| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Mallinckrodt Institute of Technology, St Louis Children's Hospital, Washington University School of Medicine, 510 S Kingshighway Blvd, St Louis, MO 63110. Received October 7, 1998; revision requested November 11; revision received January 25, 1999; accepted March 2. S.D. supported in part by a grant from the Society for Pediatric Radiology Research and Education Foundation. Address reprint requests to S.D. (e-mail: don@mirlink.wustl.edu).
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Oleic acid was intravenously injected into three rabbits at each of four doses: 0, 0.02, 0.04, and 0.06 mL/kg. Two hours later, chest computed radiographs and screen-film radiographs were obtained at 60 kVp and 1.1 mAs. Additional computed radiographs were obtained after reducing milliampere seconds or by reducing milliampere seconds and increasing the kilovolt peak, which reduced bone marrow exposure by up to 20%. The presence of pulmonary opacities, "truth," was established by the wet-dry weight ratio and by chest computed tomography (CT). The radiographs were masked and randomized. Four observers rated the images for the presence of parenchymal opacities with a dichotomous score and judged the quality of the radiographs on a scale from 1 (worst) to 6 (best). Cochran Q tests and McNemar tests were used to analyze the differences in paired comparisons. Image quality was evaluated with logistic regression analysis.
RESULTS: There was no significant difference between truth and observer ability to detect opacity for either modality or for any exposure (P > .05). There was no significant difference between computed radiography and screen-film radiography for image quality (P > .05).
CONCLUSION: Computed radiography is equivalent to screen-film radiography in the detection of pulmonary edema. Radiation exposure reduction of 20% can be achieved without affecting pulmonary edema detection or image quality.
Index terms: Computed tomography (CT), comparative studies, 60.1211 • Lung, fluid, 60.711 • Radiography, comparative studies, 60.1211, 60.1215 • Radiography, in infants and children, 60.1215 • Radiography, storage phosphor, 60.1215
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Computed radiography is replacing screen-film radiography in neonatal bedside examinations because it has a wider exposure latitude than screen-film radiography (2–4). Because of the wider exposure latitude of computed radiography, the potential exists to reduce the radiation dose to the patient (5). Investigators have demonstrated that in reducing milliampere seconds there is a concomitant increase in noise (quantum mottle), which could be confused with findings of diseases such as respiratory distress syndrome (6,7). One study of hyaline membrane disease (8) that compared exposure reduction of computed radiography by using stacked cassettes without independent proof of disease, however, demonstrated a lower sensitivity for detection of hyaline membrane disease with only a 90% dose reduction, whereas image quality decreased significantly at a 50% dose reduction.
Computed radiography has been compared with screen-film radiography in the evaluation of pneumothorax in an animal model that simulates neonatal pneumothorax (9,10); to our knowledge, no specific disease has been studied in the neonate. The neonate is a poor model because often the neonate has multiple radiographic findings and because of concern for radiation exposure (4). To simulate opacities in a neonate, a rabbit model of pulmonary edema can be created by intravenously administering oleic acid, a free fatty acid (11).
The purpose of this study was (a) to compare computed radiography with screen-film radiography in the detection of pulmonary edema and (b) to determine if exposure can be reduced while maintaining equivalent diagnostic accuracy for pulmonary edema.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Experimental Technique
A 4-mm-thick, contiguous, baseline helical chest CT scan (Somatom Plus; Siemens, Erlangen, Germany) was obtained in all rabbits from the apices of the lungs through the dome of the diaphragm. The chest CT scan was used to evaluate for opacity; any rabbit with baseline opacity in the lungs was not used for this study. After CT was performed, oleic acid was injected intravenously. Three rabbits served as control animals and underwent the entire procedure except for the injection of oleic acid. Three rabbits were injected at each of three oleic acid doses (nine total): 0.02, 0.04, and 0.06 mL/kg.
Chest radiographs were obtained 2 hours after the intravenous injection of oleic acid. Computed radiographs and screen-film radiographs were obtained with the same radiographic unit (Multiplainographic II with Polydoros 800 generator; Siemens), with a focal spot of 0.3 mm. Screen-film radiographs were obtained with T-MAT L/RA film (Eastman Kodak, Rochester, NY) and Curix Opthos D cassettes (Agfa, Ridgefield Park, NJ). The computed radiographic system was the Ektascan Storage Phosphor 400 (Eastman Kodak).
