with analog film-screen systems, the lower spatial resolution of the digital selenium and digital luminescence radiography systems does not affect detectability of fissures and bone lesions in porcine bone. Selenium is effective in skeletal radiography for detecting fissures and bone lesions. With digital selenium and digital luminescence radiography, the surface dose can be cut to half that required for 200-speed film-screen systems without losing any diagnostically relevant information.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Analog film-screen systems have a spatial resolution of up to 15 line pairs per millimeter (lp/mm). However, they are limited by a narrow, diagnostically usable exposure range that is characterized by the linear portion of the optical density curve. This narrow range can result in false exposures [5].
With regard to the maximum spatial resolution mathematically possible, the digital radiography systems available thus far have proven inferior to analog film-screen systems [5]. Storage phosphor radiography was introduced in the early 1980s [6]. The system is cassette-based and therefore in contrast to the existing systems for digital selenium radiography compatible with existing radiographic equipment. Instead of conventional film-screen combinations, storage phosphor radiography uses a photostimulable phosphor screen as the image receptor. The maximum spatial resolution of digital luminescence radiography mathematically possible ranges between 2.5 and 5 lp/mm.
Selenium radiography is a new digital detection method for chest imaging. This method eliminates the noise sources associated with multiple conversions by (in contrast to digital luminescence radiography) directly converting X-ray quanta into carriers of electric charges and by virtue of the fact that the selenium is in an amorphous state [7,8,9]. The literature contains a plethora of reports on the possible and potential uses of digital selenium radiography in chest imaging [10,11,12,13].
The digital selenium radiography system (Thoravision; Philips Medical Systems, Hamburg, Germany) we investigated has a mathematic spatial resolution of 2.8 lp/mm, which is less than that of analog film-screen systems. This fact begs the question, will the lower spatial resolution compared with analog film-screen systems lead to a diagnostically relevant loss of information in bone radiography, or can such a loss be compensated by improved quantum efficiency and dynamic range?
We attempted to answer this question by comparing the digital detector system Thoravision with various film-screen systems and digital luminescence radiographic methods. The test applications involved the imaging of cortical and cancellous bone and the detection of artificially created fissures and lesions in bones removed from domestic pigs.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
|
|
Digital Selenium Radiography
The Thoravision digital chest radiography unit uses a selenium-coated drum
as the radiographic detector. The selenium surface is positively charged
before imaging. The charge of the core of the metal drum is negative. Selenium
is an element that acts as a photosemiconductor. In other words, because the
action of the X rays produces an electric conductivity proportional to the
radiation intensity, an equilibrium of local charges occurs during X-ray
exposure, thereby creating a charged pattern proportional to the dose on the
selenium surface. After exposure, the electric charges are measured. The
resulting signal is converted from analog to digital and transferred to the
image processor
[7,8,9].
We used the following geometric parameters for selenium radiographic imaging: film focus distance, 215 cm; maximum image size, 43 x 49 cm; matrix size, 2166 x 2488 pixels; and pixel size, 0.2 mm. In the Thoravision system, analog signals are converted to digital values with a precision of 14 bits. To avoid rounding errors, we performed calculations with 15 bits, which are stored in 2 bytes. Before starting the study, the device's image processing algorithms were optimized to the reproduction of bone with heightened fine-detail contrast and noise suppression. The selected settings were not changed throughout the study. The digitalized image was printed on a laser printer with a reproduction matrix of 4256 x 5174 pixels (Matrix LR 3300; Agfa Medical Division, Cologne, Germany; film: Scopix LT 2B Dayl.-A1, Agfa) and was stored on an optical storage plate.
Digital Luminescence Radiography
The technical and physical principles of digital luminescence radiography
are known [5,
6,
14]. The PCR AC3 (Philips
Medical Systems) was used as the detector system; it has an imaging matrix of
1770 x 2370 pixels at an image depth of 12 bits. Cassettes with a format
of 18 x 24 cm were used, the pixel size of the luminescence radiographic
films was 0.1 mm, and the nominal spatial resolution was 5 lp/mm. The pictures
were documented on a laser imager (Imation Dryview 8700; Kodak Health Imaging,
Stuttgart, Germany; film size, 35 x 43 cm; pixel size, 78 mm) with a
reproduction matrix of 4620 x 5596 pixels. The postprocessing of the
digitalized images was optimized to the visualization of long bones and
performed in the same fashion throughout the study.
Analog Film-Screen Systems
The analog images were produced with 200- and 400-speed film-screen
combinations (film: Curix HT 1000 G Plus SF, Agfa; screens: Ortho medium and
regular, Agfa) and with mammography film (Min-R 2000; Kodak Health Imaging).
The spatial resolution was 8 lp/mm for the 200-speed film-screen system, 6
lp/mm for the 400-speed film-screen system, and 15 lp/mm for mammography
film.
