Conventional High-Resolution CT Versus Helical High- Resolution MDCT in the Detection of Bronchiectasis

Transcription

1 High- Resolution CT Versus MDCT in Detecting Bronchiectas is Chest Imaging Original Research A C M E D E N T U R I C A L I M A G I N G AJR 2006; 187: X/06/ American Roentgen Ray Society Y O Jonathan D. Dodd 1 Carolina A. Souza Nestor L. Müller Dodd JD, Souza CA, Müller NL F Keywords: bronchiectasis, high-resolution CT, lung disease, MDCT DOI: /AJR Received April 28, 2005; accepted after revision June 24, All authors: Department of Radiology, Vancouver General Hospital, 855 W. 12th Ave., Vancouver, British Columbia, Canada V6K 1R4. Address correspondence to N. L. Müller. Conventional High-Resolution CT Versus Helical High- Resolution MDCT in the Detection of Bronchiectasis OBJECTIVE. The purpose of this study was to compare conventional high-resolution CT (HRCT) with helical 16-MDCT in the detection of bronchiectasis. MATERIALS AND METHODS. We retrospectively evaluated 80 patients including 61 with bronchiectasis (mean age, 64 years; range, years) and a control group of 19 patients with normal MDCT of the chest. Two sets of images were blindly, randomly analyzed by two observers: contiguous 1-mm slices (MDCT set) and 1-mm slices every 10 mm (HRCT set) derived from the MDCT set. Images were scored independently for presence, extent, and severity of bronchiectasis, followed by a consensus interpretation. Kappa analysis assessed inter- and intraobserver agreement. MDCT was the radiologic gold standard. RESULTS. Of the 61 patients with bronchiectasis diagnosed with MDCT, 56 (92%) were positive for bronchiectasis on conventional HRCT. Seven patients had positive MDCT scans only, and two patients had positive HRCT scans only. Of 479 lobes, 59 were positive for bronchiectasis on MDCT and negative on HRCT, and 19 lobes were positive for bronchiectasis on HRCT and negative on MDCT (p < ). MDCT showed 25 more lobes with cylindric, 11 more lobes with varicose, and four more lobes with cystic bronchiectasis than did HRCT. Sensitivity, specificity, and positive and negative predictive values of HRCT in detecting bronchiectasis were 71%, 93%, 88%, and 81%, respectively. Interobserver agreement for presence, extent, and severity of bronchiectasis ranged from moderate to good for MDCT (kappa values, 0.64, 0.5, and 0.48, respectively) and poor to good for HRCT (kappa values, 0.65, 0.46, and 0.25, respectively). CONCLUSION. Contiguous helical 16-MDCT with 1-mm collimation is superior to HRCT at 10-mm intervals in showing the presence and extent of bronchiectasis. ronchiectasis is defined pathologically as an irreversible, abnormal B dilatation of the bronchi [1]. It can be related to a wide variety of causes, such as infection, airway obstruction from tumor or foreign body, and impairment of mucociliary clearance as in cystic fibrosis [2]. Symptoms and signs include chronic cough, recurrent infection, and hemoptysis [3]. High-resolution CT (HRCT) using 1- to 1.5-mm-collimation sections at 10-mm intervals has been regarded as the radiologic gold standard in diagnosis [4, 5]. Using this protocol, sensitivities and specificities approaching 98% and 93 99%, respectively, have been reported [4 6]. Limitations of conventional HRCT studies in the detection of bronchiectasis are related to the gaps between noncontiguous slices and imaging degradation by breathing and cardiac motion artifacts [7 9]. Single-detector helical CT has been shown to increase the detection rate of bronchiectasis and decrease motion artifact compared with conventional HRCT [8, 10, 11]. The section thickness in these studies ranged from 3 to 5 mm, and the increase in sensitivity was relatively small. The advent of MDCT has allowed narrower collimation and faster acquisitions during a single breath-hold with near isometric z- axis resolution [12, 13]. Despite the faster scanning time, MDCT image quality is equal to that of conventional HRCT [14]. Chooi et al. [15] showed an increased level of confidence and interobserver agreement in diagnosing bronchiectasis in seven patients using 4-MDCT. We hypothesized that helical 16- MDCT using 1-mm contiguous slices would lead to an increased detection of bronchiectasis compared with conventional HRCT. The aim of this study was to compare conventional HRCT with MDCT in the detection of bronchiectasis and in the assessment of disease extent and severity. 414 AJR:187, August 2006

