ORIGINAL ARTICLE
Estimation of Pleural Fluid Volumes on Chest Radiography Using Computed Tomography Volumetric Analysis An Update of the Visual Prediction Rule Joseph G. Mammarappallil, MD, PhD,* Sarah A. Anderson,w Kerry A. Danelson, PhD,w Joel A. Stitzel, PhD,w and Caroline Chiles, MD* Purpose: The purpose of the study was to determine the volumes of pleural fluid (PF) required to produce visible menisci in the lateral and posterior costophrenic angles (CPA) and obscure the hemidiaphragms (HD) on upright frontal and lateral chest radiographs (CXRs), using volumetric analysis of chest computed tomography (CT). Materials and Methods: A total of 98 patients with small pleural effusions on chest CT, in whom CXRs were obtained within a 24hour interval, were selected for retrospective analysis. PF within each hemithorax was quantified using a semiautomatic method of image segmentation. A cardiothoracic radiologist scored each hemithorax on each CXR from 0 to 3 (0—normal CPA, 1—fluid meniscus below the HD, 2—fluid meniscus at the level of the HD, 3—fluid opacity obscures the HD). Each CXR category was correlated with CT-determined PF volumes. Results: A mean of 20 mL of PF was present on CT without a visible correlate on CXR. A meniscus below the HD on CXR correlated with roughly 100 mL; a meniscus occurring at the HD correlated with roughly 250 mL; a meniscus obscuring the HD correlated with a mean of approximately 650 mL. There were large standard deviations for all PF volumes. Conclusions: We provide guidelines for estimating PF volumes on upright frontal and lateral CXRs. We also confirm that the lateral radiograph is more sensitive for detection of small pleural effusions, with blunting of the posterior CPA only correlating with a mean of 26 mL of PF.
Thoracentesis is recommended for almost all patients presenting with a new pleural effusion, with exceptions made for patients with typical congestive heart failure or for patients with “very small” pleural effusions.4 Ultrasound-guided thoracentesis is recommended for “small” pleural effusions and offers a lower risk for complications but at an increased cost.5 Clinicians encountering pleural effusions often ask radiologists to estimate the volume of PF present to determine whether thoracentesis can be performed safely at the patient’s bedside or whether ultrasound guidance is necessary. In 1996, Blackmore et al6 devised a prediction rule for estimating pleural effusion volume on the basis of the presence or absence of a meniscus on chest radiographs (CXRs). Using computed tomography (CT) as the gold standard, they calculated that a meniscus became visible on the lateral CXR at a volume of approximately 50 mL. The meniscus could be identified on the frontal radiograph at a volume of 200 mL, and the meniscus obscured the hemidiaphragm (HD) at a volume of about 500 mL. Although multiple authors have subsequently reported techniques for estimating PF volumes on sonography and CT,7–11 the CXR remains the most commonly obtained,12 and usually initial, imaging modality for most patients. The CXR volume predictions provided by Blackmore and colleagues remain in widespread use. Given the intervening
Key Words: pleural effusion, thoracic radiography, x-ray computed tomography, computer-assisted image processing
(J Thorac Imaging 2015;30:336–339)
F
luid accumulates in the space between the parietal and visceral pleural layers of the thorax whenever there is an increase in pleural fluid (PF) formation or a decrease in PF absorption. Using a pleural lavage technique, Noppen et al1 measured PF volumes in 5 subjects and determined that the normal amount of physiological fluid within the right and left hemithorax was 12.6 ± 5.3 mL (range: 3.9 to 17.9 mL) and 11.8 ± 6.8 mL (range:2.8 to 18.3 mL), respectively. Multiple pathologic processes, including congestive heart failure, infection, trauma, and malignancy, can lead to a pathologic accumulation of fluid in the pleural space.2,3 From the *Department of Radiology, Wake Forest University School of Medicine, Winston-Salem, NC; and wCenter for Injury Biomechanics, Virginia Tech-Wake Forest University, Blacksburg, VA and Winston-Salem, NC. The authors declare no conflicts of interest. Reprints: Caroline Chiles, MD, Department of Radiology, Wake Forest University School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157 (e-mail: [email protected]). Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
FIGURE 1. Scoring system for CXRs: A score of 0 corresponded to a normal CPA. A score of 1 indicated that a meniscus was visible and blunted the CPA below the level of the HD. A score of 2 indicated that the CPA was blunted at the level of the HD. A score of 3 indicated that the HD was completely obscured.
