| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Departments of Radiology (J.T.D., E.S., H.G.C., A.H.B., C.E.F., C.E.R.) and Biomedical Engineering (J.T.D., E.S., R.J.W., A.H.B., C.E.F.), Duke University Medical Center, 161F Bryan Research Bldg, Research Drive, DUMC Box 3302, Durham, NC 27710. Received December 11, 2001; revision requested February 18, 2002; revision received April 1; accepted May 23. J.T.D. supported in part by grant R01 CA80490 from the National Cancer Institute. E.S. supported in part by grant R21 CA91806 from the National Cancer Institute. Address correspondence to J.T.D. (e-mail: jtd@scott.mc.duke.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
MATERIALS AND METHODS: End points for optimization included the ratio of tissue contrast to bone contrast and a figure of merit (FOM) equal to the square of the signal-to-noise ratio of tissue divided by incident exposure to the patient. Studies were conducted with both computer spectrum modeling and experimental measurement in narrow-beam and full-field exposure conditions for four tissue thicknesses (8–32 cm). Three parameters that affect spectra were considered: the atomic number (Z) of filter material (Z = 13, 26, 29, 42, 50, 56, 64, 74, and 82), kilovoltage (from 50 to 150 kVp), and filter thickness (from 0.25 to 2.00 half-value layer [HVL]).
RESULTS: Computer modeling and narrow-beam experimental data showed similar trends for the full range of parameters evaluated. Spectrum model results showed that copper filtration at 120 kVp or more was optimum for FOM. The ratio of contrasts showed a trend to be higher with higher kilovoltage and only a minor variation with filter material. Full-field experimental results, which reflect the added contribution of x-ray scatter, differed in magnitude but not trends from the narrow-beam data in all cases except the ratio of contrasts in the mediastinum.
CONCLUSION: The best performance overall, including both FOM and ratio of contrasts, was at 120 kVp with 1-HVL copper filtration (0.2 mm). With this beam spectrum and an increase in tube output (ie, milliampere seconds) of about 50%, a chest radiograph can be obtained with image quality approximately equal to that with a conventional spectrum but with about 25% less patient exposure.
© RSNA, 2002
Index terms: Flat panel detector, 60.121 • Radiography, digital, 60.121
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
The phosphor used in these detectors has not been used previously in chest radiography; therefore, the question arises of whether the standard x-ray spectrum used for chest imaging is optimum for use with these devices. Cesium iodide has been used for a long time in image intensifiers, but chest imaging with screen-film radiography has most commonly been performed with rare-earth phosphors, such as Gd2O2S. The k edges of these materials, which reflect the energy of greatest x-ray absorption efficiency, are 36, 33, and 50 keV for cesium, iodine, and gadolinium, respectively. Thus, there may be a different spectrum that would optimize the detection efficiency for cesium iodide than that for screen-film radiography.
Another reason to consider the issue of optimizing the spectrum for these digital detectors is that the recorded image intensities in the lung and dense patient regions are accessible for image processing as a result of the linear response of the detectors over a wide range of image intensities. A variety of image processing algorithms may be used to enhance the relative visibility of both lung and mediastinum structures, irrespective of the gross transmission in those regions; thus, flexibility is added in the selection of kilovoltage (5). The choice of kilovoltage with these detectors may be based on relative signal-to-noise ratio (SNR) in the various regions of the chest and overall patient dose without the concomitant constraint of regional contrast, as with screen-film radiography.
In addition to consideration of the influence of kilovoltage on image quality with the devices, it is also possible to consider the optimum beam filtration. In the United States, the amount of beam filtration required for radiographic examinations including chest imaging is usually mandated by the states to ensure that the x-ray spectra are sufficiently hardened to prevent excessive skin dose. In the state of North Carolina, for example, a chest radiography system needs to have a minimum of 2.5 mm of aluminum-equivalent intrinsic filtration and generate a beam with a half-value layer (HVL) that exceeds 3.2 mm of aluminum at 120 kVp (6). This standard is consistent with that of an international standards organization (7). Since these values are typical of x-ray tubes and collimators at these kilovoltages, chest radiography is typically performed with minimal or no added beam filtration. The benefit of beam filtration on patient dose has been demonstrated previously (8–12), however, and any determination of optimum spectrum should include consideration of both kilovoltage and added filtration.
