| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Thoracic Imaging |
1 From the Department of Diagnostic Radiology, University Hospital of Regensburg, Franz-Josef-Strauss-Allee 11, 93042 Regensburg, Germany (O.W.H., I.B., N.Z., S.F.); Department of Radiology, UCSD Medical Center, San Diego, Calif (C.B.S.); Department of Radiology, Hospital Hohe Warte, Bayreuth, Germany (M.S.); and Department of Radiology, Ulm Army Hospital, Ulm/Donau, Germany (M.V.). From the 2004 RSNA Annual Meeting. Received October 9, 2004; revision requested December 21; revision received February 6, 2005; accepted March 7. Address correspondence to O.W.H. (e-mail: o.hamer{at}gmx.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: The study was approved by the institutional ethics committee. All prospectively enrolled patients signed the written consent form. Chest radiographs in two projections were acquired at 125-kVp tube voltage and 2.5-µGy detector entrance dose. The experimental group (38 patients) was imaged with 0.3-mm copper filtration; the control group (38 patients) was imaged without copper filtration. An additional 12 patients were imaged with and without copper filtration and served as paired subject-controls. Three readers blinded to group and clinical data independently evaluated the radiographs for image quality on a digital display system. Twelve variables (six for each radiographic projection) were assigned scores on a seven-point ordinal scale. Scores between experimental and control groups were compared: Logistic regression analysis and Mann-Whitney U test were used for unpaired patients; and Wilcoxon and McNemar test, for paired patients. In all, 72 comparisons were determined (36 [12 variables x three readers] for unpaired patients and 36 for paired patients). In a phantom study, radiation burden of experimental protocol was compared with that of control protocol by using Monte Carlo calculations.
RESULTS: For 70 of 72 comparisons, digital radiographs obtained with copper filtration were of similar image quality as radiographs obtained without copper filtration (P = .123 to P > .99). For two of 72 comparisons, one observer judged the experimental protocol superior to the control protocol (P = .043, P = .046). Patient dose reduction estimated with Monte Carlo calculations was 31%. Use of copper filtration increased exposure times by 48% for posteroanterior views and by 34% for lateral views.
CONCLUSION: Subjectively equivalent chest radiographic image quality was found with estimated 30% dose reduction after addition of 0.3-mm copper filtration with flat-panel cesium iodide–amorphous silicon technology.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The development of digital flat-panel detectors that are based on cesium iodide and amorphous silicon may offer a more promising solution to this problem. This technology permits wide exposure latitude, high contrast resolution, and high detective quantum efficiency. It is widely considered superior to both screen-film and storage phosphor radiography (9–23), because it provides high image quality even when patient exposure is low.
We hypothesized that high-quality, low-dose chest radiographs could be obtained by using a cesium iodide–amorphous silicon system with beam filtration without a reduction in tube voltage. Thus, the purpose of our study was to compare image quality and estimated dose for chest radiographs obtained by using a cesium iodide–amorphous silicon flat-panel detector at fixed tube voltage and detector entrance dose with and without additional 0.3-mm copper filtration.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
In February of 2003, we established a new protocol for chest radiography: tube voltage, 125 kVp; detector entrance dose, 2.5 µGy; and additional copper filtration, 0.3 mm. This protocol was adopted only for clinical reasons. For our study, use of copper filtration was considered the experimental protocol, whereas the same technique without copper filtration was considered the control protocol. Both protocols met national German guidelines approved for adult chest radiography.
Control Protocol: Prospective Analysis
On three consecutive days in July 2003, we randomly selected 50 ambulatory patients who were referred for chest radiography and recruited them to serve as the control study group. The only inclusion criteria were ambulatory status, age older than 18 years, and referral for chest radiography. There were no exclusion criteria. The randomly selected patients were asked by a clinical coordinator to participate in the study on their arrival at the radiology department. Patients were fully informed and gave written consent on a form approved by the institutional ethics committee.
Experimental Protocol: Retrospective Analysis
With our radiology information system (software version 7.3; Medos, Langenselbold, Germany), we determined that 12 patients who had undergone chest radiography with the control protocol also had undergone radiography at another time during the study period with the experimental protocol. These 12 patients were chosen for the experimental study group but also served as patients with paired control studies. Thus, these 12 were also part of the control group. To complete the experimental study group, we randomly selected 38 additional patients who had undergone chest radiography with the experimental protocol during the study period. The only inclusion criteria for these 38 patients were ambulatory status and age older than 18 years. There were no exclusion criteria. As already noted, the ethics committee waived consent for this retrospective component of our study.