336 | www.thoracicimaging.com J Thorac Imaging Volume 30, Number 5, September 2015 Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Thorac Imaging
Volume 30, Number 5, September 2015
technological advances in CT as well as segmentation and volumetry since that report, we sought to update these PF volume estimates.13–15
MATERIALS AND METHODS A search of the radiology database utilizing word recognition software yielded 110 chest CT reports created between May 1, 2011 and July 13, 2012 for pleural effusions described as “small,” for which upright 2-view CXRs were
Estimation of Pleural Fluid Volumes
obtained within a 24-hour interval. Patients with only portable or single-view CXRs were excluded. None of the cases had thoracentesis or chest tube placement between the 2 imaging procedures. Twelve cases were excluded when review of either the CT or the CXR suggested that the fluid was loculated (N = 9) or subpulmonic (N = 3). The remaining 98 patients were selected for retrospective analysis, for a total of 196 hemithoraces. Pleural effusions could be either unilateral or bilateral, so the analysis included not only hemithoraces without PF present but also contralateral
FIGURE 2. Frontal (A) and lateral (B) CXRs demonstrate normal lateral and posterior CPAs (score = 0). Frontal (C) and lateral (D) CXRs demonstrate a meniscus blunting the right lateral and left posterior CPAs, at a level below the HD (score = 1). Frontal (E) and lateral (F) CXRs demonstrate blunting of the left lateral and posterior CPAs, with the meniscus at the level of the HD (score = 2). Frontal (G) and lateral (H) CXRs demonstrate PF obscuring the HD (score = 3).
Copyright
r
2015 Wolters Kluwer Health, Inc. All rights reserved.
www.thoracicimaging.com |
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
337
J Thorac Imaging
Mammarappallil et al
Volume 30, Number 5, September 2015
TABLE 1. PF Volumes as Determined by Volumetric CT on Both Frontal and Lateral CXRs, Reported as Mean, SD, and Minimum and Maximum
Frontal CXR CXR Score 0 1 2 3 Total
Mean Volume (SD) (mL) 19.76 139.72 266.02 642.66
(25.70) (214.64) (237.20) (377.12)
Lateral CXR
Range (mL)
N
0-185.14 0-928.64 16.34-741.51 98.54-1340.97
124 38 20 14 196
hemithoraces with pleural effusions that were larger than “small.” PF within each hemithorax was quantified using a semiautomatic method of image segmentation. The segmentation methodology relies on Hounsfield Unit (HU) levels in the image. Given that most pleural effusions are near water density, we selected pixels of attenuation values ranging from 20 to 20 HU within the pleural space. Nandalur et al16 found a mean PF attenuation of 15.7 HU (SD 5.4 HU), with slightly lower values in transudates (mean 12.5 HU, SD 6.3) than in exudates (mean 17.1 HU, SD 4.4). An upper limit of 20 HU was chosen to avoid including pleural thickening within the PF volume measurement. The pixels selected on each slice are of known sizes, ranging from 0.582 to 0.977 mm pixel width and height, on the basis of either a 488 488 or 512512 matrix size. The volumes were then calculated by multiplying the cross-sectional area by the voxel height. Voxel height was determined by the slice thickness of the CT scan. Scan slice thickness ranged from 0.5 to 2.5 mm. We developed a scoring system for the possible meniscus levels for both the frontal and lateral CXRs (Fig. 1). A score of 0 corresponded to a normal costophrenic angle (CPA). Scores of 1, 2, and 3 indicated that a meniscus was visible, and either blunted the CPA below the level of the HD, blunted the CPA at the level of the HD, or completely obscured the HD, respectively (Fig. 2). A cardiothoracic radiologist, blinded to the CT findings, reviewed the CXRs and scored each CPA from 0 to 3. The
FIGURE 3. Box plot demonstrating the volume of pleural effusion in the y axis vs. the corresponding CXR scores in the x axis for PA radiographs. Outlier data outside of the quartile determinations are designated as a solid black circle. PA indicates posteroanterior.
