Comparison of Quality of Multiplanar Reconstructions and Direct Coronal Multidetector CT Scans of the Lung

Transcription

1 Osamu Honda 1 Takeshi Johkoh 1 Shuji Yamamoto 2 Mitsuhiro Koyama 1 Noriyuki Tomiyama 1 Takenori Kozuka 1 Seiki Hamada 1 Naoki Mihara 1 Hironobu Nakamura 1 Nestor L. Müller 3 Received January 2, 22; accepted after revision March 29, Department of Radiology, Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka, , Japan. Address correspondence to O. Honda. 2 Department of Radiology, Osaka University Hospital, Osaka, , Japan. 3 Department of Radiology, Vancouver Hospital and Health Sciences Centre and University of British Columbia, 855 W. 12th Ave., Vancouver, B. C. V5Z 1M9, Canada. AJR 22;179: X/2/ American Roentgen Ray Society Comparison of Quality of Multiplanar Reconstructions and Direct Coronal Multidetector CT Scans of the Lung OBJECTIVE. The purpose of this study was to compare the quality of coronal multiplanar reconstructions with the quality of direct coronal thin-section multidetector CT (MDCT) scans. MATERIALS AND METHODS. Axial multidetector CT (MDCT) scans were obtained through the entire lung in 1 normal autopsy lung specimens using an MDCT scanner. Four protocols were used:.5-mm collimation with a.5-mm reconstruction interval;.5-mm collimation with a.3-mm reconstruction interval; 1-mm collimation with a.5-mm reconstruction interval; and 2-mm collimation with a 1-mm reconstruction interval. Multiplanar reconstruction images with.5-mm slice thickness were obtained from the four types of data sets. Direct coronal thin-section CT of the same 1 autopsy lung specimens was performed using.5-mm scan collimation, a.3-mm reconstruction interval, a 25.6-cm field of view, and a matrix. Two independent observers compared the image quality of each of the four coronal multiplanar reconstruction sets with that of direct coronal thin-section CT scans. The observers analyzed visualization of anatomic features and artifacts. RESULTS. The total image quality of the multiplanar reconstructions obtained from.5-mm collimation data with or without.3-mm overlapping reconstruction was equal to that of direct coronal thin-section CT scans in all 2 interpretations. The image quality of multiplanar reconstruction images from.5-mm collimation data either with or without overlapping reconstruction was superior to multiplanar reconstruction images obtained from 1- or 2-mm collimation scans ( p <.1, Fisher s exact test). Stairstep artifacts in multiplanar reconstructions using.5-mm collimation without overlapping reconstruction were equal to those with overlapping reconstruction and were fewer than those on 1- or 2-mm collimation (p <.1, Mann-Whitney U test). CONCLUSION. The image quality of coronal multiplanar reconstructions from isotropic voxel data obtained using.5-mm collimation, with or without overlapping reconstruction, is similar to that of direct coronal thin-section CT scans. T o improve the quality of multiplanar reconstruction images, axial multidetector CT (MDCT) images must be reconstructed in an overlapping fashion [1 3]. In 1995, Kalender [1] predicted that isotropic scanning would become possible. MDCT scanners can rapidly scan large longitudinal (z) volumes with high z-axis resolution [4], and MDCT scanners show smaller stairstep artifacts than single-detector CT scanners [5]. MDCT scanners allow contiguous thin-section CT to be performed through the entire lung and isotropic imaging that is, equal spatial resolution in the x y plane and the z-axis. MDCT scanners can potentially allow the depiction of images in any desired plane with a resolution similar to that in the transverse plane. Optimal resolution on thin-section CT is achieved using the highest matrix size ( ) and the lowest field of view that will include the entire region of interest. CT performed using 1-mm collimation and a matrix requires a field of view of 51.2 cm to obtain isotropic voxel data sets. The introduction of MDCT scanners with the ability to obtain.5-mm-thick sections allows the reduction of the field of view to 25.6 cm. To our knowledge, no study has evaluated the image quality of multiplanar reconstruction lung images obtained from isotropic voxel data sets. The purpose of this study was to compare the image quality of coronal multiplanar reconstructions of MDCT scans performed through normal autopsy lung specimens with that of direct coronal thin-section MDCT scans. AJR:179, October

