Gastrointestinal Imaging Original Research

Transcription

1 Gastrointestinal Imaging Original Research Han et al. CT of Patients With Metal Hip Prostheses Gastrointestinal Imaging Original Research Seung Chol Han 1 Yong Eun Chung 1 Young Han Lee 1 Kwan Kyu Park 2 Myeong Jin Kim 1 Ki Whang Kim 1 Han SC, Chung YE, Lee YH, Park KK, Kim MJ, Kim KW Keywords: abdominal dual-energy CT, Metal Artifact Reduction software, metal hip prosthesis, pelvic cavity, phantom DOI: /AJR Received March 28, 2013; accepted after revision December 29, This work was supported by Yonsei University Research Fund (grant ). 1 Department of Radiology, Research Institute of Radiological Science, Severance Hospital, Yonsei University College of Medicine, 50 Yonsei-ro, Seodaemun-gu, Seoul , Republic of Korea. Address correspondence to Y. E. Chung (yelv@yuhs.ac). 2 Department of Orthopaedic Surgery, Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea. This article is available for credit. AJR 2014; 203: X/14/ American Roentgen Ray Society Metal Artifact Reduction Software Used With Abdominopelvic Dual-Energy CT of Patients With Metal Hip Prostheses: Assessment of Image Quality and Clinical Feasibility OBJECTIVE. The objective of our study was to determine the feasibility of using Metal Artifact Reduction (MAR) software for abdominopelvic dual-energy CT in patients with metal hip prostheses. MATERIALS AND METHODS. This retrospective study included 33 patients (malefemale ratio, 19:14; mean age, 63.7 years) who received total hip replacements and 20 patients who did not have metal prostheses as the control group. All of the patients underwent dualenergy CT. The quality of the images reconstructed using the MAR algorithm and of those reconstructed using the standard reconstruction was evaluated in terms of the visibility of the bladder wall, pelvic sidewall, rectal shelf, and bone-prosthesis interface and the overall diagnostic image quality with a 4-point scale. The mean and SD attenuation values in Hounsfield units were measured in the bladder, pelvic sidewall, and rectal shelf. For validation of the MAR interpolation algorithm, pelvis phantoms with small bladder lesions and metal hip prostheses were made, and images of the phantoms both with and without MAR reconstruction were evaluated. RESULTS. Image quality was significantly better with MAR reconstruction than without at all sites except the rectal shelf, where the image quality either had not changed or had worsened after MAR reconstruction. The mean attenuation value was changed after MAR reconstruction to its original expected value at the pelvic sidewall (p < 0.001) and inside the bladder (p < 0.001). The SD attenuation value was significantly decreased after MAR reconstruction at the pelvic sidewall (p = 0.019) but did not show significant differences at the bladder (p = 0.173) or rectal shelf (p = 0.478). In the phantom study, all lesions obscured by metal artifacts on the standard reconstruction images were visualized after MAR reconstruction; however, new artifacts had developed in other parts of the MAR reconstruction images. CONCLUSION. The use of MAR software with dual-energy CT decreases metal artifacts and increases diagnostic confidence in the assessment of the pelvic cavity but also introduces new artifacts that can obscure pelvic structures. W ith the number of elderly adults in society increasing, the number of patients with hip joint problems undergoing total hip replacement has gradually increased. In the United States, 231,000 total hip replacements were performed in 2007, and in Korea this number increased by approximately 39.9% from 2007 to 2011 [1, 2]. Metal hip prostheses can cause artifacts on CT images that appear as dark and bright streaks or bands [3]. From a radiologic perspective, metal artifacts due to hip prostheses can significantly limit the accuracy of assessment because they could be masking abnormal focal lesions [3]. Despite recent developments in CT technology, overcoming the inclusion of metal ar- tifacts in CT still presents a challenge. Since the early 1980s, several methods have been developed, including increasing the tube current (ma) or tube voltage (kvp), using adaptive filtering, and using iterative methods, for reducing metal artifacts and improving image quality [4]. Although these techniques have been somewhat successful in reducing metal artifacts, the clinical use of these techniques is limited because of their associated increases in radiation dose, decreases in lowcontrast resolution, and increases in computational times [4]. Recently, metal artifact reduction techniques with dual-energy CT have been reported to effectively reduce metal artifacts [5, 6]. However, most studies on metal artifact reduction, including those us- 788 AJR:203, October 2014