Mean Volume (SD) (mL) 17.23 101.81 248.68 642.66
(22.92) (185.59) (234.01) (377.12)
Range (mL)
N
0-185.14 0-928.64 0-741.51 98.54-1340.97
103 57 22 14 196
lateral and posterior CPAs were scored independently of each other, but at the same reading session. Comparison with prior CXRs was allowed when these were available. Each CXR category was then correlated with CT-determined PF volumes.
RESULTS PF volumes corresponding with CXR scores of 0 to 3 are shown in Table 1 and are reported as the mean PF volume in mL, SD, and minimum and maximum. A mean of 20 mL of PF was present on CT without a visible correlate in either the lateral or posterior CPA on CXR. A meniscus below the HD in either the lateral or posterior CPA on CXR correlated with roughly 100 mL; a meniscus occurring at the HD in either CPA correlated with roughly 250 mL; a meniscus obscuring the HD in either CPA correlated with a mean of approximately 650 mL. There were large standard deviations for all PF volumes. In addition, there were cases scored as 1 or 2 on the CXR but for which CT showed no PF. On a per-patient level, there were 22 cases in which the posterior CPA was scored as >0, and the lateral CPA was scored as 0. In these 22 cases, the mean PF volume was 26.0 mL, with an SD of 26.9 mL; total fluid volumes ranged from 0 to 84.1 mL. To further evaluate the data from our scoring system, box-plot graphs were generated, demonstrating the mean, the SD, the number of outliers, and the range of volume for each score (Figs. 3, 4). The average time between the 2 imaging studies was 6.53 hours (SD 6.86 h).
FIGURE 4. Box plot demonstrating the volume of pleural effusion in the y axis vs. the corresponding CXR scores in the x axis for lateral radiographs. Outlier data outside of the quartile determinations are designated as a solid black circle.
338 | www.thoracicimaging.com Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Thorac Imaging
Volume 30, Number 5, September 2015
DISCUSSION Radiologists are frequently asked to estimate the volume of PF present on a CXR. Estimation of PF volume can be helpful clinically in evaluating the need for pharmacologic or procedural intervention. Blackmore et al6 developed a visual prediction rule for PF volume estimation, using a test set of 24 pleural effusions in 16 patients, which they then validated in a set of 22 pleural effusions in 16 patients. Their grading system was at a patient-hemithorax level, with the score based on the combined features of the posterior and lateral CPAs. A score of zero correlated with no visible meniscus, a score of 1 represented a visible meniscus in the posterior CPA only, a score of 2 represented a visible meniscus in both the posterior and lateral CPAs, and a score of 3 represented a meniscus at the level of the HD including possible obscuration of the HD. We also used a 4-point scoring system but graded the posterior and lateral CPAs independently. In 1996, PF volume was determined by using an electronic cursor to outline the pleural effusion on 10-mmthick axial CT images.6 Each cross-sectional area was multiplied by 10 mm and then summed to yield total volume. For our analysis, slice thickness ranged from 0.5 to 2.5 mm. Both the thinner slices and the semiautomatic segmentation method available today provide greater precision in measurement of fluid volumes.13,17 The absence of a visible meniscus on either the frontal or lateral radiograph does not exclude the presence of a pleural effusion. We calculated that a mean of 20 mL of PF was present without a visible meniscus in either the posterior or lateral CPA, with up to 185 mL present in 1 case. This is comparable to the results of Blackmore et al,6 who observed 203 mL in 1 case without visible blunting of either CPA. These results are interesting given that over the last 18 years we would have expected the quality of the CXR to have been significantly improved resulting in better detection of pleural effusions. We speculate that some subpulmonic effusions do not produce the characteristic lateral shift of the HD on the frontal CXR and therefore go unrecognized. Our analysis confirms that the lateral CXR is more sensitive than the frontal for detection of small pleural effusions. There were 22 cases in which the posterior CPA was scored as >0, but there was no visible meniscus in the lateral CPA on the frontal radiograph. Total PF volumes in these cases were small, with a mean of 26 mL, in contrast to the approximately 100 mL of PF in patients with visible menisci below the level of the Hs in both lateral and posterior CPAs. Limitations of our study include the lack of lateral decubitus radiographs, which are even more sensitive for PF detection. Moskowitz et al18 determined that as little as 5 mL of fluid mixed with contrast material could be seen on cadaver radiographs performed in the lateral decubitus position. Additional limitations of our study are scoring of each CXR by a single observer and a relatively small number of cases scored as 2 or 3 (n = 20 and 14, respectively). These results are not generalizable to portable CXRs or those performed with the patient in a supine position. It is also possible that there could have been changes in PF volumes within the 24-hour interval allowed between the 2 imaging studies.