2 Honda et al. Materials and Methods Ten lungs of patients who did not have any history of pulmonary disease were autopsied, inflated, and fixed by the method of Markarian and Dailey [6]. The lungs were distended through the main bronchus with a fixative fluid that contained polyethylene glycol 4, 95% ethyl alcohol, 4% formalin, and water in a proportion of 1:5:2:3. The specimens were immersed in the fixative for 2 days. The fixed lungs were then air-dried. Four of these 1 lungs were right lungs and six were left lungs. Axial continuous MDCT scans were obtained on an MDCT scanner (Aquillion-VZ; Toshiba, Tokyo, Japan) using the following protocols:.5-mm collimation with a.5-mm reconstruction interval;.5-mm collimation with a.3-mm reconstruction interval; 1-mm collimation with a.5-mm reconstruction interval; and 2-mm collimation with a 1- mm reconstruction interval. The scanned distances were all 25.6 cm. All scans were obtained using a 6:1 pitch,.5 sec per rotation, 12 kv, 1 mas, a cm field of view, a matrix, and a high-frequency reconstruction algorithm. As the control for image quality, direct coronal thin-section CT scans of the same 1 autopsied lungs were obtained on the same MDCT scanner. Direct coronal thin-section CT was performed using.5-mm scan collimation, a 6:1 pitch, a.3-mm reconstruction interval,.5 sec per rotation, 12 kv, 1 mas, a 25.6-cm field of view, a matrix, and a high-frequency reconstruction algorithm. All images were transferred to a commercially available workstation (Advantage Windows 3.; General Electric Medical Systems, Milwaukee, WI), and coronal multiplanar reconstruction images with.5-mm slice thickness and.5-mm slice intervals were obtained from the four axial MDCT data sets. All coronal multiplanar reconstruction images were reconstructed to fit the direct coronal thin-section CT images on the workstation. All four sets of coronal multiplanar reconstructions and the direct coronal thin-section CT scans were photographed separately on hard copies using a window width of 12 H and a window level of 7 H. After the CT examination, transverse sections were obtained from the same lungs every 1 mm and were stained with H and E and elastica van Gieson stains for histopathologic assessment. Macroscopic and microscopic assessment of the lungs showed no parenchymal abnormalities. Image quality of each coronal multiplanar reconstruction image was evaluated as being equal to or inferior to that of the direct coronal thin-section CT scan by two independent pulmonary radiologists who paid special attention to visualization of the following six normal structures: the interlobular septa, the central bronchovascular bundles coursing vertically, the central bronchovascular bundles coursing horizontally, the peripheral bronchovascular bundles, the interlobar fissures, and the pulmonary parenchyma. In addition, the total image quality of each coronal multiplanar reconstruction image was subjectively evaluated as being equal to or inferior to that of direct coronal thin-section CT scan. Stairstep artifacts of each multiplanar reconstruction image were graded as 1, none; 2, mild; or 3, severe, compared with the direct coronal thin-section CT scan. We required that at least 4 weeks elapse between each set of interpretations. The difference in image quality of each multiplanar reconstruction image was analyzed using Fisher s exact probability test. The scores of stairstep artifacts of each multiplanar reconstruction were evaluated using the Mann-Whitney U test. The interobserver variation of the image quality between two reviewers was evaluated using kappa statistics. The interobserver agreement was classified as follows: poor,.2; fair,.21.4; moderate,.41.6; good,.61.8; and excellent, The interobserver variation for noise and stairstep artifacts was analyzed using Spearman s correlation coefficient. Results The two observers had good agreement (κ =.76) for the evaluation of image quality and stairstep artifacts (Spearman s correlation coefficient, r >.7; p <.1). Therefore, the statistical analysis of the comparison of image quality among the various sets of multiplanar reconstruction images was performed by combining the results of the two observers. TABLE 1 Quality of Multiplanar Reconstructions and Direct Coronal Thin-Section CT Images Multiplanar Reconstruction Protocol Interlobular Septa Coronal Thin-Section CT Images (n = 2) Bronchovascular Bundles The image quality results are summarized in Table 1. The total image quality of multiplanar reconstructions obtained from the.5-mm collimation data sets with or without.3-mm overlapping reconstruction was equal to that of the direct coronal thin-section CT scans in all 2 interpretations (1 lungs, two observers). The image quality of the other multiplanar reconstructions from.5-mm collimation data sets with or without.3-mm overlapping reconstructions was also equal to that of direct coronal thin-section CT scans except in one case of interlobular septa. The total image quality of multiplanar reconstruction images