2 High-Resolution CT Versus MDCT in Detecting Bronchiectasis Materials and Methods We evaluated retrospectively 61 patients (31 men, 30 women; mean age, 64 years; range, years) who had a diagnosis of bronchiectasis made by helical high-resolution 16-MDCT of the chest by experienced chest radiologists in our department between 2003 and Nineteen patients (10 men, 9 women; mean age, 49 years; range, years) with normal MDCT scans were included as a control group. All MDCT scans were obtained on a Somatom 16- MDCT scanner (Siemens Medical Solutions) with 1- mm thickness, 15-mm table speed per rotation, 0.5- second gantry rotation time, 120 kvp, and 130 mas. Scans were performed at full inspiration from lung apex to base. Contiguous slices were reconstructed at 1-mm intervals using a high-spatial-frequency boneedge enhancing algorithm. Lung windows with a width of 1,500 and a level of 700 H were applied to all images. Images were interpreted on a workstation. Two separate sets of images were generated: one set of contiguous 1-mm slices from apex to lung base (MDCT set) and a second set of 1-mm slices at 10- mm intervals (HRCT set) derived from the MDCT set. The two sets of images were analyzed in random order by two independent observers experienced in CT scoring who were blinded to patient demographics, clinical findings, and reported CT results. Images were evaluated for bronchiectasis in all six lobes (the lingula was counted as a separate lobe). Bronchiectasis was defined as a bronchus with an internal diameter larger than its accompanying pulmonary artery, lack of tapering of the bronchial lumen for longer then 2 cm, and visualization of a bronchus within 1 cm of the costal pleura [4]. The extent of bronchiectasis was scored from 0 to 3 (0 = no bronchiectasis, 1 = bronchiectasis involving less than a third of the lobe, 2 = bronchiectasis involving between a third and two thirds of the lobe, 3 = bronchiectasis involving more than two thirds of the lobe). The severity of bronchiectasis was scored from 0 to 3 (0 = normal, 1 = cylindric, 2 = varicose, 3 = cystic). Bronchiectasis was scored as cylindric when dilated bronchi had relatively uniform caliber and roughly parallel walls, varicose when dilated bronchi had a beaded appearance, and cystic when dilated bronchi had a cystic or saccular appearance [16]. When lobes had more then one type of bronchiectasis, the more severe type was recorded. Images were initially scored independently, and a second interpretation was performed by consensus. Inter- and intraobserver agreement for prevalence, extent, and severity of bronchiectasis for HRCT and MDCT was assessed using weighted kappa analysis. Kappa scores were interpreted as less than 0.20 = poor, = fair, = moderate, = good, and = very good [17]. Agreements were calculated per patient and per lobe. Univariate correlations between the presence, extent, and severity of bronchiectasis were performed using Spearman s rank correlation. Prevalence of bronchiectasis between MDCT and HRCT was compared using the chi-square test. Results Seven patients had positive MDCT scans and negative conventional HRCT scans. In five of these patients, only one lobe was involved; in the remainder, two lobes were involved. All of these patients had cylindric bronchiectasis involving less than a third of a lobe. Two patients were scored in consensus as having a positive HRCT and negative MDCT. Both patients appeared to have cylindric bronchiectasis affecting less than a third of the right upper lobe on HRCT. Analysis of the MDCT, however, showed normal tapering of the bronchi. All lobes had a higher prevalence of bronchiectasis with MDCT than HRCT (Table 1). Of the 479 lobes analyzed (one patient had a right upper lobectomy previously), bronchiectasis was identified in consensus in 202 lobes with MDCT and 162 lobes with HRCT (p < ). The majority of these scans had cylindric bronchiectasis involving less than a third of a lobe (Fig. 1). Subsequent review of the images showed that this was principally because small areas of bronchiectasis visualized on MDCT were located in between slices on HRCT. Nineteen lobes had HRCT scans interpreted in consensus as positive for bronchiectasis that were negative with MDCT (Fig. 2). The majority of these scans were interpreted on HRCT as cylindric bronchiectasis involving less than a third of a lobe. Subsequent review of the images showed that these