2 CT of Patients With Metal Hip Prostheses ing dual-energy techniques, have focused on the bone-prosthesis interface and periprosthetic areas [5 8]. Metal artifacts caused by hip prostheses could affect diagnostic accuracy in the following areas of the pelvic cavity: the rectal shelf, which is one of the most common areas of seeding metastases; the pelvic sidewall, which is a lymph node bearing area drained from the lower pelvic region; and the bladder wall, which may harbor primary urothelial carcinomas or metastases. Considering that primary or metastatic cancers more commonly occur in older populations, who are also more likely to need metal hip prostheses, the feasibility of a metal artifact reduction algorithm for the evaluation of pelvic lesions should be evaluated separately. One study found that metal artifact reduction could be effectively applied in the pelvic cavity [6]. In that study, metal artifact detection and segmentation were performed using reformatted projection views and 2D interpolation was used to fill in the missing data [6]. However, this metal artifact reduction algorithm is still in the research phase and is not yet commercially available. Recently, commercially available Metal Artifact Reduction (MAR) software (GE Healthcare) with gemstone spectral imaging (GSI) CT which reduces metal artifacts through metal segmentation, the estimation of missing information, and the completion of dual-energy data has been developed. The use of MAR software with GSI dual-energy CT have been shown to be feasible and to improve the delineation of prostheses and the periprosthetic area [5]. However, none of the studies to date has examined the performance of the MAR software with GSI dual-energy CT for the evaluation of the pelvic cavity. We hypothesized that the MAR application could improve the quality of pelvic cavity images by reducing metal artifacts. Hence, the purpose of this study was to verify the feasibility of the MAR algorithm with GSI dual-energy CT for the assessment of the pelvic cavity in patients with metal hip prostheses by evaluating images of clinical patients and a pelvis phantom. Materials and Methods Patients This single-center study was conducted with the approval of our hospital institutional review board. Informed consent was waived because of the retrospective design of this study. Patients who had unilateral or bilateral metal hip prostheses or screws for internal fixation at the femur neck and who underwent abdominal GSI dual-energy CT from January 2012 to June 2012 were included in this study. For comparison, we also assessed 20 additional patients without a metal prosthesis in the abdominal cavity as the control group. Metal Artifact Reduction Software With Gemstone Spectral Imaging Dual-Energy CT Although the precise MAR algorithm is not available to end-users, previous articles [9, 10] have presented an overview of the technical aspects of the MAR software. GSI dual-energy CT data were obtained for MAR reconstruction simply by using predefined MAR protocols in the console. After data acquisition, interleaved high-kvp and low-kvp projections were split into low-kvp and high-kvp projections. Metal segmentation and metal mask creation were performed using the high-kvp projections because the boundaries between the metal and the adjacent tissue were better depicted and because more accurate definition of the metal shape could be obtained than with the low-kvp data. After the removal of the metal-contaminated samples from the projection domain in both high-kvp and low-kvp data, the MAR algorithm applies an iterative process to correct for the corrupted data. This process includes the estimation of missing samples by the forward projection of the segmented metal images, reconstruction of a preliminary image with the estimated projection, and formulation of the final image with the metal mask and preliminary image. There are several distinct technical characteristics of the GSI MAR algorithm. First, the GSI MAR algorithm takes full advantage of the inherent beam-hardening correction capability of GSI and focuses only on the correction of the photon starvation aspect of metal artifacts. Projection samples contaminated by the beam-hardening effect contain valuable information and can be fully used in the reconstruction after correction, whereas data corrupted by photon starvation have little useful information and should be replaced. Second, using high-kvp projection data allows better estimation of metal segmentation. The improved metal segmentation also benefits the metal mask generation process of the low kvp. Third, the MAR processing