Copyright
r
Estimation of Pleural Fluid Volumes
The large standard deviations noted in our PF volume estimation suggest caution when using the visual prediction rule on a clinical basis. Radiologists should be aware that visible menisci within the CPAs may be artificially created by scarring, or muscle slips of the HD, as our analysis revealed cases scored as 1 or 2, without measureable fluid on the corresponding CT. In conclusion, we update the visual prediction rule for estimating PF volumes on upright frontal and lateral CXRs. We also confirm that the lateral radiograph is more sensitive for detection of small pleural effusions, with blunting of the posterior CPA only correlating with a mean of 26 mL of PF. REFERENCES 1. Noppen M, De Waele M, Li R, et al. Volume and cellular content of normal pleural fluid in humans examined by pleural lavage. Am J Respir Crit Care Med. 2000;162:1023–1026. 2. Porcel JM, Light RW. Pleural effusions. Dis Mon. 2013;59:29–57. 3. Light RW. Pleural effusions. Med Clin North Am. 2011;95: 1055–1070. 4. Sahn SA. The value of pleural fluid analysis. Am J Med Sci. 2008;335:7–15. 5. Jones PW, Moyers JP, Rogers JT, et al. Ultrasound-guided thoracentesis: is it a safer method? Chest. 2003;123:418–423. 6. Blackmore CC, Black WC, Dallas RV, et al. Pleural fluid volume estimation: a chest radiograph prediction rule. Acad Radiol. 1996;3:103–109. 7. Mergo PJ, Helmberger T, Didovic J, et al. New formula for quantification of pleural effusions from computed tomography. J Thorac Imaging. 1999;14:122–125. 8. Eibenberger KL, Dock WI, Ammann ME, et al. Quantification of pleural effusions: sonography versus radiography. Radiology. 1994;191:681–684. 9. Hazlinger M, Ctvrtlik F, Langova K, et al. Quantification of pleural effusion on CT by simple measurement. Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub. 2014;158: 107–111. 10. Moy MP, Levsky JM, Berko NS, et al. A new, simple method for estimating pleural effusion size on CT scans. Chest. 2013;143:1054–1059. 11. Moffett BK, Panchabhai TS, Anaya E, et al. Computed tomography measurements of parapneumonic effusion indicative of thoracentesis. Eur Respir J. 2011;38:1406–1411. 12. McAdams HP, Samei E, Dobbins J III, et al. Recent advances in chest radiography. Radiology. 2006;241:663–683. 13. von Falck C, Meier S, Jordens S, et al. Semiautomated segmentation of pleural effusions in MDCT datasets. Acad Radiol. 2010;17:841–848. 14. Yao J, Han W, Summers RM. Computer aided evaluation of pleural effusion using chest CT images. Proc IEEE Int Symp Biomed Imaging. 2009;2009:241–244. 15. Yao JBJ, Summers RM. Automatic segmentation and measurement of pleural effusions on CT. IEEE Trans Biomed Eng. 2013;60:1834–1840. 16. Nandalur KR, Hardie AH, Bollampally SR, et al. Accuracy of computed tomography attenuation values in the characterization of pleural fluid: an ROC study. Acad Radiol. 2005;12: 987–991. 17. Prionas ND, Ray S, Boone JM. Volume assessment accuracy in computed tomography: a phantom study. J Appl Clin Med Phys. 2010;11:3037–3051. 18. Moskowitz H, Platt RT, Schachar R, et al. Roentgen visualization of minute pleural effusion. An experimental study to determine the minimum amount of pleural fluid visible on a radiograph. Radiology. 1973;109:33–35.