bronchi were tapering normally on MDCT. TABLE 1: Prevalence and Distribution of Bronchiectasis with HRCT and MDCT CT Protocol Lingula Total HRCT MDCT Note Results are by consensus. HRCT = high-resolution CT. Most patients showed a greater extent of bronchiectasis on MDCT compared with HRCT (Table 2). MDCT showed 36 more lobes with bronchiectasis involving less than a third of a lobe and five more lobes with bronchiectasis involving between a third and two thirds of a lobe. The detection of the extent of bronchiectasis involving more than two thirds of a lobe, seen in three lobes, was equal between MDCT and conventional HRCT. One patient who had between a third and two thirds of a lobe scored in consensus on MDCT had more than two thirds of a lobe scored in consensus with HRCT. Most patients showed an increase in the severity of bronchiectasis with MDCT compared with HRCT (Table 3). MDCT depicted 25 more lobes with cylindric, 11 more lobes with varicose, and four more lobes with cystic bronchiectasis than HRCT. One patient had varicose bronchiectasis scored in one lobe in consensus on HRCT, which was scored as cylindric in consensus on MDCT. The interobserver agreements in the interpretation of HRCT and MDCT images were moderate to good for assessment of prevalence of bronchiectasis per patient and per lobe (Table 4) and for assessment of the extent of bronchiectasis. The interobserver agreement for severity of bronchiectasis was fair for HRCT and moderate for MDCT. All interobserver agreements were statistically significant at a p < level. For prevalence of bronchiectasis, intraobserver agreement per patient for observer 1, observer 2, and in consensus was moderate, moderate, and good, respectively (Table 5). For extent of bronchiectasis, intraobserver agreement for observer 1, observer 2, and in consensus was moderate, good, and very good, respectively. For severity of bronchiectasis, intraobserver agreement for observer 1, observer 2, and in consensus was fair, good, and good, respectively. All intraobserver agreements were statistically significant at a p < level. Sensitivity, specificity, and positive and negative predictive values of HRCT in detecting bronchiectasis were 71%, 93%, 88%, and 81%, respectively (Table 6). All lobes showed moderate but significant correlations with each other for prevalence of bronchiectasis at the p < 0.01 level (Table 7). The extent and severity of bronchiectasis per lobe also correlated with each other (R = 0.51; p <0.01). Discussion Conventional HRCT using 1- to 1.5-mm slices every 10 mm has been regarded as the AJR:187, August

3 radiologic gold standard for detecting bronchiectasis [4, 5]. The HRCT findings of bronchiectasis include identification of an enlarged internal bronchial diameter, lack of tapering of a bronchus as it travels toward the periphery of the lung, and identification of airways in the periphery of the lung [2]. Limitations include breathing and cardiac pulsation artifacts, particularly in the pericardiac area; missing small focal areas of bronchiectasis in between HRCT sections; misinterpretation of mucus plugs in bronchiectatic segments as nodules; and using the bronchoarterial diameter ratio close to bifurcations of A bronchi or vessels because the two may not divide at the same level. This study shows that high-resolution helical 16-MDCT using 1-mm contiguous slices results in a significant increase in the detection of bronchiectasis compared with HRCT using 1-mm slices at 10-mm intervals. MDCT de- B C Fig year-old man suspected clinically of bronchiectasis. A C, Conventional high-resolution CT (HRCT) images show no evidence of bronchiectasis in right upper lobe. (Fig. 1 continues on next page) 416 AJR:187, August 2006

4 High-Resolution CT Versus MDCT in Detecting Bronchiectasis tected bronchiectasis in 202 lobes compared with 162 lobes with HRCT. This was principally because small areas of bronchiectasis visualized on MDCT were located in between slices on HRCT. Lack of distal tapering of a bronchus, considered the most sensitive sign of bronchiectasis, was often difficult to visualize with conventional HRCT [4, 18]. Similarly, D F identifying small ectatic bronchi in the periphery of the lung was easier on MDCT images. False-positive HRCTs were caused by vessels branching before bronchi, thus resulting in an apparent increase in bronchoarterial diameter ratios and lack of adequate visualization of the