and the reconstruction of the GSI dataset can be combined to produce a final image that is better than the images that either algorithm could produce alone in terms of artifact reduction. CT Protocol All studies were performed with a GSI dual-energy 64-MDCT scanner (Discovery CT750HD, GE Healthcare). Contrast medium (2 ml/kg of body weight) was injected via the antecubital vein using a power injector with a fixed injection duration of 30 seconds. A bolus-tracking technique was used, and the portal venous phase images were obtained 55 seconds after the attenuation measurement reached 100 HU in the abdominal aorta. The CT parameters were as follows: preset protocol, GSI 1 (rotation time, 0.5 second; detector coverage, 40 cm; volume CT dose index, mgy); tube voltage, fast kilovoltageswitching between 80 and 140 kvp; tube current, less than 630 ma; pitch, 1.375; and slice thickness, 1.25 mm. Imaging data were sent to a workstation (Advantage, GE Healthcare) and monochromatic images with energy levels of 77 kev were reconstructed with and without the use of the MAR algorithm (reconstruction thickness, 2.5 mm). GSI dual-energy CT can generate material-density images that show the amount of two basis material contents by mathematic calculation of the acquired datasets from high-energy and low-energy data. Monochromatic (kiloelectron volt) images were then synthesized from materialdensity images, and mass attenuation coefficients of each material for energy levels between 40 and 140 kev were calculated with dedicated software (GSI Viewer, version 2.0, GE Healthcare) [11]. An energy level of 77 kev was chosen because, at that level, the monochromatic image could be adjusted to approximate the contrast of a 120-kVp image. Finally, monochromatic image sets were sent to a PACS (Centricity Radiology RA1000, GE Healthcare) for image review. Objective Image Analysis Quantitative image analysis was performed with a PACS workstation. ROIs were drawn as single circles positioned within the bladder at the ipsilateral pelvic sidewall and rectal shelf, and the mean and SD attenuation values in Hounsfield units were measured. The location of the ROI was carefully selected, and the ROI was drawn to be as large as possible without including adjacent organs or structures. The ROI of the pelvic sidewall was located in the fat at the level of the femoral head. The ROI of the rectal shelf was located in the fat of the rectouterine pouch in women and at the rectovesical pouch in men. The ROI of the bladder was drawn within the bladder. Subjective Image Analysis Two radiologists (one with 6 years of experience and the other with 5 years of experience in abdominal radiology) independently reviewed two image sets at 1-month intervals and with random ordering to avoid recall bias. Images were evaluated in the same window settings (window level, 60 HU; window width, 340 HU). The degree of the metal artifacts was assessed at the following sites: the contralateral and ipsilateral bladder wall, the contralateral and ipsilateral pelvic sidewall, the rectal shelf, and the bone-prosthesis interface. The ipsilateral side was defined as the side with the metal prosthesis. If metal pros- AJR:203, October

3 Han et al. TABLE 1: Quantitative Analysis of Quality of Images Reconstructed With Standard Reconstruction and With Metal Artifact Reduction (MAR a ) Software Criteria for Quantitative Analysis theses were present in both hips, the ipsilateral side was defined as the right side of the patient. The degree of metal artifacts at each site and the overall diagnostic image quality were evaluated using the following 4-point scale: 1, no artifacts affecting image interpretation; 2, minor artifacts not interfering with diagnostic decision making; 3, major artifacts affecting visualization of major structures, but diagnosis is possible; and 4, severe artifacts that deter accurate diagnosis. Phantom Study Because there were no patients with focal lesions in the pelvic cavity in our study group, a phantom study was designed to validate the MAR interpolation algorithm for visualization of focal lesions masked by metal artifacts. The pelvic bones of an ox were obtained from an abattoir. A water-filled balloon with a diameter of approximately 12.5 cm was used to simulate the human bladder. Small balloons were filled with a mixture of iodine contrast material and water (90 HU) to mimic focal lesions such as urothelial carcinomas or metastases with sizes of 1 and 2 cm. The large balloon mimicking the bladder was inverted, and the small balloons mimicking the focal lesions were attached to the inner sides of the large balloon using glue. The balloon was then reinverted and filled with water, with