Patient Data
One of the authors (O.W.H.) reviewed patients' medical records and entered each patient's age, sex, weight, and height in a computerized database. Body mass index was calculated as weight in kilograms divided by height in square meters. Other clinical factors that may potentially influence chest radiographic quality (mastectomy, breast implants, median sternotomy, and other chest wall abnormalities) also were recorded.
Image Acquisition
Patients were imaged in standard posteroanterior and left lateral projections. Thus, a total of 100 radiographic examinations, comprising a total of 200 radiographs, were performed for this study: 50 sets of posteroanterior and lateral radiographs in the experimental group and 50 sets in the control group.
The radiographic system (Axiom Aristos; Siemens Medical Solutions, Forchheim, Germany) consisted of an x-ray tube (focal spot size, 0.6), a wall stand, and a cesium iodide–amorphous silicon detector (Pixium 4600; Trixell, Moirans, France) mounted behind a phototimer sensor (Iontomat; Siemens Medical Solutions) and a dedicated, stationary antiscatter grid (80 lines per centimeter; ratio, 15:1). Matrix size was 3000 x 3000 pixels, with a pixel pitch of 143 µm, which led to an active imaging area of 43 x 43 cm. The theoretical spatial resolution limit was 3.5 line pairs per millimeter. A dose-area measuring chamber (KermaX-plus; Wellhoefer, Schwarzenbruck, Germany) was integrated into the x-ray collimator.
The inherent filtration provided by the tube housing, collimator, dose-area measuring chamber, and light localizer was equivalent to 2.9 mm of aluminum. Control group radiographs were acquired without additional copper filtration. For the experimental group, a 0.3-mm-thick copper filter was integrated in the tube housing on the tube side of the dose-area measuring chamber. Thus, filtration in the experimental group included the filtration in the control group plus 0.3 mm of additional copper filtration.
Focus to image receptor distance was 180 cm. Tube voltage was fixed at 125 kVp. By using the phototimer, detector entrance dose was set at 2.5 µGy, the dose that would be applied with a 400-speed screen-film system. At a tube voltage of 125 kVp, the influence of an additional filtration of 0.3 mm of copper on the phototimer sensor cutoff dose is negligible (within 5%, data provided by the manufacturer). The peak tube current was 400 mA. The dose-area product, the tube current–time product, and exposure time were automatically recorded on the Digital Imaging and Communications in Medicine header of each image. One of the authors (O.W.H.) manually entered the dosimetry data into a spreadsheet. The frequency with which exposure time exceeded 40 msec (the upper limit recommended by the American College of Radiology [24]) was determined for the experimental and the control groups.
Image Processing
In the cesium iodide–amorphous silicon detector, a 500-µm-thick layer of thallium-doped cesium iodide was implemented for conversion of x-rays to light. The impinging light was converted to an electric charge in the photodiodes of the amorphous silicon matrix. The charge was read out with dedicated electronics and then converted to a digital signal with 14-bit resolution (16 384 gray levels).
Images acquired with and without copper filtration underwent digital image processing by using the algorithm routinely used for chest radiography in our institution: The first step of digital image processing was a flat-field and dark-current correction at a single-pixel level. This processing step is performed automatically and cannot be controlled by the user. The next step was the so-called signal normalization. The purpose of signal normalization was to assign normalized values, which are independent of detector entrance dose settings, to the pixel values measured behind the phototimer sensors. For the present study, signal normalization for both imaging protocols was performed according to the identical fixed factor. In the third processing step, organ-specific contrast and latitude optimization was performed by applying a nonlinear gradation curve (lookup table). The lookup table was identical for both imaging protocols. This gradation processing reduced the image resolution from 14 bits to 12 bits (4096 gray levels). The fourth processing step was multiband frequency processing. Two spatial filters were applied to compress the latitude (so-called harmonization) and to enhance detail contrast and sharpness of fine structures (so-called detail or edge enhancement). Each filter was specified by kernel size and gain factor. Filters, gain factors, and kernel sizes were identical for both imaging protocols. The fifth processing step was windowing for adjustment of contrast and brightness. This was performed on the basis of a histogram analysis by using an autowindowing algorithm. This algorithm calculated the optimum window width and center to be used by the viewing software for display of the final image. Window width was set to be as small as possible but without clipping relevant gray values. The window center was exactly in the middle of the window width. The digital image processing system used in the present study is vendor specific but similar to other systems. All processing parameters were part of the routinely used preconfigured organ-specific program and were identical for both imaging protocols. After exposure release, all images were automatically processed without manual intervention.