The following exposure parameters were used for analog and digital images for ribs: mammography film, 48 kV and 7.1 mAs; 200-speed film-screen system, 43 kV and 4 mAs; 400-speed film-screen system, 43 kV and 2.5 mAs; digital luminescence radiography, 43 kV and 4 mAs; and digital selenium radiography, 44 kV and 16 mAs. The parameters for femurs were mammography film, 56 kV and 16 mAs; 200-speed film-screen system, 49 kV and 8 mAs; 400-speed film-screen system, 49 kV and 5 mAs; digital luminescence radiography, 49 kV and 8 mAs; and digital selenium radiography, 48 kV and 32 mAs. The film focus distance was 1.05 m for the analog images and digital luminescence radiography. The selenium radiographic exposure values were adapted to the focus detector distance, which was 2.15 m. Additionally, low-dose images were prepared for digital luminescence and digital selenium radiography in which the detector dose corresponded to the values used in the 400-speed film-screen system.
Evaluation
Analog and digital images were separated and made anonymous. The resulting
700 pictures were randomly coded and given to four independent reviewers for
interpretation (one specialist in radiology, two residents in radiology in
their fifth year of training, and one resident in emergency surgery in his
fourth year of training). The reviewers interpreted the radiographs directly
on the viewboxes under standard conditions without using any additional
illumination lamps. The following rating scale was used for assessing the
image quality for visualizing cortical and cancellous bone: 1, well visualized
in all sections; 2, well visualized in some sections; 3, sufficiently
visualized; 4, inadequately visualized; and 5, not assessable.
The four reviewers assessed the localization of the fissures and bone lesions artificially created in the animal specimens according to the following criteria: 1, definite fissure or cortical lesion; 2, probable fissure or cortical lesion; 3, uncertain; 4, probably not a fissure or cortical lesion; and 5, definitely not a fissure or cortical lesion. The reviewers were also required to state the number of definitely and probably distinguishable bone lesions.
Statistical Methods
The image quality of cortical and cancellous bone as assessed by each
reviewer was analyzed separately using paired t tests for the mean. A
global assessment was obtained by averaging the four results for each bone.
Receiver operating characteristic (ROC) analyses were performed to compare the
diagnostic accuracy of the imaging techniques. A total of 112 ROC curves were
constructed for each of the four reviewers, seven techniques, two types of
bone, and fissure or osteolyses. The Mean row in Tables
2,3,4,5
corresponds to the area under the ROC curve of the averaged ratings of the
four reviewers.
|
|
|
|
A nonparametric approach described by DeLong et al. [15] was used for paired comparison of the correlated areas under the ROC curves derived from the seven techniques. A computer code implementing the approach of DeLong et al. was downloaded from the Internet [16]. A p value of less than 0.05 was considered significant [15].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Image quality did not score more poorly when 50% of the surface dose of a 200-speed film-screen system was used. Scores for digital luminescence radiography of cortical bone were ribs, 2.09 and femurs, 1.86; and of cancellous bone were ribs, 2.22 and femurs, 2.09. Scores for digital selenium radiography of cortical bone were ribs, 2.10 and femurs, 2.05; and of cancellous bone were ribs, 2.39 and femurs, 2.14.
Evaluation of Fissures
Tables 2 and
3 show the areas under the ROC
curves with respect to fissures for each individual reviewer and averaged
across all reviewers according to the individual radiology methods. No
statistically significant difference was seen in diagnostic performance on the
basis of the image quality of the individual examination methods for either
ribs or femurs.
Evaluation of Bone lesions
Tables 4 and
5 show the areas under the ROC
curves with respect to lesions for the individual reviewers and averaged
across all reviewers for the individual radiology methods. Figure
2A,2B,2C,2D
presents the ROC curves generated for the individual reviewers for
interpretation of hind leg bone lesions. No statistically significant
difference was seen among the diagnostic reliability of the individual
examination methods in detecting bone lesions when all reviewers were
considered together and separately.
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Murphey et al. [4] carried out studies on digitalized skeletal images and proved that a pixel size of at least 0.16 mm, corresponding to a nominal spatial resolution of 2.8 lp/mm, was necessary to detect nondisplaced fractures. The spatial resolution requirements for detection of subperiosteal lesions are higher—that is, a pixel size of at least 0.1 mm, corresponding to a nominal spatial resolution of 5.7 lp/mm [3, 4]. By contrast, a more recent article by Murphey et al. [21], who used a primary digitalizing device, established that a spatial resolution of 2.5-5 lp/mm, corresponding to a pixel size of 0.2-0.1 mm, is sufficient for detection in musculoskeletal imaging.
This finding has several explanations. First, the geometric resolution was measured on lead grids that, as high-contrast phantoms, are not suitable for imaging of osseous details that do not contrast well with their surroundings [22]. Second, it was shown that the signal-to-noise ratio is a more decisive parameter for the correct detection of poor-contrast objects [22]. From the literature it is known that the signal-to-noise ratio for digital selenium radiography is best among all methods compared in our study [8, 11]. The requirements for comparatively high spatial resolution in diagnostic skeletal radiography have been established in part by studies on digitalized radiographic projections [3, 4]. However, secondary digitalization can store only the information that was originally present on the analog image. The dynamic range of analog projection-based radiographic imaging is markedly narrower than all the digital reference methods. In other words, digitalized images do not use the good contrast resolution available with direct digital techniques.