careful attention not to detach the small balloons from the large one. The pelvic bone was placed in an enclosed plastic container, and the bladder balloon was placed within the pelvic cavity and fixed with an adhesive bandage so that the locations of the focal lesions could be easily changed (Fig. 1). The metal prosthesis was placed in the acetabulum of the pelvic bone and fixed with clay. The location Patient Group Standard Reconstruction Images b MAR Reconstruction Images b of the clay was carefully selected so that the clay would not be included in the FOV of the CT scans. The phantom was imaged with the same imaging parameters used in the patient study. We obtained CT scans under several presumptive situations by adjusting the location of the balloons as follows: situation 1, a 1-cm lesion fully obscured by a dark streak; situation 2, a 2-cm lesion partly obscured by a dark streak at its central portion; situation 3, a 2-cm lesion partly obscured by a dark streak at its peripheral portion; and situation 4, a metal prosthesis placed on both sides and a 2-cm lesion fully obscured by a dark streak. Standard Reconstruction Images of the Control Group b p c p d Attenuation (HU) Mean Ipsilateral pelvic sidewall ± ± ± 9.3 < Bladder 16.0 ± ± ± 6.8 < Rectal shelf 47.0 ± ± ± SD Ipsilateral pelvic sidewall 30.9 ± ± ± < Bladder 34.1 ± ± ± Rectal shelf 23.6 ± ± ± < a GE Healthcare. Mean ± SD. c From paired Student t test between standard reconstruction and MAR reconstruction. d From independent Student t test between MAR reconstruction images from patient group and standard reconstruction images from control group. Fig. 1 CT image of phantom without hip prosthesis. Note small balloon containing diluted iodine contrast material (arrow) mimicking urothelial carcinoma or metastasis of bladder. Statistical Analysis The acquired data were evaluated using statistics software (SPSS, version 18.0, IBM Software). The Wilcoxon signed rank test was used to assess the differences in the image quality graded by two radiologists before and after application of the MAR algorithm. The paired Student t test and independent Student t test were used to compare mean attenuation values and SD attenuation values between the original images and the MAR-applied images in the same patient and between the patient group and control group, respectively. The interobserver variability was evaluated by linear-weighted kappa statistics. A kappa value of indicated slight agreement; , fair agreement; , moderate agreement; , substantial agreement; and , almost perfect agreement [12]. A p value of < 0.05 was considered statistically significant. 790 AJR:203, October 2014

4 CT of Patients With Metal Hip Prostheses Results Patients A total of 33 patients were included in this study. The patient population included 19 men and 14 women with a mean age of 63.7 years (range, years). Six patients had bilateral metal hip prostheses, and 27 patients had unilateral prostheses. All the patients were referred for pre- or postoperative CT evaluation of stomach cancer (n = 5); colorectal cancer (n = 10); breast cancer (n = 3); anal cancer (n = 1); lung cancer (n = 8); pseudomyxoma peritonei (n = 1); hepatocellular carcinoma (n = 1); or other benign causes such as diarrhea (n = 1), postoperative intestinal obstruction (n = 1), acute pyelonephritis (n = 1), or benign stricture of the gastrointestinal tract (n = 1). As we mentioned, an additional 20 patients without a metal prosthesis in the abdominal cavity were included as the control group for comparison. Objective Image Analysis The mean attenuation was significantly increased after MAR reconstruction from to 74.6 HU in the ipsilateral pelvic sidewall (p < 0.001) and from 16.0 to 13.9 HU in the bladder (p < 0.001). The mean attenuation of the rectal shelf was decreased after MAR reconstruction, but this difference did not show statistical significance (p = 0.469). The SD attenuation value was decreased in all sites after MAR reconstruction, but there was a statistically significant difference only in the ipsilateral pelvic sidewall (p = 0.019). Although the mean and SD attenuation values of the patient group were close to those of the control group after MAR reconstruction, there was still a significant difference in both the mean and SD attenuation values compared with those of the control group (p < 0.05, Table 1). Subjective Image Analysis Both radiologists judged the quality of the images after MAR reconstruction to be significantly better than the quality of the images with standard reconstruction at the bone-prosthesis interface, ipsilateral pelvic sidewall, and ipsilateral bladder wall (p < 0.05) (Table 2 and Fig. 2). However, the image quality at