The resulting processed images were transferred to a workstation (Magic-View; Siemens Medical Solutions) equipped with a diagnostic monitor (Simomed HM, 21 inch; Siemens Medical Solutions) with a resolution of 3000 x 3000 pixels and a maximum screen white luminance of 260 candela per square meter.
Review of all radiographs for this study was performed by using this workstation and monitor in full-resolution mode (no interpolation or down-sampling) at constant room illumination. The contrast ratio of the monitor, which included the potential influence of ambient light, was greater than 100:1, which meets the guidelines of the German regulatory board.
Diagnostic Interpretation of Radiographs
A radiologist (O.W.H., the diagnostic radiologist, with 4 years of experience in chest radiographic interpretation) reviewed the images from 100 chest radiographic examinations for diagnostic interpretation. The diagnostic radiologist was blinded to whether the examinations were performed with or without additional filtration. The clinical history and, when available, findings from additional examinations, such as computed tomography, were considered in the diagnostic interpretation. On the basis of this review, the radiologist documented all normal and abnormal radiographic findings in the study examinations.
Assignment of Scores for Image Quality
After completion of the diagnostic radiologist's interpretation, three other radiologists, the readers (M.V., N.Z., I.B.) independently reviewed the images from the 100 radiographic examinations. The posteroanterior and lateral images were reviewed together. The examinations, each of which included a posteroanterior view and a lateral view, were presented in random order. The three readers included a board-certified radiologist (M.V., with 7 years of experience in general radiology) and two residents (I.B. and N.Z., each with 4 years of experience in general radiology). Window width and window level settings were manually adjusted as necessary by the readers on a patient-by-patient basis. The readers were free to vary their viewing distance on the basis of individual preferences. A time limit for image review was not imposed.
Each reader assigned a score to the images without knowledge of clinical history, diagnostic interpretation, group status, or the other readers' scores. Image quality was assessed subjectively by using a scoring system modified from the European Guidelines on Quality Criteria for Diagnostic Radiographic Images and similar to other published scoring systems for chest radiographic assessment (17,18,22,23,25–29). The scoring system required that only normal structures be assessed. For example, in patients with circumscribed pulmonary consolidation, only nonconsolidated lung was evaluated. To ensure that the three readers evaluated the same anatomic regions and avoided pathologically affected areas, the diagnostic radiologist attended the evaluations. The diagnostic radiologist pointed out areas of pathologic abnormalities but communicated no other information.
Scoring System
For each posteroanterior and lateral radiograph, six separate variables were assigned scores, as summarized in Table 1. The first five variables represented different sets of anatomic structures: The score reflected visualization of structure borders and intrinsic details. The sixth variable was global image quality: The score reflected overall impression of image contrast, noise, and motion blur. Thus, 12 variables were assigned scores per examination by each reader, six for the posteroanterior projection and six for the lateral projection.
|
In addition, for the paired patients, the seven-point scores were dichotomized into a binary scale: 1, excellent (score of 1 on the seven-point scale), or 2, less than excellent (score of 2–7 on the seven-point scale). Dichotomization was performed to permit statistical analysis of the paired samples by using the McNemar test, as will be discussed later.
The observed frequency of each possible ordinal (scores 1–7) and dichotomized (binary scores 1 and 2) score for each reader and for each of the 12 variables (six for the posteroanterior view and six for the lateral view) was tabulated. The mean frequency (±the standard deviation) for each score for each reader was calculated separately for the unpaired and paired patients in the experimental and control groups.
Statistical Analysis
Interobserver agreement.—Interobserver agreement was analyzed by using the weighted 
test for multiple observers.
Statistical comparisons for unpaired patients.—Statistical comparisons for the unpaired patients (38 vs 38) were performed by using (a) unpaired Student t test for continuous data (age, body mass index, dosimetry), (b) ordinal logistic regression for ordinal data (seven-point image quality scores), (c) Mann-Whitney U test for ordinal data (seven-point image quality scores) if logistic regression was not applicable because one or more variables were constant, and (d) 
2 or Fisher exact tests (depending on expected frequencies) for nominal data (sex, presence of pathologic findings, frequency of exposure times that exceeded the threshold recommended by the American College of Radiology).