2015 Wolters Kluwer Health, Inc. All rights reserved.
www.thoracicimaging.com |
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
339
Estimation of Pleural Fluid Volumes on Chest Radiography Using Computed Tomography Volumetric Analysis An Update of the Visual Prediction Rule Joseph G. Mammarappallil, MD, PhD,* Sarah A. Anderson,w Kerry A. Danelson, PhD,w Joel A. Stitzel, PhD,w and Caroline Chiles, MD* Purpose: The purpose of the study was to determine the volumes of pleural fluid (PF) required to produce visible menisci in the lateral and posterior costophrenic angles (CPA) and obscure the hemidiaphragms (HD) on upright frontal and lateral chest radiographs (CXRs), using volumetric analysis of chest computed tomography (CT). Materials and Methods: A total of 98 patients with small pleural effusions on chest CT, in whom CXRs were obtained within a 24hour interval, were selected for retrospective analysis. PF within each hemithorax was quantified using a semiautomatic method of image segmentation. A cardiothoracic radiologist scored each hemithorax on each CXR from 0 to 3 (0—normal CPA, 1—fluid meniscus below the HD, 2—fluid meniscus at the level of the HD, 3—fluid opacity obscures the HD). Each CXR category was correlated with CT-determined PF volumes. Results: A mean of 20 mL of PF was present on CT without a visible correlate on CXR. A meniscus below the HD on CXR correlated with roughly 100 mL; a meniscus occurring at the HD correlated with roughly 250 mL; a meniscus obscuring the HD correlated with a mean of approximately 650 mL. There were large standard deviations for all PF volumes. Conclusions: We provide guidelines for estimating PF volumes on upright frontal and lateral CXRs. We also confirm that the lateral radiograph is more sensitive for detection of small pleural effusions, with blunting of the posterior CPA only correlating with a mean of 26 mL of PF.
Thoracentesis is recommended for almost all patients presenting with a new pleural effusion, with exceptions made for patients with typical congestive heart failure or for patients with “very small” pleural effusions.4 Ultrasound-guided thoracentesis is recommended for “small” pleural effusions and offers a lower risk for complications but at an increased cost.5 Clinicians encountering pleural effusions often ask radiologists to estimate the volume of PF present to determine whether thoracentesis can be performed safely at the patient’s bedside or whether ultrasound guidance is necessary. In 1996, Blackmore et al6 devised a prediction rule for estimating pleural effusion volume on the basis of the presence or absence of a meniscus on chest radiographs (CXRs). Using computed tomography (CT) as the gold standard, they calculated that a meniscus became visible on the lateral CXR at a volume of approximately 50 mL. The meniscus could be identified on the frontal radiograph at a volume of 200 mL, and the meniscus obscured the hemidiaphragm (HD) at a volume of about 500 mL. Although multiple authors have subsequently reported techniques for estimating PF volumes on sonography and CT,7–11 the CXR remains the most commonly obtained,12 and usually initial, imaging modality for most patients. The CXR volume predictions provided by Blackmore and colleagues remain in widespread use. Given the intervening
Key Words: pleural effusion, thoracic radiography, x-ray computed tomography, computer-assisted image processing
(J Thorac Imaging 2015;30:336–339)
F
luid accumulates in the space between the parietal and visceral pleural layers of the thorax whenever there is an increase in pleural fluid (PF) formation or a decrease in PF absorption. Using a pleural lavage technique, Noppen et al1 measured PF volumes in 5 subjects and determined that the normal amount of physiological fluid within the right and left hemithorax was 12.6 ± 5.3 mL (range: 3.9 to 17.9 mL) and 11.8 ± 6.8 mL (range:2.8 to 18.3 mL), respectively. Multiple pathologic processes, including congestive heart failure, infection, trauma, and malignancy, can lead to a pathologic accumulation of fluid in the pleural space.2,3 From the *Department of Radiology, Wake Forest University School of Medicine, Winston-Salem, NC; and wCenter for Injury Biomechanics, Virginia Tech-Wake Forest University, Blacksburg, VA and Winston-Salem, NC. The authors declare no conflicts of interest. Reprints: Caroline Chiles, MD, Department of Radiology, Wake Forest University School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157 (e-mail: [email protected]). Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
FIGURE 1. Scoring system for CXRs: A score of 0 corresponded to a normal CPA. A score of 1 indicated that a meniscus was visible and blunted the CPA below the level of the HD. A score of 2 indicated that the CPA was blunted at the level of the HD. A score of 3 indicated that the HD was completely obscured.