from.5- mm collimation data sets either with or without overlapping reconstruction was superior to that of multiplanar reconstructions obtained from 1- and 2-mm collimation data sets (Fisher s exact probability test, p <.1) (Fig. 1). Stairstep artifacts in multiplanar reconstruction images are summarized in Table 2. The scores of stairstep artifacts in multiplanar reconstructions obtained from.5-mm collimation data sets without overlapping reconstruction were equal to those obtained from.5-mm collimation data sets with overlapping reconstruction. The scores of stairstep artifacts in multiplanar reconstructions obtained from.5- mm collimation data sets with overlapping reconstruction were superior to those from 1- and 2-mm collimation data sets (Mann-Whitney U test, p <.1). Discussion Helical CT has been shown to have a spatial resolution similar to that of conventional CT in the imaging plane and a greater resolution than that of conventional CT in the z- axis [1]. However, until recently multiplanar reconstruction or three-dimensional images from axial helical CT data sets have been accompanied by stairstep artifacts along the z- axis [7]. Resolution in the z-axis can be Major or Minor Fissure Pulmonary Parenchyma Total Image Quality Vertical Central Horizontal Central Peripheral.5-mm collimation,.3-mm overlap mm collimation, no overlap mm collimation,.5-mm overlap mm collimation, 1-mm overlap 1 Note. Total quality of multiplanar reconstructions from.5-mm-collimation data sets either with or without overlapping reconstruction was superior to that of multiplanar reconstructions obtained from 1- and 2-mm-collimation data sets (Fisher s exact probability test, p <.1). Image quality of multiplanar reconstructions from.5-mm-collimation data sets either with or without overlapping reconstruction was superior to that from 1- and 2-mm-collimation data sets. 876 AJR:179, October 22

3 Multidetector CT and Multiplanar Reconstructions of the Lung A B Fig. 1. Lung of 68-year-old male cadaver. A, Direct coronal thin-section multidetector CT image of lung with.5-mm collimation and pitch of 6:1 clearly shows interlobular septum (arrowhead ), central and peripheral bronchovascular bundles, and major fissure (arrow ). B, Coronal multiplanar reconstruction image from.5-mm-collimation data sets with.3-mm overlapping reconstruction shows image quality similar to that of A. C, Coronal multiplanar reconstruction image from.5-mm-collimation data set without overlapping reconstruction shows image quality similar to that of B. (Fig. 1 continues on next page) C AJR:179, October

4 Honda et al. D E Fig. 1. (continued) Lung of 68-year-old male cadaver. D, On coronal multiplanar reconstruction image from 1-mm-collimation data set with.5-mm overlapping reconstruction, interlobular septum (arrowhead ) is ambiguous compared with direct coronal thin-section CT image (A) and coronal multiplanar reconstruction images from.5-mm-collimation data set with (B) and without (C).3-mm overlapping reconstruction. E, Coronal multiplanar reconstruction image from 2-mm-collimation data set with 1-mm overlapping reconstruction shows considerable noise. Interlobular septum, central and peripheral bronchovascular bundles, and major fissure (arrow ) are ambiguous. improved by using overlapping reconstructions [1, 2]. However, depiction of anatomic structures in arbitrarily chosen planes with the same image quality as in the transverse plane requires isotropic imaging. In this study, we used.5-mm collimation, a 25.6-cm field of view, and a matrix. Pixel size in the x y plane was calculated as follows: 256-mm field of view / 512 pixels =.5 mm. Therefore, isotopic imaging has been theoretically achieved. The results of the study showed similar image quality for coronal multiplanar images reconstructed from axial scans obtained using.5-mm collimation and for direct coronal CT scans. Furthermore, these multiplanar reconstructions had no stairstep artifacts. These findings confirm the hypothesis that isotropic imaging was achieved. Caldemeyer et al. [8] reported that the diagnostic efficacy of coronal multiplanar reconstruction images with.5-mm collimation on dual-detector helical CT was comparable to or superior to that of direct coronal CT scans with mm collimation on temporal bone. Venema et al. [9] also reported that coronal multiplanar reconstructions from axial dual-detector helical CT obtained with.5-mm collimation could replace direct coronal sequential CT scans of petrosal bone. In their study, overlapping reconstruction was used. However, our results suggest that overlapping reconstruction is not necessary for coronal multiplanar recon- TABLE 2 structions from isotropic voxel data sets for lung parenchyma. Our study has several