The x-ray beam was collimated to the rabbit's chest, and the collimation remained fixed for the duration of the experiment. The source-film distance was 40 inches (101.6 cm). Screen-film radiography was optimized as previously described (9) at an exposure of 60 kVp at 1.1 mAs. Computed radiographs were obtained at the equivalent dose of 60 kVp at 1.1 mAs and at the following exposures: 60 kVp at 0.9 mAs, 70 kVp at 0.9 mAs, 70 kVp at 0.56 mAs, and 81 kVp at 0.56 mAs, our equipment's lowest milliampere second setting. The images were processed by following the pediatric chest protocol of our institution. For each exposure, the mean air entrance exposure was measured in air with a dosimeter (model 35050A; Keithley Instruments, Cleveland, Ohio), and bone marrow exposure was calculated (12,13).
Immediately after obtaining the radiographs, a final chest CT scan was obtained with the same technique used for the baseline chest CT. The rabbits were immediately sacrificed with a dose of 150 mg/kg pentobarbital sodium (Veterinary Laboratories, Lennexa, Kan). The lungs were removed in block, weighed wet, dried in a warmer, and weighed dry 3 days later. The wet-dry weight ratio was then calculated.
Image Interpretation
Four radiologists served as observers for the study (G.D.S., D.M.L., T.E.H., W.H.M.). Three observers were pediatric radiologists with American Board of Radiology certificates of added qualifications in pediatric radiology. The fourth observer was a fellow in pediatric radiology.
The images were masked and randomized and were viewed in two sessions. The first session consisted of viewing only the optimally exposed screen-film radiographs and the equivalently exposed computed radiographs (96 observations: four observers x two image types x 12 rabbits). The second session consisted of reviewing the optimally exposed screen-film radiographs and all the computed radiographic exposures (288 observations: four observers x six image types x 12 rabbits). The observers evaluated the images independently. They then determined whether opacity was present and gave only a dichotomous "yes" or "no"response. They also evaluated image quality on a six-point ordinal scale (1, worst, to 6, best). Finally, the observers were asked to determine whether the image was a computed radiograph or a screen-film radiograph.
Statistical Analysis
Truth for the presence of pulmonary edema was established by considering both the wet-dry weight ratio and the postinjection chest CT scan. One radiologist (S.D.) who was not an observer for the image review portion of this examination and who did not refer to the radiographs, observers' ratings, wet-dry weight ratios, or oleic acid injection volumes evaluated the chest CT scans for opacities.
The responses of each observer were tested against truth. Tests for intraobserver variation between the first and second sessions were performed. Interobserver variation was also tested for both the first and second sessions. For the first session, responses for computed radiographs were tested against the responses for screen-film radiographs for each observer. In the second session, the responses for each computed radiograph were tested against the responses for screen-film radiographs for each observer.
The Cochran Q test (STATISTICA; StatSoft, Tulsa, Okla) was used initially to test for differences in findings among multiple paired frequencies (ie, among observers, among image types [computed radiographs and screen-film radiographs], and among image types and truth). This analysis tests for differences in more than two dependent samples. It is an extension of the McNemar test for paired comparisons.
If the Cochran Q test resulted in a P value that indicated a significant difference (
.05), each pair was tested individually with the McNemar test. Exact P values were calculated with the McNemar test (STATXACT TURBO; Cytel, Cambridge, Mass). True-positive (TP), true-negative (TN), false-positive (FP), and false-negative (FN) rates were determined for optimally exposed screen-film radiographs, equivalently exposed computed radiographs, and reduced-exposure computed radiographs. Variation in image quality among image types was evaluated by using logistic regression analysis performed with JMP software (SAS, Cary, NC).
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
|
|
|
|
|
|
|
|
|
|
For the first session, there was no significant interobserver variation for computed radiography, but there was significant interobserver variation between observer 1 and observers 2 and 3 for screen-film radiography (observer 1 vs observer 2, P = .03; observer 1 vs observer 3, P = .02). Observer 1 had four FN responses with screen-film radiography that the other two observers did not have. In addition, observer 2 had two FP responses with screen-film radiography and observer 3 had three FP responses with screen-film radiography that observer 1 did not have.