The purpose of our study was to ascertain the optimum x-ray spectrum for chest radiography with a cesium iodide–amorphous silicon flat-panel detector, including effects from both filtration and kilovoltage.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
To optimize performance, it was necessary to select end points for measurement that reflect inherent quality in chest radiographs. Factors that are typically considered for optimized chest radiographs include increased SNR, reduced relative contrast of the bone structures (to avoid obscuring the lucent lung), and reduced patient dose. From these considerations, we selected two quantities for measurement: the ratio of tissue contrast to bone contrast (hereafter, ratio of contrasts) and a figure of merit (FOM), defined as the squared SNR for soft tissue divided by incident patient exposure.
For the ratio of contrasts of tissue to bone, the replacement bone contrast was used. The latter was defined as the contrast measured when a thickness of soft tissue was replaced by an equivalent thickness of bone, since this is the mechanism with which the contrast of ribs is produced in a chest image (the ribs are encased in surrounding soft tissue). The replacement bone contrast may be written as the following equation:
|
t,B=t is the signal intensity recorded in a region of tissue thickness T = t - t and bone thickness B = t.
The contrast of tissue relative to air was used for the tissue contrast since soft-tissue contrast in the lungs is typically a result of the differences in integral tissue thickness relative to air from point to point in the lungs, as follows:
|
t is the signal intensity recorded in a region of tissue thickness T = t + t.
The FOM was chosen to reflect the SNR performance relative to patient risk, as follows:
|
t) to noise (the SD of signal intensity values). Xinc is the x-ray exposure incident on the patient (at a prescribed distance of 150 cm from the focal spot), which was used as a surrogate measure of patient risk. For detectors with small components of additive noise, where the majority of noise is a result of x-ray quantum fluctuations, the ratio of squared SNR to exposure is essentially independent of the exposure level; thus, FOM would be expected to be a measure of performance that would not change with incident exposure. On the basis of reported measurements (3,4), the detective quantum efficiency of the detector used in the current study demonstrated little influence from exposure in the range used in these experiments. This finding indicates that this FOM should be relatively insensitive to change with exposure, as desired. The chest radiography device used for these studies was a prototype flat-panel system (GE Medical Systems, Milwaukee, Wis) with a cesium iodide–amorphous silicon detector panel with performance equivalent to that with the commercially released version of this manufacturer’s product (Revolution XQ/i). The imaging system was equipped with an antiscatter grid with a grid ratio of 13:1 and 78 lines per centimeter. For image acquisitions in which the grid was removed (narrow-beam experiments discussed later), the device was first calibrated with the grid removed to ensure the application of proper gain-correction maps.
Computer Spectrum Modeling
The computer spectrum model (DXSPEC) used in this study was originally written by Ergun et al (13) and was later modified (J.T.D.). The model has been used in other published studies (14) and is described in the Appendix. The model is used to estimate the signal intensity and variance of energy deposited in a detector and is based on published attenuation data for Compton scattering and photoelectric absorption. The model includes k-shell fluorescence and reabsorption events.
The spectrum model was used to generate values of the ratio of contrasts of tissue to bone and FOM described earlier for a range of kilovoltage, filter material, filter thickness, and background tissue thickness. Specifically, kilovoltages ranged from 50 to 150 kVp in increments of 10 kVp; filter thicknesses were evaluated at 0.0, 0.25, 0.5, 0.75, 1.0, 1.5, and 2.0 HVL for each filter material; and background tissue thicknesses of 8, 16, 24, and 32 cm were used to represent lung, mediastinum, and abdominal thicknesses for average-sized and large patients. Filter materials that were chosen for evaluation comprised elements that were feasible for use in radiography (ie, primarily available in metallic or simple compound form) and that had atomic numbers (Z) that spanned the range of available elements at fairly even intervals. The following elements were used: aluminum (Z = 13), iron (Z = 26), copper (Z = 29), molybdenum (Z = 42), tin (Z = 50), barium (Z = 56), gadolinium (Z = 64), tungsten (Z = 74), and lead (Z = 82).
To compute the differential signal intensities of bone and tissue that were needed to calculate the SNR and ratio of contrasts, incremental thicknesses of 1 mm of tissue and 1 mm of bone were used. Formulas given by the International Commission on Radiation Units and Measurements (15) were used in the model for the elemental compositions of tissue (given as the average composition of soft tissue for male and female subjects) and bone (given as the average composition of the second and sixth ribs).
The thickness of cesium iodide was equal to that in the detector system, a value which is proprietary to the manufacturer.
Narrow-Beam Validation Experiments
A series of measurements were performed to validate the computer spectrum model for a limited set of filter materials (aluminum, copper, molybdenum, tin, and lead) and background acrylic thicknesses (8, 16, 24, and 32 cm). Measurements were also made by using the filter material determined to be best in the computer simulation study at a range of kilovoltages (90, 105, 120, 135, and 150 kVp) and filter thicknesses (0.25, 0.5, 1.0, and 2.0 HVL). These experiments were performed in narrow-beam geometry without an antiscatter grid, to simulate the conditions of the computer spectrum model. For these experiments, various thicknesses of methyl methacrylate (acrylic) and bone-simulating plastic (SB3; Gammex-RMI, Middleton, Wis) were used to simulate soft tissue and bone. The exact thicknesses and elemental recipes for these materials were used to generate a computer spectrum model for these particular experimental validation studies.
The image data were acquired with a uniform acrylic slab of the desired thickness that was placed 88 cm away from the focal spot. The source-to-image distance was 180 cm. The beam was collimated to an 8.5 x 8.5-cm field of view on the surface of the slab. For each experimental condition examined, a pair of images was acquired, one with three small target objects placed at the beam entrance side of the phantom toward the center of the field of view and one without any target objects. The objects were a 25-mm-diameter acrylic disk (6.02 mm thick), a 10 x 10-mm bone-simulating plastic square (2.00 mm thick), and a 7 x 7-mm lead square (4 mm thick) that was used to measure scatter fraction. The milliampere second settings for the two images of the image pair were identical and were based on the value determined at previous exposures with the automatic exposure control of the system. Thus, the exposures detected for various experimental conditions were similar and were in a range within which the electronic noise of the detector was negligible (3). The exposure incident on the phantom was measured independently for each of the experimental conditions by using a calibrated ion chamber (model 10X5-6 ionization chamber and model 1015 x-ray monitor; Radcal, Monrovia, Calif).
A comparison of the images acquired with and those acquired without the target objects was used to evaluate the signal intensity and the noise. Pixel values were normalized between regions in each pair of images to eliminate background trends and, hence, to give a more reliable measurement. The mean and SD of pixel values in regions of interest located within and adjacent to the regions of the target objects were then measured in each image pair. The sizes of the regions of interest were 70 x 70 pixels (14 x 14 mm) for the acrylic object and the background signal intensity, 40 x 40 pixels (8 x 8 mm) for the bone-simulating plastic object, and 30 x 30 pixels (6 x 6 mm) for the lead object. The noise values in single images were deduced from those in the normalized image pairs by means of standard error propagation methods.
The SNR, FOM, and ratio of contrasts were computed from the measured values. To compare the results of the narrow-beam experiments with those of the spectrum model, the experimental results for SNR were adjusted to correct for the measured amount of scattered radiation, as follows:
|
Wide-Beam Experimental Measurements
A second set of experimental measurements was performed to ascertain what effect, if any, would be noticed on the relative performance of kilovoltage and filter if full-field exposures were made, which would include the effects of scattered radiation on contrast and SNR. These experiments were performed with an all-acrylic chest phantom designed to emulate the gross attenuation and scattering characteristics of the lung, heart or mediastinum, and subdiaphragm regions but without anatomic detail (16). The phantom was constructed with thicknesses of 8, 16, and 17 cm in the lung, heart or mediastinum, and subdiaphragm regions, respectively. Tissue- and bone-simulating target objects and lead squares that were identical to those used in the narrow-beam experiments were placed in each of two locations in the upper and lower lung regions and in two locations in the upper and lower heart or mediastinum regions of the phantom. The objects were placed on the beam entrance side of the phantom, which was placed immediately in front of the detector. All image data were acquired with the antiscatter grid in place.
Experimental conditions were identical to those in the narrow-beam experiments, except for the background acrylic thickness, which was fixed in the regions examined (lung- and mediastinum-equivalent regions). Two images were acquired for each experimental condition, one with target objects and one without. The tissue and bone signal intensities, noise, scatter fraction, SNR, FOM, and ratio of contrasts were calculated in each of the four regions by using a method that was identical to that used in the narrow-beam experiments, except that the SNR and FOM values were not adjusted for the presence of scattered radiation. Thus, these reported values represent the "real" values anticipated in chest radiographs. The data from the two lung regions and the two mediastinum regions were averaged to obtain the final results.
In addition to the numeric analyses described earlier, three posteroanterior images of an anthropomorphic chest phantom (Radiology Support Devices, Long Beach, Calif) were acquired for subjective evaluation. Images were acquired at 120 kVp with (a) no added filtration, with standard exposure levels incident on the chest phantom (equivalent to a phototimed exposure); (b) copper filtration (0.2 mm of copper at 120 kVp), with 25% less exposure to the phantom; and (c) no added filtration, with milliampere seconds reduced to achieve a 25% exposure reduction to the phantom. The phantom was supplemented with two 2.5-cm-thick acrylic slabs, one on each side of the phantom, to provide scatter radiation levels equivalent to those in clinical chest radiographs. The raw image data were postprocessed by using a typical chest radiography look-up table (sigmoid shape with log latitude similar to that with commercially available asymmetric screen-film combinations), with minor adjustments to generate images with similar global gray-scale representation. No edge enhancement or other frequency processing was applied.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
Findings with the computer spectrum model revealed definite maxima and minima for FOM as a function of filter material and kilovoltage. Figure 1 shows the results with 1 HVL of filter material (note that equivalent transmissions, as measured in units of HVL, were used to facilitate comparisons of various filter materials). A clear minimum at atomic number of about 50–60 was noted in each graph. There was a relatively broad maximum at low atomic number, with a peak in the region of copper and another peak at the highest atomic number measured. In all cases, the low atomic number peak near copper was greater than or equal to the high atomic number peak. In the neighborhood of the low atomic number peak, the best performance was obtained at 100–130 kVp for the 8-cm tissue thickness and at 140–150 kVp for the 32-cm tissue thickness. The variation with kilovoltage was relatively gradual, with a range of kilovoltages for which approximately equivalent performance was noted.
|
|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
Nonetheless, in a study such as this in which there are so many comparisons that experimental or Monte Carlo estimations for every single data point are not feasible, a semiempirical method for spectrum estimation, such as that used in the present study, seems a reasonable approach, on the basis of the excellent agreement in data trends between simulation and experiment. Despite the modest accuracy of the model for absolute measurements, the same conclusions regarding spectrum optima were reached in the limited set of experimental data as were reached in the computer simulation. Thus, we believe the conclusions after the study were unaffected by the accuracy of the computer spectrum model.
Findings in these studies show that use of a copper filter at about 120 kVp results in the best performance for chest imaging with cesium iodide–amorphous silicon detectors when FOM and the ratio of tissue-to-bone contrasts are considered simultaneously. Findings also demonstrated that use of as thick a copper filter as possible resulted in improved performance; however, there will clearly be a limit to what filter thickness can be used because of limitations of tube loading and required exposure time. For cases in which there is adequate tube current from the generator to accommodate the additional tube loading without increased exposure time, there should not be a reduction in sharpness from motion blur in the resulting images. With large patients, however, it would likely be necessary to increase the exposure time unacceptably with the larger filter thicknesses. For this reason, we anticipate that 1 HVL of filter material is probably the maximum feasible thickness, considering the range of patient thicknesses encountered and the output from a typical 80-kW generator. Furthermore, changes in FOM and ratio of contrasts were gradual beyond about 1 HVL of copper and, in consideration of tube loading, we decided that 1 HVL (0.2 mm) was the practical optimum.
To evaluate the interdependency of patient exposure, SNR, and tube loading, the experimental data can be viewed in terms of the following equation for normalized FOM (nFOM):
|
is the fractional filter transmission and FOMfilter and FOM0 are the FOM with and without, respectively, the filter in the beam. This quantity represents the change in squared SNR after the addition of filtration while milliampere seconds are constant. To maintain constant SNR with the filter in place, the milliampere seconds would have to be increased by the reciprocal of nFOM, which would result in a change in patient entrance exposure given by /nFOM. At the recommended filtration and kilovoltage (0.2 mm of copper, 120 kVp), nFOM is about 0.66 (0.633 in the lungs and 0.668 in the mediastinum region). Thus, with the recommended spectrum, the milliampere seconds would need to be increased by 52% (ie, 0.66-1, or equivalently 1/0.66) to produce an image with SNR comparable to that obtained with a conventional spectrum, which would result in a decrease in patient exposure of 24% (ie, 0.5/0.66, because = 0.5 for 1 HVL of copper). Or equivalently, a doubling of the milliampere seconds with the optimum spectrum in place would yield patient exposure comparable to that with the conventional spectrum but with squared SNR improved by 27% in the lungs and 34% in the dense patient regions. It should be noted that these increases in squared SNR would correspond to increases in SNR of 13%–16%, which are small enough that they might be difficult to appreciate visually on the images in the presence of anatomic background (17). As a result, we conclude that the most likely use of the optimum filtration will be to reduce patient exposure rather than to improve SNR. This reduction in exposure will likely become increasingly important as advanced applications such as dual-energy imaging and digital tomosynthesis become widely available. Exposure saved with the optimum spectrum might be used for the extra images required with the advanced applications without increasing the total exposure from the entire chest examination.
In regard to estimating patient risk, it is important to select a measure of risk that can be determined easily for each combination of kilovoltage and filter material. It would be desirable, though not feasible, to measure the actual effective dose equivalent for each spectrum under consideration. Measurements of SNR versus estimated dose equivalents were used in a previous study of optimum kilovoltage for use with computed radiography, but the filtration remained the same for each kilovoltage (18). In the present study, in which both kilovoltage and filter material were changing, it was not possible to estimate the effective dose equivalent for each of these beams. Owing to the difficulty of estimating absorbed dose theoretically or of measuring it experimentally, we chose a more practical solution by using patient entrance exposure as the surrogate measure of patient risk for each of the spectra considered. However, it should be noted that this measure, while reasonable, is not the same as measurement of the response of the system as a function of absorbed dose or effective dose equivalent.
The current study was focused on only the intrinsic image quality characteristics of chest radiography. In digital radiography, postprocessing of the image data can have a substantial effect on image impression. Changes in the beam quality to optimize intrinsic image characteristics, as addressed in this study, may require a modification of the image postprocessing parameters to make the gray-scale appearance of the images equivalent to that in images that are now acquired in the clinic. With that modification, the standard for optimized chest imaging with the spectrum determined in this study could be practiced with the cesium iodide–amorphous silicon flat-panel detectors currently available.
Practical application: Findings in this study demonstrate that the optimum x-ray spectrum for chest radiography with cesium iodide–amorphous silicon flat-panel detectors is 120 kVp and 0.2 mm of copper. Implementation of the optimum spectrum in clinical chest imaging should permit a reduction in patient entrance exposure of approximately 25% with image quality consistent with the current radiography standard. The dose saved will have increasing clinical importance as further developments occur in the field of advanced applications aimed at improving diagnosis with flat-panel detectors.
|
APPENDIX |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
The signal intensity measured for this study was considered to be the estimated energy deposited in a layer of cesium iodide, which represented the detector. The estimated deposited energy included the effects of photoelectric absorption and absorption of k fluorescence, average energy deposited per Compton event, and a cascade of these effects within the thickness of cesium iodide. The relative detector thickness available for these interactions was chosen to reflect a reasonable physical model of the interactions, that is, the k-fluorescence events were considered to be emitted isotropically, and the Compton-scattered photons emitted within the detector were considered to be emitted at the average Compton angle. The detector was subdivided into three layers, and the photoelectric, k-fluorescence, and Compton events were all considered to occur at the midpoint of each of these layers. The energy absorbed in each layer was computed, as well as the number of subsequent absorptions for the k-fluoresced and Compton-scattered photons that were expected to occur from each layer to the edge of the detector. The composite absorbed energy was computed as the sum of the initial and secondary absorption events for each of the layers.
An estimate of the noise measured in the signal intensity was added to the computer spectrum model (J.T.D.). The estimate of noise was computed as the summed energy-weighted variance associated with each energy bin from all the primary and secondary absorption events described earlier. The initial x-ray tube output spectrum, including both bremsstrahlung and characteristic photon intensities, was modified (J.T.D.) to include empiric tube spectra published by Boone and Seibert (19). The inherent filtration of the spectrum model was adjusted until it produced HVLs equivalent to those measured for the actual x-ray system used in these experiments.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: FOM = figure of merit, HVL = half-value layer, SNR = signal-to-noise ratio
Author contributions: Guarantors of integrity of entire study, J.T.D., E.S., H.G.C.; study concepts, all authors; study design, J.T.D., E.S., H.G.C., R.J.W., A.H.B., C.E.F.; literature research, J.T.D., E.S., H.G.C.; experimental studies, E.S., H.G.C., R.J.W., A.H.B., C.E.F.; data acquisition, J.T.D., E.S., H.G.C., R.J.W., A.H.B., C.E.F.; data analysis/interpretation, J.T.D., E.S., H.G.C., C.E.R.; manuscript preparation, J.T.D., E.S., H.G.C.; manuscript definition of intellectual content, all authors; manuscript editing, J.T.D., E.S., H.G.C.; manuscript revision/review and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H Cole and C P Lawinski Developments in radiography and fluoroscopy Imaging, June 1, 2006; 18(2): 75 - 81. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. W. Hamer, C. B. Sirlin, M. Strotzer, I. Borisch, N. Zorger, S. Feuerbach, and M. Volk Chest Radiography with a Flat-Panel Detector: Image Quality with Dose Reduction after Copper Filtration Radiology, November 1, 2005; 237(2): 691 - 700. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Samei, J. T. Dobbins III, J. Y. Lo, and M. P. Tornai A framework for optimising the radiographic technique in digital X-ray imaging Radiat Prot Dosimetry, May 17, 2005; 114(1-3): 220 - 229. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Bath, M. Hakansson, J. Hansson, and L. G. Mansson A conceptual optimisation strategy for radiography in a digital environment Radiat Prot Dosimetry, May 17, 2005; 114(1-3): 230 - 235. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Doyle, C. J. Martin, and D. Gentle Dose-image quality optimisation in digital chest radiography Radiat Prot Dosimetry, May 17, 2005; 114(1-3): 269 - 272. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I D Honey, A Mackenzie, and D S Evans Investigation of optimum energies for chest imaging using film-screen and computed radiography Br. J. Radiol., May 1, 2005; 78(929): 422 - 427. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Uffmann, U. Neitzel, M. Prokop, N. Kabalan, M. Weber, C. J. Herold, and C. Schaefer-Prokop Flat-Panel-Detector Chest Radiography: Effect of Tube Voltage on Image Quality Radiology, May 1, 2005; 235(2): 642 - 650. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