Statistical comparisons for paired patients.—Statistical comparisons for the paired patients (12 vs 12) were performed by using the (a) paired Student t test for continuous data (dosimetry), (b) Wilcoxon signed rank test for ordinal data (seven-point image quality scores), and (c) McNemar test for dichotomous variables (dichotomized two-point image quality scores and frequency of exposure times that exceeded the American College of Radiology threshold).
In regard to the analysis of subjective image quality, differences in observed frequencies of scores between groups were assessed statistically by using nonparametric tests as noted previously. The 12 variables that were assigned scores were treated as independent and were analyzed separately for the three readers. Thus, 12 independent comparisons were determined for each reader (one for each variable that was assigned a score). Because there were three readers, a total of 36 independent comparisons were analyzed (12 comparisons per reader x 3 readers). The mean observed frequencies (averaged over the three readers) were reported but were not used for statistical comparisons.
A two-tailed P value of .05 or less was considered to represent a statistically significant difference. Because the authors sought to demonstrate statistical equivalency between the experimental and control protocols, the authors did not correct for multiple comparisons. This was a conservative decision that maximized the probability of finding statistical advantages of the control protocol over the experimental protocol.
Power analysis.—We retrospectively calculated the power at the 
= .05 level to detect a clinically meaningful difference in image quality, which was defined retrospectively as an absolute 25 percentage point difference in frequency of excellent scores (score of 1) for the experimental group versus the control group. The power calculations were repeated for each observed frequency in the control group. In the unpaired sample, this was achieved by using the method introduced by Whitehead (30). In the paired sample, this was achieved by treating the ordinal data as continuous and by calculating the power of a Student t test (as a signed rank test approximation). We assumed actual sample sizes. Because 36 comparisons were made (between each of the 12 variables evaluated by each of the three readers), we also estimated the power to detect clinically meaningful differences in image quality in at least six of the 36 comparisons. This last estimation was performed by assuming that the power of each of the 36 observations was independent; a conservative decision was determined to fix the power of each comparison as the lowest observed power.
All statistics except the weighted and power analyses were calculated by using statistical software (SPSS 10.0 for Windows; SPSS, Chicago, Ill). The weighted was calculated by using exact methods. The power calculation was performed by using software that was available online at the Web site of the Department of Statistics of the University of California Los Angeles, Los Angeles, Calif (www.stat.ucla.edu/).
Phantom Study
A phantom study was performed to estimate patient dose by using the experimental versus the control radiographic protocols. The phantom, which consisted of multiple polymethyl methacrylate plates fixed together without intervening gaps, was mounted in the detector center in direct contact with the front plate and is similar to chest phantoms in other published studies (14,31–33).
Polymethyl methacrylate has a mean attenuation of 125 HU. Physical density can be estimated from Hounsfield units (HU) (34,35) by using the following equation: density = HU + 1000/1000, where physical density is expressed in grams per milliliter.
For example, polymethyl methacrylate has an estimated physical density of 1.125 g/mL, which matches the published density for polymethyl methacrylate (36). Thus, if one assumes that the attenuation values of human fat, soft tissue, and lung tissue are –100 HU, 50 HU, and –800 HU, respectively (corresponding to estimated densities of 0.9 g/mL, 1.05 g/mL, and 0.2 g/mL), then a 1-cm slab of polymethyl methacrylate causes the same x-ray attenuation as 1.25 cm of fat, 1.07 cm of soft tissue, and 5.6 cm of lung tissue. Three phantoms were constructed by using different numbers of plates to achieve total thicknesses of 7.5 cm, 12.5 cm, and 21.5 cm. The phantoms were chosen to simulate the following: a 7.5-cm-thick phantom, which corresponds to a thin patient imaged in the posteroanterior projection (15 cm of lung tissue, 3.2 cm of soft tissue, and 2 cm of fat); a 21.5-cm-thick phantom, which corresponds to a large patient imaged in the lateral projection (19 cm of lung tissue, 12 cm of soft tissue, and 7.5 cm of fat); and a 12.5-cm-thick phantom, which corresponds to a midsized variant with features that are between those of the small phantom and those of the large phantom.
The radiographic system was identical to that used for patients. Images were acquired with (experimental protocol) and without (control protocol) 0.3-mm copper filtration. The field of view was 40 x 40 cm. Other parameters were kept constant as described previously.