For the second session, the only significant interobserver variation was between observers 1 and 4 for the computed radiograph at 70 kVp and 0.9 mAs exposure (P = .02). Observer 1 had four FN responses for this exposure and for all other computed radiographic exposures and screen-film radiographs. Observer 4 had only one FN response with the computed radiograph at 70 kVp and 0.9 mAs, and it was for one of the images for which observer 1 had an FN response. Observer 4 had four FP responses for the computed radiograph at 70 kVp and 0.9 mAs, of which observer 1 had none. Observer 4 also had three of these FP responses with screen-film radiographs.
Analysis of the FN observations showed that 47 of the 66 FN observations (Table 2) were for two rabbits, one rabbit each with an oleic acid dose of 0.02 mL/kg and 0.06 mL/kg. Subjectively, these rabbits had the least severe pulmonary edema as observed with CT. In the first session, seven of 16 images for these two rabbits had FN responses; four of these occurred with screen-film radiographs and three occurred with equivalently exposed computed radiographs. In the second session, 40 of 48 images for these rabbits had FN responses. Of these, eight responses were for screen-film radiographs, six for equivalently exposed computed radiographs, eight for 60-kVp at 0.9-mAs computed radiographs, five for 70-kVp at 0.9-mAs computed radiographs, six for 70-kVp at 0.56-mAs computed radiographs, and seven for 81-kVp at 0.56-mAs computed radiographs. No trend was identified in the 38 FP observations.
There was no significant difference in image quality between screen-film radiographic and any computed radiographic exposure (P = .16)(Table 3). Masking the images prevented selection bias. The observers could not distinguish screen-film radiographs from computed radiographs. The observers identified correctly the image type as screen-film radiograph only 60% of the time. For equivalent-dose computed radiographs, they identified correctly the image type only 44% of the time. When they reviewed all types of computed radiographs, they were more likely to identify computed radiographs as screen-film radiographs than as computed radiographs (125 times vs 115 times).
|
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
We believe the New Zealand white rabbit was an ideal animal model for this study. With it, we were able to create a range of subtle to obvious pulmonary edema with independent pathologic proof. Two of our rabbits had subtle pulmonary edema, as evidenced by their lower wet-dry weight ratios and CT-demonstrated opacities. In these rabbits, the observers were unable to detect the pulmonary edema with either screen-film radiography or computed radiography. The rabbits' responses to the oleic acid dose, however, appeared to be complex in that the wet-dry weight ratios did not correlate linearly with the administered doses. As an example, one rabbit that received a dose of 0.06 mL/kg failed to demonstrate pulmonary edema, whereas one that received a dose of 0.02 mL/kg did demonstrate pulmonary edema.
There was no significant difference between observer response and truth for any observer in either session 1 or 2, and there was no significant intra- or interobserver difference between any of the computed radiographic and screen-film radiographic exposures for session 2. One observer, however, tended to identify images as negative for opacity and accounted for significant interobserver screen-film radiographic variation in session 1, interobserver computed radiographic (70 kVp at only 0.9 mAs) variation in session 2, and intraobserver screen-film radiographic–computed radiographic variation in session 1. This observer had the same four FN responses for all screen-film radiographs in both sessions 1 and 2 and all computed radiographs in session 2. For session 1 computed radiographs, this observer had one of these FN responses but additionally had three FP responses. The observer's FN responses accounted for all significant differences in the study and demonstrated no pattern of superior performance for screen-film radiography or for computed radiography.
Computed radiography was equivalent to screen-film radiography in the detection of pulmonary edema. Contrary to concern that noise in computed radiography may be misinterpreted as disease (7), in our study there was no tendency for FP responses to account for significant differences between computed radiographs at any exposure and screen-film radiographs.
Furthermore, for image quality, there were no significant differences between any computed radiographic and screen-film radiographic exposure. Our conclusion is similar to that drawn from findings of a study of hyaline membrane disease (8). In that study, only computed radiographs were evaluated; screen-film radiographs were not obtained. The investigators found that as dose is reduced there is no increased FP result rate.