336 | www.thoracicimaging.com J Thorac Imaging Volume 30, Number 5, September 2015 Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Thorac Imaging
Volume 30, Number 5, September 2015
technological advances in CT as well as segmentation and volumetry since that report, we sought to update these PF volume estimates.13–15
MATERIALS AND METHODS A search of the radiology database utilizing word recognition software yielded 110 chest CT reports created between May 1, 2011 and July 13, 2012 for pleural effusions described as “small,” for which upright 2-view CXRs were
Estimation of Pleural Fluid Volumes
obtained within a 24-hour interval. Patients with only portable or single-view CXRs were excluded. None of the cases had thoracentesis or chest tube placement between the 2 imaging procedures. Twelve cases were excluded when review of either the CT or the CXR suggested that the fluid was loculated (N = 9) or subpulmonic (N = 3). The remaining 98 patients were selected for retrospective analysis, for a total of 196 hemithoraces. Pleural effusions could be either unilateral or bilateral, so the analysis included not only hemithoraces without PF present but also contralateral
FIGURE 2. Frontal (A) and lateral (B) CXRs demonstrate normal lateral and posterior CPAs (score = 0). Frontal (C) and lateral (D) CXRs demonstrate a meniscus blunting the right lateral and left posterior CPAs, at a level below the HD (score = 1). Frontal (E) and lateral (F) CXRs demonstrate blunting of the left lateral and posterior CPAs, with the meniscus at the level of the HD (score = 2). Frontal (G) and lateral (H) CXRs demonstrate PF obscuring the HD (score = 3).
Copyright
r
2015 Wolters Kluwer Health, Inc. All rights reserved.
www.thoracicimaging.com |
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
337
J Thorac Imaging
Mammarappallil et al
Volume 30, Number 5, September 2015
TABLE 1. PF Volumes as Determined by Volumetric CT on Both Frontal and Lateral CXRs, Reported as Mean, SD, and Minimum and Maximum
Frontal CXR CXR Score 0 1 2 3 Total
Mean Volume (SD) (mL) 19.76 139.72 266.02 642.66
(25.70) (214.64) (237.20) (377.12)
Lateral CXR
Range (mL)
N
0-185.14 0-928.64 16.34-741.51 98.54-1340.97
124 38 20 14 196
hemithoraces with pleural effusions that were larger than “small.” PF within each hemithorax was quantified using a semiautomatic method of image segmentation. The segmentation methodology relies on Hounsfield Unit (HU) levels in the image. Given that most pleural effusions are near water density, we selected pixels of attenuation values ranging from 20 to 20 HU within the pleural space. Nandalur et al16 found a mean PF attenuation of 15.7 HU (SD 5.4 HU), with slightly lower values in transudates (mean 12.5 HU, SD 6.3) than in exudates (mean 17.1 HU, SD 4.4). An upper limit of 20 HU was chosen to avoid including pleural thickening within the PF volume measurement. The pixels selected on each slice are of known sizes, ranging from 0.582 to 0.977 mm pixel width and height, on the basis of either a 488 488 or 512512 matrix size. The volumes were then calculated by multiplying the cross-sectional area by the voxel height. Voxel height was determined by the slice thickness of the CT scan. Scan slice thickness ranged from 0.5 to 2.5 mm. We developed a scoring system for the possible meniscus levels for both the frontal and lateral CXRs (Fig. 1). A score of 0 corresponded to a normal costophrenic angle (CPA). Scores of 1, 2, and 3 indicated that a meniscus was visible, and either blunted the CPA below the level of the HD, blunted the CPA at the level of the HD, or completely obscured the HD, respectively (Fig. 2). A cardiothoracic radiologist, blinded to the CT findings, reviewed the CXRs and scored each CPA from 0 to 3. The
FIGURE 3. Box plot demonstrating the volume of pleural effusion in the y axis vs. the corresponding CXR scores in the x axis for PA radiographs. Outlier data outside of the quartile determinations are designated as a solid black circle. PA indicates posteroanterior.