entire vessel when in the plane of an image in relation to its accompanying bronchus. Fig. 1 (continued) 67-year-old man suspected clinically of bronchiectasis. D F, Corresponding MDCT images at level of B. At this level, bronchiectasis (arrow) in D and E abuts the costal margin in F. This resulted in a false-negative conventional HRCT scan. That conventional HRCT can miss small areas of bronchiectasis is well known. In one radiologic pathologic correlative study, 22 patients underwent HRCT followed by surgical resection for bronchiectasis [4]. HRCT resulted in six false-negative results from 47 lobes with bronchiectasis [4]. Although HRCT allowed detection of some abnormal- E AJR:187, August

5 ity in all false-negative cases (three had consolidation, two had an apparent mass, and one had an apparent cavity), it did not depict the presence of bronchiectasis. Some of these limitations have been overcome using helical CT. For example, in one study of 50 patients evaluated for suspected bronchiectasis, single-row helical CT found bronchiectasis in 90 lung segments compared with 77 segments with HRCT [8]. Interobserver agreement was A B C D significantly better with helical than with conventional HRCT. A disadvantage of single-row helical CT is partial volume averaging. In the study mentioned earlier, the effective slice thickness of 3.7 mm resulted in small focal areas of bronchiectasis seen on conventional HRCT being missed on singlerow helical CT in seven lung segments [8]. We routinely use 1-mm slices in the evaluation of bronchiectasis. Remy-Jardin et al [19]. E Fig year-old man suspected clinically of bronchiectasis. A, In conventional high-resolution CT (HRCT), image because branching points of vessels are in between slices, their relationships to accompanying bronchi are not clearly seen. Both observers scored the bronchus (arrow) as bronchiectasis in right middle lobe. B E, Corresponding MDCT images show the vessel branching before its accompanying bronchus (arrow), resulting in an apparent increased bronchoarterial diameter ratio. This resulted in a false-positive HRCT scan. compared 1.25-mm and 3-mm slice thickness in 40 patients with bronchiectasis, and they found no significant difference in overall detection. However, 11 cylindric bronchiectatic lung zones (16%) were detected with 1.25-mm slices that were not detected with 3-mm slices (p < 0.01). In our study, 59 lobes (29%) had cylindric bronchiectasis with MDCT only. This discrepancy in results is likely related to differing MDCT protocols between the two studies. 418 AJR:187, August 2006

6 High-Resolution CT Versus MDCT in Detecting Bronchiectasis TABLE 2: Extent of Bronchiectasis with HRCT and MDCT CT Protocol Lingula Total HRCT < 1/3 of lobe /3 2/3 of lobe > 2/3 of lobe MDCT < 1/3 of lobe /3 2/3 of lobe > 2/3 of lobe Note Results are by consensus. HRCT = high-resolution CT. TABLE 3: Severity of Bronchiectasis with HRCT and MDCT CT Protocol Lingula Total HRCT Cylindric Varicose Cystic MDCT Cylindric Varicose Cystic Note Results are by consensus. HRCT = high-resolution CT. Whereas we used a contiguous imaging protocol, Remy-Jardin et al. used incremental 10- mm intervals between slices. Our interobserver agreement for prevalence of bronchiectasis was moderate per patient and good per lobe. In the study by Lucidarme et al. [8], interobserver agreement between conventional and single-detector helical CT for prevalence of bronchiectasis per patient and per lobe was either good or very good. Our interobserver agreement for extent of bronchiectasis was generally lower with helical and higher with conventional HRCT than that of Lucidarme et al. This difference may be related to different patient populations or types of analysis. Lucidarme et al. excluded the right middle lobe and lingula in their analysis of the extent of bronchiectasis. This may have led to an underrepresentation of the extent of bronchiectasis per patient. Our interobserver agreement for severity of bronchiectasis was generally similar to that of Lucidarme et al. Correlation analysis showed that the presence of bronchiectasis in one lobe was associated with a greater likelihood of bronchiectasis in the remaining lobes. Extent of bronchiectasis correlated significantly with severity of bronchiectasis. From a practical viewpoint, this suggests that detection of bronchiectasis in one lobe should prompt careful assessment of the other lobes for