The dose-area product was measured once (single trial) for the three phantoms by using each radiographic protocol. Entrance dose was then calculated from the dose-area product by applying the inverse-square law and the known field of view. Monte Carlo calculations were performed to estimate the energy absorbed in the center of the phantom. In detail, we performed computer calculations that simulated the propagation of single photons with the Monte Carlo method through complex geometric phantoms of arbitrary materials, with consideration of all relevant photon interactions, such as photo-effect, coherent and incoherent scatter, as well as K fluorescence. For calculations, we used a software program (Mocassim; Siemens Medical Solutions). Statistical comparisons of entrance dose and central energy absorption for the three phantoms were not determined because only single-trial measurements were performed.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
The mean interval between examinations in the 12 paired patients was 93 days (range, 39–147 days). In eight (67%) of 12 paired patients, radiographs were normal at both examinations. In the other four (33%), there was at least one radiographic finding (Table 3). The findings were identical at both examinations.
|
For the 12 paired controls, mean dose-area product in the experimental group was 49% less (3.5 µGy · m2 ± 1.0 vs 6.8 µGy · m2 ± 3.2 for the posteroanterior projection, P = .001) and 55% less (14.4 µGy · m2 ± 8.2 vs 32.2 µGy · m2 ± 27.4 for the lateral projection, P = .030) than it was in the control group. Mean tube current–time product was 42% greater (2.0 mAs ± 0.5 vs 1.4 mAs ± 0.5 for the posteroanterior projection, P < .001) and 23% greater (7.3 mAs ± 3.7 vs 5.9 mAs ± 4.5 for the lateral projection, P = .28) in the experimental group than it was in the control group. Mean exposure time was 51% greater (5.3 msec ± 1.5 vs 3.5 msec ± 1.7 for the posteroanterior projection, P < .001) and 24% greater (22.7 msec ± 11.8 vs 18.4 msec ± 14.6 for the lateral projection, P = .30) in the experimental group than it was in the control group.
Averaged for the unpaired and paired patients, exposure time increased by 48% for the posteroanterior view and by 34% for the lateral view when copper filtration was added. In the experimental group, exposure time exceeded the recommended American College of Radiology threshold of 40 msec for two of 38 lateral radiographs (42 msec, 62 msec) in unpaired patients and one of 12 lateral radiographs (49 msec) in paired patients. Patients' body mass indexes were 27.3, 36.8, and 30.1 kg/m2, respectively. In the control group, exposure time exceeded 40 msec for none of 38 lateral radiographs in unpaired patients and one of 12 lateral radiographs (59 msec) in paired patients. This patient's body mass index was 37.4 kg/m2. Differences in frequency were not significant for either the unpaired (two of 38 vs none of 38, P = .625) or paired (one of 12 vs one of 12, P > .99) comparisons. Maximum exposure time for posteroanterior radiographs was 8 msec for the experimental protocol and 7 msec for the control protocol.
Image Quality
Weighted interobserver agreement between the three readers was 99.7%. Because the expected weighted agreement was also high (99.6%), the weighted value was relatively low (0.246).
In regard to the observed frequency of the readers' ordinal scores for the 12 analyzed variables (six for the posteroanterior view and six for the lateral view), all variables in all images were of good (score of 3), good to excellent (score of 2), or excellent (score of 1) quality. There were no scores of 4 or higher for unpaired (Fig 1 ) or paired (Fig 2) examinations.
|
|
|
|
In the paired examinations, mean frequencies for a score of 1 ranged from 9.0 (75%) of 12.0 to 12.0 (100%) of 12.0 in the experimental group and 8.0 (67%) of 12.0 to 11.7 (98%) of 12.0 in the control group. Frequencies for a score of 2 ranged from 0 (0%) of 12.0 to 3.0 (25%) of 12.0 in the experimental group and 0.3 (3%) of 12.0 to 3.3 (28%) of 12.0 in the control group. Frequencies for a score of 3 were 0 (0%) of 12.0 for all variables in the experimental group and ranged from 0 (0%) of 12.0 to 0.3 (3%) of 12.0 in the control group. By using the Wilcoxon signed rank test, only one of the 36 comparisons showed a statistically significant difference: The experimental protocol was judged superior to the control protocol (P = .046) by one of the three readers for delineation of the variable airways in the lateral view. Differences for all other comparisons were statistically insignificant, with P values ranging from P = .157 to P > .99. The control protocol was not judged significantly superior to the experimental control in any of the 36 comparisons. Moreover, global image quality scores for the posteroanterior and lateral projections were virtually identical in both groups (P values ranged from P = .317 to P > .99). By using the McNemar test for analysis of dichotomized data, no comparison was significantly different (P = .500 to P > .99). Motion degradation was not observed. An example for images with and without copper filtration in the same patient is provided in Figure 3.