Mean Volume (SD) (mL) 17.23 101.81 248.68 642.66
(22.92) (185.59) (234.01) (377.12)
Range (mL)
N
0-185.14 0-928.64 0-741.51 98.54-1340.97
103 57 22 14 196
lateral and posterior CPAs were scored independently of each other, but at the same reading session. Comparison with prior CXRs was allowed when these were available. Each CXR category was then correlated with CT-determined PF volumes.
RESULTS PF volumes corresponding with CXR scores of 0 to 3 are shown in Table 1 and are reported as the mean PF volume in mL, SD, and minimum and maximum. A mean of 20 mL of PF was present on CT without a visible correlate in either the lateral or posterior CPA on CXR. A meniscus below the HD in either the lateral or posterior CPA on CXR correlated with roughly 100 mL; a meniscus occurring at the HD in either CPA correlated with roughly 250 mL; a meniscus obscuring the HD in either CPA correlated with a mean of approximately 650 mL. There were large standard deviations for all PF volumes. In addition, there were cases scored as 1 or 2 on the CXR but for which CT showed no PF. On a per-patient level, there were 22 cases in which the posterior CPA was scored as >0, and the lateral CPA was scored as 0. In these 22 cases, the mean PF volume was 26.0 mL, with an SD of 26.9 mL; total fluid volumes ranged from 0 to 84.1 mL. To further evaluate the data from our scoring system, box-plot graphs were generated, demonstrating the mean, the SD, the number of outliers, and the range of volume for each score (Figs. 3, 4). The average time between the 2 imaging studies was 6.53 hours (SD 6.86 h).
FIGURE 4. Box plot demonstrating the volume of pleural effusion in the y axis vs. the corresponding CXR scores in the x axis for lateral radiographs. Outlier data outside of the quartile determinations are designated as a solid black circle.
338 | www.thoracicimaging.com Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Thorac Imaging
Volume 30, Number 5, September 2015
DISCUSSION Radiologists are frequently asked to estimate the volume of PF present on a CXR. Estimation of PF volume can be helpful clinically in evaluating the need for pharmacologic or procedural intervention. Blackmore et al6 developed a visual prediction rule for PF volume estimation, using a test set of 24 pleural effusions in 16 patients, which they then validated in a set of 22 pleural effusions in 16 patients. Their grading system was at a patient-hemithorax level, with the score based on the combined features of the posterior and lateral CPAs. A score of zero correlated with no visible meniscus, a score of 1 represented a visible meniscus in the posterior CPA only, a score of 2 represented a visible meniscus in both the posterior and lateral CPAs, and a score of 3 represented a meniscus at the level of the HD including possible obscuration of the HD. We also used a 4-point scoring system but graded the posterior and lateral CPAs independently. In 1996, PF volume was determined by using an electronic cursor to outline the pleural effusion on 10-mmthick axial CT images.6 Each cross-sectional area was multiplied by 10 mm and then summed to yield total volume. For our analysis, slice thickness ranged from 0.5 to 2.5 mm. Both the thinner slices and the semiautomatic segmentation method available today provide greater precision in measurement of fluid volumes.13,17 The absence of a visible meniscus on either the frontal or lateral radiograph does not exclude the presence of a pleural effusion. We calculated that a mean of 20 mL of PF was present without a visible meniscus in either the posterior or lateral CPA, with up to 185 mL present in 1 case. This is comparable to the results of Blackmore et al,6 who observed 203 mL in 1 case without visible blunting of either CPA. These results are interesting given that over the last 18 years we would have expected the quality of the CXR to have been significantly improved resulting in better detection of pleural effusions. We speculate that some subpulmonic effusions do not produce the characteristic lateral shift of the HD on the frontal CXR and therefore go unrecognized. Our analysis confirms that the lateral CXR is more sensitive than the frontal for detection of small pleural effusions. There were 22 cases in which the posterior CPA was scored as >0, but there was no visible meniscus in the lateral CPA on the frontal radiograph. Total PF volumes in these cases were small, with a mean of 26 mL, in contrast to the approximately 100 mL of PF in patients with visible menisci below the level of the Hs in both lateral and posterior CPAs. Limitations of our study include the lack of lateral decubitus radiographs, which are even more sensitive for PF detection. Moskowitz et al18 determined that as little as 5 mL of fluid mixed with contrast material could be seen on cadaver radiographs performed in the lateral decubitus position. Additional limitations of our study are scoring of each CXR by a single observer and a relatively small number of cases scored as 2 or 3 (n = 20 and 14, respectively). These results are not generalizable to portable CXRs or those performed with the patient in a supine position. It is also possible that there could have been changes in PF volumes within the 24-hour interval allowed between the 2 imaging studies.