further disease, and the more extensive the identified bronchiectasis, the more severe it is likely to be. Most studies evaluating MDCT for bronchiectasis have assessed multiplanar reconstructions in the detection of and confidence in diagnosing bronchiectasis compared with conventional transverse imaging. Chooi et al. [15] assessed seven patients with bronchiectasis using a combination of transverse, coronal, and sagittal multiplanar reconstructions and found an improvement in the level of confidence in the diagnosis of bronchiectasis and interobserver agreement. Sung et al. [20] compared combined transverse and coronal multiplanar reconstructions with transverse imaging alone, and found a significantly higher detection rate of bronchiectasis using the combined protocol. We did not include multiplanar reconstructions in our MDCT protocol, which might have improved the detection rate of bronchiectasis and interobserver agreement on MDCT. Our study had several limitations. It was retrospective so we were unable to correlate the CT TABLE 4: Interobserver Agreement for Bronchiectasis with HRCT and MDCT Kappa Value CT Protocol Per Per Patient HRCT Prevalence Extent 0.46 Severity 0.25 MDCT Prevalence Extent 0.50 Severity 0.48 Note All results are significant at the p < level. Dash ( ) indicates not applicable. HRCT = highresolution CT. TABLE 5: Intraobserver Agreement for Bronchiectasis Between HRCT and MDCT Kappa Value Observers Per Per Patient Observer 1 Prevalence Extent 0.56 Severity 0.31 Observer 2 Prevalence Extent 0.69 Severity 0.47 Consensus Prevalence Extent 0.85 Severity 0.45 Note All results are significant at the p <.001 level. HRCT = high-resolution CT. Dash ( ) indicates not applicable. TABLE 6: Statistical Analysis of Bronchiectasis Detected with HRCT and MDCT HRCT Finding MDCT Finding Positive Negative Positive Negative Note Results are by consensus using MDCT as the gold standard. Statistical performance of highresolution CT (HRCT) for detecting bronchiectasis was as follows: prevalence, 42%; sensitivity, 71% (95% confidence interval [CI], 64 78%); specificity, 93% (95% CI, 90 96%); positive and negative predictive values, 88% and 81%, respectively. findings with those of pathologic specimens. Differences in breathing or cardiac motion arti- AJR:187, August

7 TABLE 7: Correlation Between s for Presence of Bronchiectasis Part of Lung Being Evaluated Lingula upper lobe middle lobe lower lobe upper lobe Lingula lower lobe Note Numbers are R values. All lobes correlated at the p < 0.01 level. Dash ( ) indicates not applicable. fact between CT protocols could not be assessed because conventional HRCT images were derived from MDCT images. One of the advantages of conventional HRCT is the short breathhold per slice (approximately 1 second), whereas thin-section MDCT of the entire thorax requires a breathhold of seconds. For dyspneic patients, this difference may become important. Finally, the radiation dose per patient was not calculated in our study. The estimated effective dose using our MDCT protocol is considerably higher than with HRCT performed at 10-mm intervals. Jung et al. [9] evaluated the effect of varying radiation doses on image quality with single-detector helical CT. Acceptable image quality was obtained at a tube current as low as 40 ma. Yi et al. [21] subsequently showed acceptable images using a modified low-dose MDCT protocol with a tube current of 70 ma. Evolving MDCT dose-reduction techniques, notably automatic tube current modulation that varies tube current in response to patient size and body habitus, result in further significant radiation dose reductions [22]. Although MDCT results in increased radiation, it also results in an increase in the detection of bronchiectasis. Detecting focal areas of bronchiectasis is important because even small areas can give rise to significant hemoptysis. In a cohort of 80 patients presenting with large or massive hemoptysis, Revel et al. [23] described 10 patients as having only minor and three patients as having no evidence of bronchiectasis on HRCT. Bronchiectasis was confirmed at surgery, pathology, or both, in all cases. In conjunction with the current study, these findings support the use of a helical, contiguous scanning protocol if minimal disease is to be detected. The risks from radiation are influenced by age and sex [24]. The risks are greatest in children and young adults and relatively small in