limitations. First, because we used autopsy lung specimens, it was not possible to evaluate the influence of motion artifacts. However, ethical concerns exist about performing repeated CT in any given patient. Furthermore, it is not possible to obtain direct coronal images of human lungs using Noise and Stairstep Artifacts in Multiplanar Reconstructions Stairstep Artifacts (n = 2) Multiplanar Reconstruction Protocol.5-mm collimation,.3-mm overlap None Mild Severe mm collimation, no overlap mm collimation,.5-mm overlap mm collimation, 1-mm overlap Note. Data are numbers of reconstructions in each scoring category. Multiplanar reconstructions obtained from.5-mmcollimation data sets without overlapping reconstruction were equal to those from.5-mm-collimation data sets with overlapping reconstruction (Mann-Whitney U test, p <.5). Multiplanar reconstructions obtained from.5-mm-collimation data sets with overlapping reconstruction were superior to those from 1- and 2-mm-collimation data sets (Mann-Whitney U test, p <.1). AJR:179, October 22

5 Multidetector CT and Multiplanar Reconstructions of the Lung commercially available CT scanners. Second, the study was limited to normal lungs. It was reported that MDCT high-resolution CT images of various modes were compared with single-detector high-resolution images in the axial mode, and high-resolution MDCT images of a 6:1 pitch were unsuitable for detecting subtle lung abnormalities on axial MDCT images of and 2.5-mm collimation [1]. In our study, we used a pitch of 6:1 and.5-mm collimation. We chose.5-mm collimation to obtain the multiplanar reconstruction images from isotropic voxels, and a 6:1 pitch to cover a large area in one breath-hold. If axial high-resolution CT is used for direct coronal high-resolution CT instead of a high-resolution MDCT pitch of 6:1, multiplanar reconstruction images from MDCT data of a 6:1 pitch would be inferior to direct coronal high-resolution axial CT scans. Further study is needed with scanning of abnormal lungs, especially lungs with subtle abnormalities. If more rapid scanning could be achieved, it would be possible to obtain superior multiplanar reconstruction images of the whole lung from isotropic voxel data. The third potential limitation is a possible bias because the observers knew which of the images were obtained using the direct coronal technique, which may have biased them to make a superior evaluation of the multiplanar reconstruction images with.5-mm collimation. However, comparing the images with direct coronal images has merit to test for the inferiority of multiplanar reconstruction images as well. As for the multiplanar reconstructions, the observers did not know how the images were scanned and reconstructed; therefore, their findings that the image quality of coronal multiplanar reconstruction images from isotropic voxel data sets without overlapping reconstruction is almost identical to that with overlapping reconstruction are reliable. In conclusion, the results of our study indicate that the image quality of coronal multiplanar reconstructions obtained from isotropic voxel data sets is similar to the image quality of direct coronal thin-section CT scans. The image quality of coronal multiplanar reconstructions from isotropic voxel data sets without overlapping reconstruction is almost identical to that of sets with overlapping reconstruction. References 1. Kalender WA. Thin-section three-dimensional spiral CT: is isotropic imaging possible? Radiology 1995;197: Kasales CJ, Hopper KD, Ariola DN, et al. Reconstructed helical CT scans: improvement in z-axis resolution compared with overlapped and nonoverlapped conventional CT scans. AJR 1995; 164: Wang G, Vannier MW. Stair-step artifacts in three-dimensional helical CT: an experimental study. Radiology 1994;191: Hu H. Multi-slice helical CT: scan and reconstruction. Med Phys 1999;26: Fleischmann D, Rubin GD, Paik DS. Stair-step artifacts with single versus multiple detector-row helical CT. Radiology 2;216: Markarian B, Dailey ET. Preparation of inflated lung specimens. In: Heitzman ER, ed. The lung: radiologic pathologic correlations, 3rd ed. St. Louis: Mosby, 1993: Brink JA. Technical aspects of helical (spiral) CT. Radiol Clin North Am 1995;33: Caldemeyer KS, Sandrasegaran K, Shinaver CN, Mathews VP, Smith RR, Kopecky KK. Temporal bone: comparison of isotropic helical CT and conventional direct axial and coronal CT. AJR 1999;172: Venema HW, Phoa SS, Mirck PG, Hulsmans FJ, Majoie CB, Verbeeten B Jr. Petrosal bone: coronal reconstructions from axial spiral CT data obtained with.5-mm collimation can replace direct coronal sequential CT scans. Radiology 1999;213: 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 21;177: AJR:179, October