In the 1990s, many authors compared the diagnostic merits of digital luminescence radiography with those of analog film-screen systems and ranked them as comparable [2, 23,24,25,26]. The indications studied spanned the fields of traumatology, rheumatology, and oncology, to name a few. Our data agree with and underscore the results of previous authors who rated the diagnostic performance of digital luminescence radiography for detecting bone lesions and fissures as equivalent or superior to analog film-screen methods [2, 23, 25].
The advantage of digital selenium radiography compared with analog film-screen systems and digital luminescence radiography is that no intermediate luminescence is required. Therefore, the noise sources associated with multiple conversions are eliminated. At least theoretically, this means that the signal-to-noise distance is also greater than with digital luminescence radiography [11]. This might be one way to improve the detail recognition that is so important in the detection of small, weakly opaque structures such as bone lesions and fissures, and might provide the opportunity for dose reduction compared with film-screen systems. Background noise is likewise reduced by the amorphous state of the selenium in comparison with microcrystalline phosphor layers, thereby achieving better imaging detail. In contrast to the conditions prevailing on amplifying screens, the thickness of the selenium layer has no effect on spatial resolution. Although the light quanta produced by film-screen systems and digital luminescence radiography are distributed through the layer and form "confusion areas," the strong electric field in selenium ensures vertical transport of the charge carrier to the selenium surface. Hence, the selenium layer itself can be regarded as a virtually perfect X-ray detector that has no adverse effects on the image signal, nor does it create any external noise in the system [8]. Digital selenium radiography allows more enhanced discrimination among subtle contrast differences and thus lends itself well to imaging of the human body.
One limitation of the Thoravision selenium system compared with the analog film-screen systems and the luminescence radiographic device we tested was its lower nominal spatial resolution (i.e., 2.8 lp/mm) [7, 8]. Nonetheless, our results show that despite a lower spatial resolution, the diagnostic reliability of digital selenium radiography is comparable with that of both digital luminescence radiography and analog film-screen systems in the detection of fissures and bone lesions in the rib and hind leg bones of domestic pigs. In a recently published work, Ludwig et al. [27] reached the same conclusion, but the detectability of only bone lesions was tested. The main difference between our study and the study by Ludwig et al. is that we additionally created fissures similar to actual fractures in a clinical setting. We also think that our bone lesions are quite similar to osteolysis in patients because the lesions in our study were created by wood screws and not by a drill. The edges of bone lesions created by a drill are much sharper and, in our opinion, not comparable to clinical osteolysis.
Although the image quality of cancellous and cortical bone in the anatomic specimens studied in images taken by the mammography method was significantly better than in all other projection methods, no diagnostic benefit was gained. Fissures and bone lesions were detected most reliably using the mammography method, but the difference from the results of the analog and digital reference images did not achieve statistical significance. By contrast, the surface dose used for mammographic imaging was higher than the dose with digital selenium and digital luminescence radiography by a factor of 3. With digital selenium and digital luminescence radiography, it was possible to cut the surface dose to the half that was required for 200-speed film-screen system without losing any diagnostically relevant information. This observation can be explained by the large dynamic range of digital imaging procedures that produces good image quality in a large exposure range. Because of the broad, linear gradation curve, there is virtually no more need for repeated examinations because of poor exposure. This fact also contributes to reducing radiation exposure in general.
In ROC analysis, areas under the ROC curves of 0.75-0.80 are assumed to provide adequate diagnostic difficulty [28]. In our study, this fact applied only to the bone lesions on the porcine femora bones. This fact might lead one to conclude that although fissures and osteolyses involve the rendering of subtle findings, the detectability was too simple. On the other hand, this observation might be explained by the good methodologic ability of all imaging techniques to detect subtle changes.
Limitations of the kind of study we are presenting might be that, as in other studies, in our experimental setting it is not possible to differentiate exactly between cortical and cancellous bone lesions. Additionally, it remains to be proven whether the results of our experimental bone study are transferable in a clinical setting to patients with bone lesions, fissures, and osteolysis that are sometimes hard to detect because of the presence of soft tissue and bowel gas. Especially, the possibility of reducing the exposure dose has to be proven in further clinical studies.
In conclusion, compared with analog film-screen systems, the lower spatial resolution of the digital selenium and digital luminescence radiography systems used in our study does not affect detectability of fissures and bone lesions in porcine bone. Selenium seems to be effective in skeletal radiography. In our experimental setting with digital selenium and digital luminescence radiography, the surface dose could be cut to half that required for 200-speed film-screen systems without losing any diagnostically relevant information.
Acknowledgments
We thank J. Koebke, Center of Anatomy, University of Cologne, and T.
Zähringer, Department of Radiology, University
of Cologne, for their support in preparing the animal specimens and producing
the radiographic images.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
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]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