male patients older than 40 years and female patients older than 50 years. Therefore, unless there is a suspicion of primary or metastatic neoplasm, we recommend that HRCT at 10-mm intervals be performed in the initial evaluation of patients under 40 years old, while men 40 years or older and women 50 years or older, particularly those presenting with severe recurrent respiratory infections or hemoptysis, should undergo MDCT in the evaluation of suspected bronchiectasis. In conclusion, helical 16-MDCT using 1-mm contiguous slices results in a significant improvement in the assessment of the presence, extent, and severity of bronchiectasis compared with conventional HRCT. Bronchiectasis detected in one lobe should prompt a careful search for further disease in the remaining lobes. References 1. Reid LM. Reduction in bronchial subdivision in bronchiectasis. Thorax 1950; 5: McGuinness G, Naidich DP. CT of airways disease and bronchiectasis. Radiol Clin North Am 2002; 40: Ooi GC, Khong PL, Chan-Yeung M, et al. High-resolution CT quantification of bronchiectasis: clinical and functional correlation. Radiology 2002; 225: Kang EY, Miller RR, Muller NL. Bronchiectasis: comparison of preoperative thin-section CT and pathologic findings in resected specimens. Radiology 1995; 195: Grenier P, Maurice F, Musset D, Menu Y, Nahum H. Bronchiectasis: assessment by thin-section CT. Radiology 1986; 161: Young K, Aspestrand F, Kolbenstvedt A. High resolution CT and bronchography in the assessment of bronchiectasis. Acta Radiol 1991; 32: Tarver RD, Conces DJ Jr, Godwin JD. Motion artifacts on CT simulate bronchiectasis. AJR 1988; 151: Lucidarme O, Grenier P, Coche E, Lenoir S, Aubert B, Beigelman C. Bronchiectasis: comparative assessment with thin-section CT and helical CT. Radiology 1996; 200: Jung KJ, Lee KS, Kim SY, Kim TS, Pyeun YS, Lee JY. Low-dose, volumetric helical CT: image quality, radiation dose, and usefulness for evaluation of bronchiectasis. Invest Radiol 2000; 35: van der Bruggen-Bogaarts BA, van der Bruggen HM, van Waes PF, Lammers JW. Assessment of bronchiectasis: comparison of HRCT and spiral volumetric CT. J Comput Assist Tomogr 1996; 20: Engeler CE, Tashjian JH, Engeler CM, Geise RA, Holm JC, Ritenour ER. Volumetric high-resolution CT in the diagnosis of interstitial lung disease and bronchiectasis: diagnostic accuracy and radiation dose. AJR 1994; 163: Fleischmann D, Rubin GD, Paik DS, et al. Stair-step artifacts with single versus multiple detector-row helical CT. Radiology 2000; 216: Honda O, Johkoh T, Yamamoto S, et al. Comparison of quality of multiplanar reconstructions and direct coronal multidetector CT scans of the lung. AJR 2002; 179: Honda O, Johkoh T, Tomiyama N, et al. High-resolution CT using multidetector CT equipment: evaluation of image quality in 11 cadaveric lungs and a phantom. AJR 2001; 177: Chooi WK, Matthews S, Bull MJ, Morcos SK. Multislice helical CT: the value of multiplanar image reconstruction in assessment of the bronchi and small airways disease. Br J Radiol 2003; 76: Naidich DP, McCauley DI, Khouri NF, Stitik FP, Siegelman SS. Computed tomography of bronchiectasis. J Comput Assist Tomogr 1982; 6: Daly LE, Bourke GJ. Interpretation and uses of medical statistics, 5th ed. London, UK: Marston Book Services, 2000: Lynch DA, Newell JD, Tschomper BA, Cink TM, Newman LS, Bethel R. Uncomplicated asthma in adults: comparison of CT appearance of the lungs in asthmatic and healthy subjects. Radiology 1993; 188: Remy-Jardin M, Amara A, Campistron P, et al. Diagnosis of bronchiectasis with multislice spiral CT: accuracy of 3-mm-thick structured sections. Eur Radiol 2003; 13: Sung YM, Lee KS, Yi CA, Yoon YC, Kim TS, Kim S. Additional coronal images using low-milliamperage multidetector-row computed tomography: effectiveness in the diagnosis of bronchiectasis. J Comput Assist Tomogr 2003; 27: Yi CA, Lee KS, Kim TS, Han D, Sung YM, Kim S. Multidetector CT of bronchiectasis: effect of radiation dose on image quality. AJR 2003; 181: Kalra MK, Maher MM, Toth TL, et al. Strategies for CT radiation dose optimization. Radiology 2004; 230: Revel MP, Fournier LS, Hennebicque AS, et al. Can CT replace bronchoscopy in the detection of the site and cause of bleeding in patients with large or massive hemoptysis? AJR 2002; 179: Mayo JR, Aldrich J, Muller NL. Radiation exposure at chest CT: a statement of the Fleischner Society. Radiology 2003; 228: AJR:187, August 2006