An advantage of computed radiography over screen-film radiography is the potential for exposure reduction (5). The acquisition of multiple radiographs results in cumulative radiation exposure, which can have biologic consequences (15,16). While the effects of low-dose radiation on the infant are difficult to measure, limiting radiation exposure to children is a goal of pediatric radiologists.
The reduction of milliampere seconds to reduce patient exposure is limited by the concurrent increase in image noise. An alternative exposure reduction method is to increase kilovolt peak while lowering milliampere seconds. Kilovolt peak affects image darkening by the fourth power. Increasing kilovolt peak from 60 to 70 increases image darkening by a factor of almost 2. This is because less of the higher-energy x-ray beam is attenuated by the patient. Thus, if one increases kilovolt peak, one can decrease milliampere seconds and achieve the same image darkening with less exposure to the patient and no increase in noise (17,18) (Table 4).
We achieved a 20% bone marrow radiation exposure reduction without affecting pulmonary edema detection or image quality. We did this by either decreasing milliampere seconds from 1.1 to 0.9 at 60 kVp or by increasing kilovolt peak from 60 to 70 with a concomitant decrease in milliampere seconds from 1.1 to 0.56 (Table 4). While not significantly different, 70 kVp at 0.56 mAs produced a higher number of TP results than 60 kVp at 0.9 mAs (Table 2). Increasing kilovolt peak from 70 to 81 while decreasing milliampere seconds to hold the dose roughly equivalent (Table 4) resulted in a decrease in the number of TP results. We theorize that this insignificant decrease could be due to the decrease in contrast with increasing kilovolt peak.
Further exploration with animal models and computed radiography for the reduction of radiation exposure is needed. First, a combination of higher kilovolt peak and even lower milliampere seconds than tested in this study may provide more exposure reduction. Second, additional filtration should be considered to reduce the number of lower-energy photons that are absorbed by the patient and do not contribute to the image. Third, use of image-processing algorithms that adjust for the loss of contrast might prove beneficial. Finally, once exposure reduction is realized in the experimental model, clinical trials will be needed.
In conclusion, computed radiography is equivalent to screen-film radiography in the detection of pulmonary edema in a rabbit model that simulates neonatal pulmonary infiltrates. Dose reduction by at least 20% is possible without loss of diagnostic accuracy or image quality.Practical application: This rabbit model of pulmonary edema in the neonate demonstrates that computed radiography can replace screen-film radiography without loss of diagnostic accuracy in detecting a common chest radiographic pattern seen with bedside neonatal radiography. Furthermore, results with this model indicate that dose reduction for neonatal imaging, especially with increased kilovolt peak, may be achieved without loss of diagnostic accuracy or decreased image quality.
|
Footnotes |
|---|
S.D. is a member of the Kodak Imaging Advisory Board.
2 Current address: Department of Radiology, Stanford University Hospital, Calif. 
Abbreviations: FN = false-negative FP = false-positive TN = true-negative TP = true-positive
Author contributions: Guarantor of integrity of entire study, S.D.; study concepts, S.D.; study design, S.D., C.F.H., T.L.S.; definition of intellectual content, S.D., C.F.H., G.D.S.; literature research, S.D.; experimental studies, S.D., T.L.S.; data acquisition, G.D.S., D.M.L., W.H.M., T.E.H.; data analysis, S.D., C.F.H.; statistical analysis, C.F.H.; manuscript preparation, S.D., C.F.H.; manuscript editing, G.D.S.; manuscript review, T.L.S., D.M.L., T.E.H.
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Don, B. R. Whiting, J. S. Ellinwood, D. H. Foos, K. A. Kronemer, and R. A. Kraus Neonatal Chest Computed Radiography: Image Processing and Optimal Image Display Am. J. Roentgenol., April 1, 2007; 188(4): 1138 - 1144. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. S. Kim, J.-G. Im, J. M. Goo, K. H. Lee, Y.-J. Lee, S. H. Kim, and S. Kim Detection of Pulmonary Edema in Pigs: Storage Phosphor versus Amorphous Selenium-based Flat-Panel-Detector Radiography Radiology, June 1, 2002; 223(3): 695 - 701. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J V Cook Radiation protection and quality assurance in paediatric radiology Imaging, December 1, 2001; 13(4): 229 - 238. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