Copyright
r
Estimation of Pleural Fluid Volumes
The large standard deviations noted in our PF volume estimation suggest caution when using the visual prediction rule on a clinical basis. Radiologists should be aware that visible menisci within the CPAs may be artificially created by scarring, or muscle slips of the HD, as our analysis revealed cases scored as 1 or 2, without measureable fluid on the corresponding CT. In conclusion, we update the visual prediction rule for estimating PF volumes on upright frontal and lateral CXRs. We also confirm that the lateral radiograph is more sensitive for detection of small pleural effusions, with blunting of the posterior CPA only correlating with a mean of 26 mL of PF. REFERENCES 1. Noppen M, De Waele M, Li R, et al. Volume and cellular content of normal pleural fluid in humans examined by pleural lavage. Am J Respir Crit Care Med. 2000;162:1023–1026. 2. Porcel JM, Light RW. Pleural effusions. Dis Mon. 2013;59:29–57. 3. Light RW. Pleural effusions. Med Clin North Am. 2011;95: 1055–1070. 4. Sahn SA. The value of pleural fluid analysis. Am J Med Sci. 2008;335:7–15. 5. Jones PW, Moyers JP, Rogers JT, et al. Ultrasound-guided thoracentesis: is it a safer method? Chest. 2003;123:418–423. 6. Blackmore CC, Black WC, Dallas RV, et al. Pleural fluid volume estimation: a chest radiograph prediction rule. Acad Radiol. 1996;3:103–109. 7. Mergo PJ, Helmberger T, Didovic J, et al. New formula for quantification of pleural effusions from computed tomography. J Thorac Imaging. 1999;14:122–125. 8. Eibenberger KL, Dock WI, Ammann ME, et al. Quantification of pleural effusions: sonography versus radiography. Radiology. 1994;191:681–684. 9. Hazlinger M, Ctvrtlik F, Langova K, et al. Quantification of pleural effusion on CT by simple measurement. Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub. 2014;158: 107–111. 10. Moy MP, Levsky JM, Berko NS, et al. A new, simple method for estimating pleural effusion size on CT scans. Chest. 2013;143:1054–1059. 11. Moffett BK, Panchabhai TS, Anaya E, et al. Computed tomography measurements of parapneumonic effusion indicative of thoracentesis. Eur Respir J. 2011;38:1406–1411. 12. McAdams HP, Samei E, Dobbins J III, et al. Recent advances in chest radiography. Radiology. 2006;241:663–683. 13. von Falck C, Meier S, Jordens S, et al. Semiautomated segmentation of pleural effusions in MDCT datasets. Acad Radiol. 2010;17:841–848. 14. Yao J, Han W, Summers RM. Computer aided evaluation of pleural effusion using chest CT images. Proc IEEE Int Symp Biomed Imaging. 2009;2009:241–244. 15. Yao JBJ, Summers RM. Automatic segmentation and measurement of pleural effusions on CT. IEEE Trans Biomed Eng. 2013;60:1834–1840. 16. Nandalur KR, Hardie AH, Bollampally SR, et al. Accuracy of computed tomography attenuation values in the characterization of pleural fluid: an ROC study. Acad Radiol. 2005;12: 987–991. 17. Prionas ND, Ray S, Boone JM. Volume assessment accuracy in computed tomography: a phantom study. J Appl Clin Med Phys. 2010;11:3037–3051. 18. Moskowitz H, Platt RT, Schachar R, et al. Roentgen visualization of minute pleural effusion. An experimental study to determine the minimum amount of pleural fluid visible on a radiograph. Radiology. 1973;109:33–35.
2015 Wolters Kluwer Health, Inc. All rights reserved.
www.thoracicimaging.com |
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
339