|
|
|
|
For the 36 paired comparisons, the power of a single comparison to detect a 25 percentage point difference in frequency of excellent scores ranged from 26.4% to 68.5% (mean power, 61.4% ± 12.3). Five comparisons had less than 50% power. Five had 50.2% power. The remaining 26 had 68.5% power. If we assume that the actual power of each comparison was 26.4% (the lowest observed power), the power to detect six or more statistically meaningful comparisons would have been 94.1%.
Phantom Study
Comparison of the experimental protocol with the control protocol resulted in an entrance dose reduction of 44% for the small phantom (13.4 µGy vs 24.0 µGy), of 46% for the midsized phantom (30.3 µGy vs 55.7 µGy), and of 47% for the large phantom (131.8 µGy vs 250.5 µGy). A reduction of the absorbed energy in the center of the phantom (Monte Carlo estimation) of 33% (13.4 µGy vs 20.1 µGy), of 31% (25.0 µGy vs 36.1 µGy), and of 29% (62.2 µGy vs 87.2 µGy) was determined for the three sizes, respectively. Hence, averaged for the three phantom sizes, reduction of absorbed energy by copper filtration was 31%.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The benefit of beam filtration, with decreasing patient radiation, has been well demonstrated (1,2,5,8). The drawback of x-ray beam hardening is reduction of image contrast. As opposed to screen-film radiography, digital radiography provides a linear response over a wide latitude and is amenable to image processing. Hence, loss of image contrast caused by x-ray beam hardening can be partially recovered with image processing, particularly with manipulation of image window width and window level. This manipulation, however, increases apparent image noise because the contrast-to-noise ratio remains constant. A common compromise is to add filtration and simultaneously to reduce tube voltage. This compromise, unfortunately, reduces the dose-saving effect of beam filtration. Previously, Strotzer et al (22) reported excellent image quality for chest radiographs acquired with standard parameters (125 kVp, 2.5 µGy, no copper filtration). Similarly, we found consistently good or excellent image quality in radiographs obtained with the same parameters (control protocol). We also found, however, virtually identical image quality after addition of 0.3-mm copper filtration (experimental protocol). Thus, our three readers judged the radiographs obtained with additional beam filtration and the standard images acquired with the conventional x-ray spectrum to be of equivalent diagnostic quality. In fact, of the 72 comparisons (36 in the unpaired and 36 in the paired cohorts), only two had significantly different scores between the experimental and the control groups; in both cases, the experimental protocol was statistically superior. Weighted agreement between the three readers exceeded 99%. The weighted statistic ( = 0.246) suggested only fair strength of agreement, however, at least in part because the expected agreement also exceeded 99%.
We retrospectively estimated the power of each comparison to achieve a clinically meaningful difference in frequency of excellent scores (a clinically meaningful difference was retrospectively defined as an absolute 25 percentage point reduction in frequency of excellent scores). Although the retrospectively estimated power of individual comparisons had a considerably broad range, 28 (78%) of 36 unpaired comparisons had greater than 85% power and 26 (72%) of 36 paired comparisons had 68.5% power. Moreover, because multiple comparisons were determined, the power to achieve statistical significance in at least six of the 36 comparisons in the cohort was relatively high even if we conservatively assumed the lowest observed power in each cohort for all comparisons. These considerations confirm that our study had an adequate power to detect clinically meaningful differences in image quality. Our clinical results are consistent with recently published computer modeling and phantom data in regard to x-ray spectrum optimization for chest radiography by using cesium iodide–amorphous silicon technology (31). Dobbins et al (31) evaluated different tube voltages (50–150 kVp in 10-kVp increments), filter materials (aluminum, copper, molybdenum, gadolinium, and tungsten), and filter thicknesses (0.0–2.0 half-value layer) by using a computer spectrum model and phantom experiments. These investigators concluded that 120 kVp with the thickest possible copper filtration provided the best performance, considering signal-to-noise ratio, entrance dose, and relatively low contrast of the bone structures (to avoid obscuring the lucent lung).