the rectal shelf after MAR reconstruction was not significantly different (for radiologist 1, p = 0.311) or was worse (for radiologist 2, p = 0.001) than after standard reconstruction (Fig. 3). The image quality of the contralateral bladder wall and the contralateral pelvic sidewall was improved, but only radiologist 1 judged there to be a statistically significant difference between the standard reconstruction images and the MAR reconstruction images. The overall diagnostic confidence of both radiologists was increased with MAR reconstruction, although statistical significance was noted by only one radiologist (radiologist 1, p < 0.001) (Table 2). Interobserver variability was decreased after application of the MAR reconstruction except at the bone-prosthesis interface (weighted κ, ). The linear-weighted kappa values for the evaluation of interobserver variability are summarized in Table 2. Phantom Study In the phantom study, the obscured lesions were successfully visualized after application of the MAR algorithm in all situations. However, new metal artifacts also developed after MAR application, obscuring areas that could be visualized before MAR reconstruction (Fig. 4). Discussion Our results showed that the MAR reconstruction effectively reduced metal artifacts due to hip prostheses, especially at the boneprosthesis interface, ipsilateral pelvic sidewall, and ipsilateral bladder wall. Compared with the standard reconstruction images, the overall image quality was also improved after MAR application. However, metal artifacts in the rectal shelf were not significantly reduced or were even worse on MAR reconstruction images compared with standard reconstruction images. In the phantom study, lesions masked by metal artifacts were successfully visualized in their original shapes after MAR application. However, new dark streak artifacts developed in areas that were previously visualized in the standard reconstruction images. Thus, both the images with and those without MAR reconstruction should be reviewed to avoid missing focal lesions in the pelvic cavity. Because loosening of the prosthesis, bone resorption, dislocation, or metallic or periprosthetic bone fractures are common problems in patients with metallic prostheses, most studies relating to metal artifact reduc- TABLE 2: Subjective Analysis of Quality of Images Reconstructed With Standard Reconstruction and With Metal Artifact Reduction (MAR a ) Software Criteria for Subjective Analysis Visibility of Score c (Mean ± SD) Radiologist 1 Radiologist 2 Score c (Mean ± SD) Weighted κ b Standard MAR p Standard MAR p Standard MAR Contralateral bladder wall 2.6 ± ± ± ± Ipsilateral bladder wall 3.1 ± ± 0.7 < ± ± Contralateral pelvic sidewall 2.1 ± ± ± ± Ipsilateral pelvic sidewall 3.1 ± ± 0.7 < ± ± Rectal shelf 2.3 ± ± ± ± Bone-prosthesis interface 3.7 ± ± 0.6 < ± ± 0.5 < Overall diagnostic confidence 3.1 ± ± 0.6 < ± ± a GE Healthcare. b A kappa value of indicates slight agreement; , fair agreement; , moderate agreement; , substantial agreement; and , almost perfect agreement. c A score of 1 indicates no artifacts affecting image interpretation; 2, minor artifacts not interfering with diagnostic decision making; 3, major artifacts affecting the visualization of major structures, but diagnosis is possible; and 4, severe artifacts that deter accurate diagnosis. AJR:203, October

5 Han et al. tion have focused on the improvement of image quality at the prosthesis itself, the boneprosthesis interface, or the periprosthetic areas [4 6, 13, 14]. However, the evaluation of organs or structures around the metallic prostheses is also important and is being performed more often because of the increasing number of older patients with metallic prostheses. In abdominopelvic CT studies, the pelvic cavity is the area that is most vulnerable to metallic artifacts due to hip prostheses, followed by the paraspinal areas due to metallic fixation of the vertebral bodies. To date, there have been only two studies on the application of a metal artifact reducing reconstruction for abdominal imaging. One study claimed that spectral imaging with a metal artifact reduction algorithm was helpful for reducing near-field artifacts and the blooming effect seen with gold fiducial markers, resulting in improved tumor visibility in the vicinity of the gold markers [10]. Another study found that overall image quality and the visual conspicuity of the bladder A base and rectum were improved by a metal artifact reduction technique using reformatted projections; however, this technique required approximately minutes and was thus not clinically feasible. The recently developed MAR algorithm with GSI dual-energy CT requires only about 5 minutes for reconstruction, which is clinically feasible. Furthermore, because this MAR reconstruction does not increase the required radiation exposure, it may be the only reconstruction algorithm that can be used in daily clinical practice [5, 15]. The attenuation values in Hounsfield units at the ipsilateral pelvic sidewall and bladder of the patient group approached those of the control group, which were assumed to be the expected normal value. In the qualitative analysis, the visual conspicuity of the ipsilateral bladder wall and of the ipsilateral pelvic sidewall was significantly increased with MAR reconstruction for both radiologists, which is comparable with the results previously described by Yu et al. [4]. This finding might be because artifacts manifesting as dark and bright streaks and bands, which have unusually high or low attenuation values, are decreased after MAR reconstruction [15]. The metal artifacts were most markedly decreased at the bone-prosthesis interface, which is in concordance with the results of previous studies [4 6, 13, 14]. At the rectal shelf, the mean and SD attenuation values were close to those of the control group after MAR reconstruction; however, there was no significant difference between the standard reconstruction images and the MAR reconstruction images. In the qualitative analysis, image quality either did not significantly change or was worse. One possible explanation for this phenomenon is that the rectal shelf is located between both ischial tuberosities or between the metal prosthesis and the contralateral ischial tuberosity, resulting in aggravation of beamhardening artifacts between bony structures or between bone and a metal prosthesis [13]. Because the MAR reconstruction algorithm B C D Fig. 2 Axial contrast-enhanced CT scans of 53-year-old man with right metal hip prosthesis undergoing CT for follow-up of anal cancer (A and B) and of 65-year-old man with right metal hip prosthesis undergoing CT for assessment of benign stricture of sigmoid colon (C and D). A D, Visibility of ipsilateral pelvic sidewall, bladder wall, rectal shelf, and bone-metal interface is markedly improved on images reconstructed with Metal Artifact Reduction software (GE Healthcare) (B and D) compared with standard reconstruction images (A and C) at same level. 792 AJR:203, October 2014

6 CT of Patients With Metal Hip Prostheses A C E Fig. 3 Axial contrast-enhanced CT scans of 39-year-old man with left metal hip prosthesis undergoing evaluation for pseudomyxoma peritonei (A and B), 75-year-old woman with left metal hip prosthesis and metal screws undergoing postoperative follow-up of colon cancer (C and D), and 78-year-old man with bilateral hip metal prostheses undergoing preoperative evaluation for rectal cancer (E and F). A and B, Standard reconstruction image (A) and image reconstructed with Metal Artifact Reduction (MAR) software (GE Healthcare) (B). Visibility of bladder wall and overall image quality decreased after MAR reconstruction. Note newly developed bright and dark streaks passing through bladder wall (arrows, B). C and D, Standard reconstruction image (C) and MAR reconstruction image (D). Note newly developed bright and dark streaks passing through rectal shelf (arrows, D). E and F, Standard reconstruction image (E) and MAR reconstruction image (F). Severe metal artifacts obscuring rectal shelf are noted, and overall image quality is not significantly changed after MAR reconstruction. Note newly developed bright and dark streaks passing through bladder wall (arrowheads, F). B D F AJR:203, October

7 Han et al. A C Fig. 4 CT images of phantom with lesions. A D, Images reconstructed with standard reconstruction (A and C) and with Metal Artifact Reduction (MAR) software (GE Healthcare) (B and D) are shown. In situation 1, 1-cm lesion is totally obscured by dark streak (arrow, A); in situation 2, peripheral portion of 2-cm lesion is partly obscured by dark streak (arrow, C). In both cases, lesions (arrowheads, B and D) masked by metal artifacts are successfully shown after MAR application. However, note newly developed bright and dark streaks (arrows, B and D) after MAR application. These streaks obscure portions of artificial bladder that are visible in standard reconstruction images. was primarily developed to compensate for photon starvation artifacts due to metal, other artifacts due to beam hardening, scattering, or partial volumes could not be effectively decreased. In addition, in patients with bilateral hip replacements, the angular range of the corrupted samples due to the metal implants is likely to be larger than in patients with unilateral hip replacements. In contrast to our results, Yu et al. [4] reported that the visual conspicuity of the rectum was increased after MAR reconstruction. This finding might be because the rectum is located above and below the pelvic cavity, and metal artifacts thus could be reduced across most of the rectum except at the level of the rectal shelf, which is located between the ischial tuberosity and the metal prostheses. According to previous studies, residual or new artifacts could present after MAR reconstruction [4, 13, 16]. There are several possible reasons for this phenomenon. First, osseous structures could be interpreted as metal because of their high attenuation value during the metal segmentation and interpolation processes [16]. Second, metal-induced scattering, partial volume, signal underrange, and beam-hardening artifacts are likely to be present in addition to photon starvation. These effects are not fully characterized or corrected by the MAR algorithm. Third, possible errors or inconsistencies in projection views introduced during the estimation of the corrupted samples might induce artifacts after MAR reconstruction [4]. The exact mechanism of appearance of residual or new artifacts after MAR reconstruction should be evaluated in future studies. The overall image quality was improved after MAR reconstruction, but both radiologists judged that there were still residual minor artifacts in the CT images (rated 2.4 and 2.9, respectively, on a 4-point scale). Furthermore, there was a significance difference in the mean and SD attenuation values even after MAR reconstruction, compared with those of the control group. These results suggest that MAR with GSI dual-energy CT could not completely eliminate metal artifacts. The appearance of metal artifacts in CT is dependent on the composition and geometry of the metal as well as the CT parameters. According to previous reports, titanium and aluminum cause fewer metal artifacts than stainless steel or gold [5, 6, 17], and thicker metal results in more artifacts [18]. Although the CT parameters were optimized to minimize metal artifacts, the composition and geometry of the metals used cannot typically be controlled. Different pat- B D 794 AJR:203, October 2014

8 terns and extents of metal artifacts might nostic accuracy of the MAR reconstruction artefact reduction in gemstone spectral imaging dualenergy CT with and without metal artefact reduction CT of Patients With Metal Hip Prostheses not be corrected by one algorithm, but techniques for the detection of pelvic lesions is needed tailored to specific metal implants could more completely remove metal artifacts [14]. In the qualitative analysis, interobserver variability worsened after MAR reconstruction except at the bone-prosthesis interface. This finding might be because of the subjective nature of the interpretation of residual or newly developed artifacts after MAR reconstruction. In terms of the bone-prosthesis interface, metal artifact reduction was evident to both radiologists (score for standard reconstruction vs score for MAR reconstruction: radiologist 1, 3.7 vs 2.5; radiologist 2, 3.8 vs 2.3), resulting in an increase in interobserver agreement after MAR reconstruction. In the phantom study, the MAR reconstruction could depict lesions that were both partially masked and fully masked by metal artifacts. Based on this result, the MAR algorithm with GSI dual-energy CT can be confirmed as having successfully interpolated the missing data. However, as in previous clinical studies, new bright and dark streaks developed after MAR application and portions of the artificial bladder that were previously visualized in the standard reconstruction image were obscured by the new artifacts. Hence, radiologists should review images with both standard and MAR reconstructions to avoid missing focal lesions masked by metal artifacts. There were several limitations to our study. First, a relatively small number of patients were included. Second, although metal composition, geometry, tube voltage, and tube current can affect metal artifacts, we did not evaluate how these characteristics and parameters affect the efficiency of the MAR reconstruction and image quality [5, 18]. However, because commonly used prosthetic hip joints and plates were included in our study, our results are likely representative of what would be seen in real clinical settings. Third, the degree and extent of metal artifacts were evaluated only in a 77-keV monochromatic image. A higher-kev image might better reduce beam-hardening artifacts, although this potential improvement may be at the cost of decreased low-contrast resolution [6]. We focused on the evaluation of nonbony lesions; hence, the loss of low-contrast resolution might interfere with the detection of pelvic lesions. Further evaluation of the relationship between the kev energy level and the diag- in future studies. Last, we could not validate the interpolation algorithm of the MAR software in patients because none of the patients had focal pelvic lesions. Although our phantom study suggested that the MAR algorithm can successfully interpolate and depict lesions masked by metallic artifacts, this finding should be validated in patient data. In conclusion, the use of MAR software with GSI dual-energy CT improved overall image quality and increased diagnostic confidence in the assessment of the pelvic cavity in patients with metal hip prostheses by decreasing metallic artifacts appearing as dark or bright streaks, although statistical significance was noted by only one radiologist in our study. However, metal artifacts in the rectal shelf were not significantly reduced or were worse compared with standard reconstruction images. Furthermore, new dark streak artifacts developed in areas that were previously visualized in the standard reconstruction images after MAR application in the clinical and phantom study. Acknowledgments Min Jeong Yun, who is an employee of GE Healthcare Korea, was involved in optimizing the CT sequences and provided technical support. We also recognize the contributions of Jiang Hsieh, CT Chief Scientist of GE Healthcare Systems, for technical advice and meaningful suggestions. References 1. Buie VC, National Center for Health Statistics. National hospital discharge survey: 2006 annual summary data from the national health care surveys. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics, 2010:v, Korean National Health Insurance Service website. Top 20 highly frequent diseases submitted to surgery. (in Korean) Published November 5, Accessed February Barrett JF, Keat N. Artifacts in CT: recognition and avoidance. RadioGraphics 2004; 24: Yu L, Li H, Mueller J, et al. Metal artifact reduction from reformatted projections for hip prostheses in multislice helical computed tomography: techniques and initial clinical results. Invest Radiol 2009; 44: Lee YH, Park KK, Song HT, Kim S, Suh JS. Metal software. Eur Radiol 2012; 22: Meinel FG, Bischoff B, Zhang Q, Bamberg F, Reiser MF, Johnson TR. Metal artifact reduction by dualenergy computed tomography using energetic extrapolation: a systematically optimized protocol. Invest Radiol 2012; 47: Liu PT, Pavlicek WP, Peter MB, Spangehl MJ, Roberts CC, Paden RG. Metal artifact reduction image reconstruction algorithm for CT of implanted metal orthopedic devices: a work in progress. Skeletal Radiol 2009; 38: Prell D, Kyriakou Y, Kachelrie M, Kalender WA. Reducing metal artifacts in computed tomography caused by hip endoprostheses using a physics-based approach. Invest Radiol 2010; 45: Pal D, Sen Sarma K, Hsieh J. Metal artifact correction algorithm for CT. In: Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC). Washington, DC: IEEE, 2013: Chang YB, Xu D, Zamyatin AA. Metal artifact reduction algorithm for single energy and dual energy CT scans. In: Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC). Washington, DC: IEEE, 2012: Pessis E, Campagna R, Sverzut JM, et al. Virtual monochromatic spectral imaging with fast kilovoltage switching: reduction of metal artifacts at CT. RadioGraphics 2013; 33: Kundel HL, Polansky M. Measurement of observer agreement. Radiology 2003; 228: Watzke O, Kalender WA. A pragmatic approach to metal artifact reduction in CT: merging of metal artifact reduced images. Eur Radiol 2004; 14: Verburg JM, Seco J. CT metal artifact reduction method correcting for beam hardening and missing projections. Phys Med Biol 2012; 57: Bamberg F, Dierks A, Nikolaou K, Reiser MF, Becker CR, Johnson TR. Metal artifact reduction by dual energy computed tomography using monoenergetic extrapolation. Eur Radiol 2011; 21: Mahnken AH, Raupach R, Wildberger JE, et al. A new algorithm for metal artifact reduction in computed tomography: in vitro and in vivo evaluation after total hip replacement. Invest Radiol 2003; 38: Joemai RM, de Bruin PW, Veldkamp WJ, Geleijns J. Metal artifact reduction for CT: development, implementation, and clinical comparison of a generic and a scanner-specific technique. Med Phys 2012; 39: White LM, Buckwalter KA. Technical considerations: CT and MR imaging in the postoperative orthopedic patient. Semin Musculoskelet Radiol 2002; 6:5 17 FOR YOUR INFORMATION This article is available for CME and Self-Assessment (SA-CME) credit that satisfies Part II requirements for maintenance of certification (MOC). To access the examination for this article, follow the prompts associated with the online version of the article. AJR:203, October