Addition of filtration, however, must be offset by an increase in tube load (the product of tube current and exposure time) to maintain adequate detector entrance dose (37). The radiographic system described here fixed the peak tube current at the maximum allowed by the manufacturer (400 mA). Thus, to maintain adequate entrance dose, exposure time was prolonged by 46% and 33% for the posteroanterior and lateral views, respectively. Similarly, tube load was increased by 35% and 32%. (The reason that the proportional changes in exposure time and tube load were different is that the current declination curve was variable and dependent on beam attenuation according to the patient's body habitus.) According to American College of Radiology standards for adult chest radiography, exposure time should not exceed 40 msec (24). For 97 of 100 images acquired with beam filtration, the exposure time was within this limit. For three patients, exposure times for images acquired in the lateral projection were longer than 40 msec. One patient in the control group also required an exposure time longer than 40 msec for the lateral view. All four patients were overweight, and three were obese. Moreover, although exposure time guidelines were exceeded in four patients, no motion artifacts were observed in these or other patients. Thus, the addition of 0.3-mm copper filtration is feasible when applied to the evaluated system.
In our clinical study, beam filtration reduced the dose-area product by approximately 55%. In the phantom experiment, entrance dose was reduced by approximately 46%. Neither the dose-area product nor the entrance dose, however, can be used to reliably assess the reduction of patient radiation with beam filtration. The reason is that both measurements gauge dose for the tissue on the tube side, where dose saving caused by beam hardening is particularly pronounced (1–3). Radiation burden for tissues adjacent to the detector (ie, at the beam exit point) is similar for filtered and unfiltered x-ray spectra when the detector entrance dose is fixed, as in our study. Therefore, we performed Monte Carlo calculations in a phantom experiment to estimate absorbed dose in the center of the phantom. This value was not considered as an absolute value but rather as a relative measure of radiation burden. The results suggest that addition of 0.3-mm copper filtration at a tube voltage of 125 kVp helped reduce patient radiation exposure by about 31%. This value is consistent with published data (5).
The best parameter to assess patient risk is effective dose equivalent. Unfortunately, it is difficult to estimate effective dose equivalent theoretically and virtually impossible to measure it experimentally in human subjects. With consideration of the previously outlined distribution of dose savings in tissue with beam filtration, we made the practical assumption that comparing absorbed energy in the center of the phantom provides a reasonable approximation of relative dose.
A limitation of our study design is that the number of paired samples was limited. Several considerations help validate our results despite this limitation. The experimental and control groups showed no significant differences in age, body mass index, and the sex distribution. Moreover, no patient had undergone median sternotomy, breast implantation, or mastectomy. Thus, it can be assumed that the body habitus was similar in the experimental and control groups. The proportion of normal and abnormal radiographs and the frequency of specific imaging findings, including diffuse abnormalities such as fibrosis and edema, were also similar in both groups. When circumscribed pathologic changes were present, a neutral radiologist ensured that the readers evaluated the same uninvolved areas. Finally, separate analyses of unpaired and paired patients revealed identical results.
Disadvantages of flat-panel detector technology include high fixed costs and large image file size (18 MB per full-size image), which slows download to review workstations and imposes considerable storage capacity requirements for picture archiving and communication systems.
In conclusion, equivalent chest radiographic image quality and marked dose reduction can be achieved by using 0.3-mm copper filtration with a cesium iodide–amorphous silicon system. Rather than adding beam filtration as reported here, an alternative strategy for decreasing patient radiation is to diminish detector entrance dose. In our study, we fixed detector entrance dose at 2.5 µGy, a standard setting in many radiology departments. Lower detector entrance doses, however, are possible. Several authors have confirmed patient dose reduction of more than 50% without beam filtration or compromise of image quality by decreasing detector entrance dose and by using cesium iodide–amorphous silicon detectors (15–18,21,38,39). These results highlight the potential clinical advantages of cesium iodide–amorphous silicon technology, which permit rational manipulation of acquisition parameters to provide excellent image quality at a lower patient dose. Further study is required to empirically compare beam filtration and detector entrance dose reduction strategies by using flat-panel detector systems.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, O.W.H., M.S., S.F., M.V.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, O.W.H., I.B., M.V.; clinical studies, O.W.H., M.S., I.B., S.F., M.V.; experimental studies, O.W.H., N.Z., M.V.; statistical analysis, O.W.H., C.B.S.; and manuscript editing, O.W.H., C.B.S., M.S., I.B., N.Z., M.V.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | < |
