- Email: [email protected]
Reduction of dental metallic artefacts in CT: Value of a newly developed algorithm for metal artefact reduction (O-MAR)
Clinical Radiology 69 (2014) e11ee16
Contents lists available at SciVerse ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Reduction of dental metallic artefacts in CT: Value of a newly developed algorithm for metal artefact reduction (O-MAR) M. Kidoh a, b, *, T. Nakaura a, b, S. Nakamura a, b, S. Tokuyasu c, H. Osakabe c, K. Harada d, Y. Yamashita b a
Diagnostic Radiology, Amakusa Medical Center, Amakusa, Kumamoto, Japan Department of Diagnostic Radiology, Graduate School of Medical Sciences, Kumamoto University, Honjo, Kumamoto, Japan c Philips Electronics, Tokyo, Japan d Department of Surgery, Amakusa Medical Center, Amakusa, Kumamoto, Japan b
art icl e i nformat ion Article history: Received 17 January 2013 Received in revised form 23 July 2013 Accepted 7 August 2013
AIM: To evaluate the image quality of O-MAR (Metal Artifact Reduction for Orthopedic Implants) for dental metal artefact reduction. MATERIALS AND METHODS: This prospective study received institutional review board approval and written informed consent was obtained. Thirty patients who had dental implants or dental fillings were included in this study. Computed tomography (CT) images were obtained through the oral cavity and neck during the portal venous phase. The system reconstructed the O-MAR-processed images in addition to the uncorrected images. CT attenuation and image noise of the soft tissue of the oral cavity were compared between the O-MAR and the uncorrected images. Qualitative analysis was undertaken between the two image groups. RESULTS: The image noise of the O-MAR images was significantly lower than that of the uncorrected images (p < 0.01). O-MAR offered plausible attenuations of soft tissue compared with non-O-MAR. Better qualitative scores were obtained in the streaking artefacts and the degree of depiction of the oral cavity with O-MAR compared with non-O-MAR. CONCLUSION: O-MAR enables the depiction of structures in areas in which this was not previously possible due to dental metallic artefacts in qualitative image analysis. O-MAR images may have a supplementary role in addition to uncorrected images in oral diagnosis. Ó 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Streak artefacts from high-attenuation objects are a common problem in computed tomography (CT). This type of artefact typically occurs due to metallic implants such as joint replacement, osteosynthesis, dental implants, or * Guarantor and correspondent: M. Kidoh, Diagnostic Radiology, Amakusa Medical Center, Kameba 854-1, Amakusa, Kumamoto 863-0046, Japan. Tel.: þ81 969244111. E-mail address: [email protected] (M. Kidoh).
dental fillings. When CT examinations are carried out in dental and maxillofacial regions in the presence of metallic prosthetic appliances in the oral cavity, the appearance of metal-induced streak artefacts is unavoidable.1e4 The metal artefacts actually comprise two main different components. One is photon starvation due to full absorption of x-ray quanta, causing zero-transmission projections. The other is beam hardening due to absorption of low-energy quanta.5 Although there is no remedy for these artefacts, in clinical practice, higher-energy quanta reduce the extent of artefacts to some degree and could constitute a potential
0009-9260/$ e see front matter Ó 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.crad.2013.08.008
e12
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
approach to improve diagnostic CT image quality in metallic implants. Different approaches towards metal artefact reduction (MAR) have been described, including physical prefiltering, water correction, and dual energy scanning. Many different post-processing algorithms for MAR have also been reported.6e9 O-MAR (Metal Artifact Reduction for Orthopedic Implants, Philips Healthcare, Cleveland, OH, USA) is a recently released commercial product that implements a robust and efficient algorithm to mitigate artefacts caused by metal objects in CT images; that is, O-MAR is the first iterative reconstruction method applied clinically. To the authors’ knowledge, there have been no published reports regarding the effectiveness of O-MAR. In addition, the effectiveness of O-MAR for non-orthopaedic metal, for example, dental fillings, has not been sufficiently analysed because the main purpose of O-MAR is to address artefacts arising from orthopaedic metal. When multiple metallic objects, such as dental fillings, are present in dental regions, the clinical application of previous MAR methods is limited.10,11 The purpose of the present study was to evaluate the image quality of O-MAR for dental metal artefact reduction in CT in oral diagnosis.
Materials and methods This prospective study received institutional review board approval; prior informed consent to participate was obtained from all patients.
Patients Between August and October 2012, 73 patients were consecutively enrolled. Serum creatinine levels were determined in all patients prior to contrast materialenhanced examinations within 3 months of the imaging examination, and the estimated glomerular filtration rate (eGFR) was calculated using the modification of diet in renal disease (MDRD) formula.12,13 Body weight (BW) was also obtained in all patients prior to the CT examination to tailor the amount of contrast medium. The inclusion criteria were a past history or suspicion of metastatic head and neck tumour or lymph node metastasis. The exclusion criteria were as follows: (1) renal failure (eGFR <60 ml/min/1.73 m2); (2) a history of a reaction to iodinated contrast media; and (3) proven or suspected pregnancy. Twenty-one patients with an eGFR <57 ml/ min/1.73 m2 were excluded and scanned with a low contrast medium dose CT protocol. Consequently, 52 patients were enrolled in the study. There were 16 men and 36 women ranging in age from 26e88 years (mean 66.2 years); their BW ranged from 33e86 kg (mean 54.4 kg). Thirty patients had dental implants or dental fillings (metal group). The other 22 patients did not have dental metal. To define “typical attenuations” of the soft tissue in the oral cavity, they were assigned to a control group (non-metal group).
CT scanning and contrast medium infusion protocols All patients were imaged using a 256-row MDCT system (Brilliance iCT, Philips Healthcare). All patients were scanned after contrast material (600 mg iodine/kg BW) had been delivered over 30 s.14 The acquisition was performed with Dose Right (Philips Healthcare) automatic current selection [reference volume CT dose index (CTDIvol), 18 mGy for 120 kVp]. The detailed imaging parameters of each protocol are shown in Table 1. Iohexol (300 mg iodine/ml; Omnipaque-300, DaiichiSankyo, Tokyo, Japan), iomeprol (350 mg iodine/ml; Iomeron-350, Eisai, Tokyo, Japan), or iopamidol (370 mg iodine/ml; Iopamiron-370, Nihon Schering, Osaka, Japan) was delivered via a 20 G catheter inserted into an antecubital vein using a power injector (DUAL SHOT GX; Nemoto-Kyorindo, Tokyo, Japan). CT images were obtained through the oral cavity and neck during the portal venous phase. An automatic bolustracking program (Bolus Pro Ultra, Philips Medical Systems) was used to time the start of the scan after contrast medium injection. Monitoring was performed at the level of the L1 vertebral body. The region of interest (ROI) cursor was placed in the abdominal aorta. Real-time (120 kVp, 15 mAs) serial monitoring studies began 10 s after the start of contrast medium injection. The scan was started 55 s after the trigger (150 HU). The patients were instructed to hold their breath with tidal inspiration during scanning. Image reconstruction was performed with 5 mm thickness in a 21 cm display field of view (FOV). The images were reconstructed using a standard filtered back projection (FBP) algorithm with a standard soft-tissue kernel (kernel B; uncorrected image). In the metal group, the system reconstructed the O-MAR-processed images in addition to the uncorrected images. In this study, the images were not reconstructed using the O-MAR algorithm in the non-metal group because O-MAR has no impact on non-metal images theoretically. The CTDIvol provided by the CT machine was recorded for each patient.
Quantitative image analysis A radiologist performed quantitative image analysis on reconstructed 5 mm thick transverse images. To evaluate Table 1 Imaging parameters. Beam collimation (mm) Slice thickness and intervals (mm) Helical pitch Table movement (mm/s) Reference CTDIvol (mGy) for ACS Rotation time (s) Effective tube current (mAs/section) Tube voltage (kVp) Total amount of contrast medium (mg iodine/kg) Injection duration (s) Bolus tracking trigger (HU) Scan delay (s)
128 0.625 5 0.8 128 18 0.5 155e490 120 600 30 150 55
CTDIvol, volume CT dose index; ACS, automatic current selection.
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
e13
the image contrast, the radiodensity values of the oral cavity were measured in a circular ROI. In addition, image noise was defined as the standard deviation (SD) in Hounsfield units (HU) of the oral cavity, which was bounded by hard palate and floor of mouth. Attempts were made to select the ROI in the soft tissue of the centre of the oral cavity of approximately 400 mm2, as this size was not so small as to be affected by pixel variability and not so large as to include tooth or air. The ROI on the section including dental artefact was measured in the metal group. To minimize bias from single measurements, the ROI was measured twice on the same image; the measurements were averaged.
Table 2 Quantitative image analysis
Qualitative image analysis
were considered statistically significant. The Kappa coefficients were calculated based on both the uncorrected and O-MAR results. The scale for the Kappa coefficients for interobserver agreement was as follows: less than 0.20 ¼ poor, 0.21e0.40 ¼ fair, 0.41e0.60 ¼ moderate, 0.61e0.80 ¼ substantial, and 0.81e1.00 ¼ near-perfect. Statistical analyses were performed with the free statistical software “R” (R, version 2.6.1; The R Project for Statistical Computing; http://www.r-project.org/).
To evaluate the image quality and the effect of O-MAR, qualitative image analysis of axial images on a picture archiving and communication system (PACS) viewer (Synapse, Fuji Film Medicals, Tokyo, Japan) was performed. Two radiologists with 9 and 5 years of experience in head and neck radiology independently graded the following four parameters: streaking artefacts, image sharpness, texture naturalness, and degree of depiction of oral cavity. The CT datasets were randomized, and the radiologists were blinded to the acquisition parameters. The presence of streaking artefacts at mediastinal window settings (window level and width of 40 and 280 HU) was examined qualitatively on a four-point subjective scale (1 ¼ extensive streak artefacts; 2 ¼ moderate streak artefacts present and interfering with the depiction of adjacent structures; 3 ¼ minimal streak artefacts present, but not interfering with the depiction of adjacent structures; 4 ¼ no streak artefacts). Image sharpness was graded with regard to the edge sharpness of the muscle tissue (1 ¼ marked blurring without definable margins; 2 ¼ blurring, but with definable margins; 3 ¼ minimal blurring; 4 ¼ sharp definition). Texture naturalness was graded (1 ¼ unnatural image texture present and unacceptable; 2 ¼ unnatural image texture present and interfering with the depiction of adjacent structures; 3 ¼ unnatural image texture present without interfering with the depiction of adjacent structures; 4 ¼ natural image texture). The degree of depiction of the oral cavity was graded (1 ¼ no depiction; 2 ¼ faint depiction; 3 ¼ good depiction; 4 ¼ excellent depiction). In cases of interobserver disagreement, final decisions were reached by consensus.
Statistical analysis All numeric values are reported as the mean SD. Wilcoxon signed rank test was used to compare the radiodensity, image noise, and qualitative data, because the KolomogoroveSmirnov test revealed a distribution of data (radiodensity in uncorrected images, and radiodensity and image noise in O-MAR images) that was different from normality (p < 0.05 for all). Levene’s test was used to compare the variation (standard deviation) of the attenuation values. McNemar’s test was used to compare the success rate of typical attenuations. Differences of p < 0.05
Uncorrected images
O-MAR images
p-Value p-Value for enhancement variability
Radiodensity e7.6 114.7a 80.2 26.7a <0.01 (HU) Image noise 222.8 222.1a 33.0 16.9a <0.01 Success rate 23.3% (7/30) 80.0% (24/30) <0.01 of “typical attenuations”b a b
<0.01
d
Data are mean standard deviation. “Typical attenuations”, 36e105 HU.
Results Patient characteristics and radiation dose There were no significant differences between the metal and non-metal groups with respect to age, male : female ratio, BW, and CTDIvol (65.7 13.2 versus 67.0 14.9; 8:22 versus 8:14; 53.7 10.2 versus 55.4 12.7; 12.6 4 versus 13.6 4.2, respectively; p > 0.05 for all).
Quantitative image analysis Table 2 and Fig 1 show the results of quantitative image analysis. In the non-metal group (22 patients who did not have dental metal), attenuations of the soft tissue in the oral cavity ranged from 36e105 HU (mean 67.7 HU). These measured data were not affected by metal objects. Therefore, 36e105 HU was defined as “typical attenuations” in oral cavity. In the metal group (30 patients who had dental implants or dental fillings), the number and rate of patients in which "typical attenuation" were achieved were compared between the uncorrected images and the O-MAR images (the rate of “typical attenuation” ). patients ¼ Numberofpatients who achieved 30 As shown in Table 2, in the metal group, the success rate of “typical attenuations” of the O-MAR images was significantly higher than that of the uncorrected images (80% versus 23.3%, p < 0.01); that is, the O-MAR attenuation was more plausible than the attenuation seen on the uncorrected images. There were significant differences in the radiodensity between the uncorrected images and the O-MAR images (7.6 114.7 versus 80.2 26.7 HU, p < 0.01). The variation (standard deviation) of the radiodensity of the O-MAR images was significantly lower than that of the uncorrected images (26.7 versus 114.7, p < 0.01). The image noise of the
e14
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16 Table 3 Qualitative image analysis Uncorrected images Streaking artifacts Image sharpness Texture naturalness Degree of depiction
1.5 3.8 3.9 1.8
0.7 0.4 0.3 0.6
O-MAR images 3.5 3.4 3.5 3.2
0.5 0.5 0.5 0.6
p Value
Kappa
<0.01 <0.01 <0.01 <0.01
0.64 0.65 0.71 0.73
Note - Data are shown as the mean standard deviation.
Discussion
corrupted data in sinogram space are replaced via interpolation.17,18 Despite their popularity, however, the reliability of many pure interpolation-based approaches decreases considerably when considering large and/or multiple metal objects.18 Metal artefact reduction methods, such as the metal deletion technique (MDT) or the selective algebraic reconstruction technique (SART), which adopt an iterative reconstruction technique, have been developed.2 These methods might effectively reduce the appearance of metal artefacts in reconstructed images compared with previous projection completion methods; however, they have not been released as commercial products because they are computationally expensive for clinical CT machines and, thus, impractical for clinical use. Iterative reconstruction techniques produce better image quality than previous methods and might have the potential to generate artefactfree CT images.16 Boas et al. evaluated an iterative technique (MDT) for reducing metal artefacts in CT. They concluded that MDT effectively yielded reduced metal streak artefacts and had the potential to improve diagnostic accuracy.2 However, despite their excellent results, commercialization has been difficult. O-MAR is a commercial product that implements an iterative reconstruction algorithm to mitigate artefacts caused by metal objects in CT images; recent rapid growth in computer power has enabled a shortening of the calculation time; as a result, O-MAR, for which processing times are clinically acceptable, is the first to be released ahead of other iterative reconstruction methods. Iterative
O-MAR images may have a supplementary role in addition to uncorrected images in oral diagnosis. Although OMAR did not totally eliminate metal artefacts, it did enable the depiction of structures in areas in which this was not previously possible due to dental metallic artefacts in qualitative image analysis. There was minimal unnatural texture or blurring induced by O-MAR in the image; however, O-MAR offered plausible attenuations of soft tissue in quantitative image analysis and better qualitative score in streaking artefacts and degree of depiction of the oral cavity compared with non-O-MAR. O-MAR may provide additional diagnostic information to clinicians upon crossreferencing the uncorrected images with the O-MAR dataset in CT in dental regions. Several corrective methods have been previously studied to reduce streak artefacts caused by high-attenuation material15,16; the vast majority of existing MAR techniques fall into the projection completion category, whereby missing/
Figure 2 There were significant differences in streaking artefacts, image sharpness, texture naturalness and degree of depiction of the oral cavity between the uncorrected images and the O-MAR images (1.5 0.7 versus 3.5 0.5; 3.8 0.4 versus 3.4 0.5; 3.9 0.3 versus 3.5 0.5; 1.8 0.6 versus 3.2 0.6, respectively; p < 0.01).
Figure 1 The image noise of O-MAR images (33 16.9) was significantly lower than that of uncorrected images (222.8 222.1; p < 0.01). There were significant differences in the radiodensity between the uncorrected images and the O-MAR images (e7.6 114.7 versus 80.2 26.7 HU, p < 0.01).
O-MAR images (33 16.9) was significantly lower than that of the uncorrected images (222.8 222.1; p < 0.01).
Qualitative image analysis Table 3 and Fig 2 show the results of the qualitative image analysis. There were significant differences in streaking artefacts, image sharpness, texture naturalness, and degree of depiction of the oral cavity between the uncorrected images and the O-MAR images (1.5 0.7 versus 3.5 0.5; 3.8 0.4 versus 3.4 0.5; 3.9 0.3 versus 3.5 0.5; 1.8 0.6 versus 3.2 0.6, respectively; p < 0.01). There were substantial interobserver agreements with respect to streaking artefacts, image sharpness, texture naturalness, and degree of depiction of the oral cavity (Kappa ¼ 0.64, 0.65, 0.71 and 0.73, respectively). Representative cases are shown in Figs 3 and 4.
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
e15
Figure 3 (a) Uncorrected image and (b) O-MAR image of a 50-year-old man with suspicion of lymph node metastasis. (b) An example of O-MAR being applied to a patient with a dental filling. On the uncorrected image, the dental region was obscured by the dark shadow caused by the metal. On the O-MAR image, the dental region was visible.
reconstruction algorithms are specifically designed to reconstruct images from incomplete projections.19,20 However, this category of algorithm is much more computationally expensive than projection completion methods, and has only recently become available for clinical use. The rapid progress of computers has resulted in major technological developments in reconstruction techniques; attempts to minimize the execution time of applications were established by the rapid development of central processing unit (CPU) chips. Owing to these developments, recently, OMAR was released for clinical use. In fact, using a standard personal computer, O-MAR image reconstruction is possible within a few minutes in most cases, a time that is acceptable for clinical use.
The most important finding of the present study is that a better qualitative score was noted in terms of the degree of depiction of the oral cavity with O-MAR compared with that in non-O-MAR, i.e., O-MAR made it possible to depict structures in areas where this was not previously possible due to metal artefacts in qualitative image analysis. When multiple metallic objects, such as dental fillings, are present in the imaging field of view, the clinical application of projection completion methods is limited.10,11 However, in the present study, O-MAR removed metal artefacts originating from dental metals effectively. Magnetic resonance imaging (MRI) can quite often uncover lesions obscured by amalgam artefact. However, O-MAR might be able to render severely degraded images useful for oral diagnosis,
Figure 4 (a) Uncorrected image and (b) O-MAR image of a 50-year-old man with suspicion of lymph node metastasis. On the uncorrected image, the dental filling caused severe metal artefacts in the CT images. On the O-MAR image, both the streak and the darkening artefacts were mitigated. Although O-MAR does not totally eliminate metal artefacts, it is capable of reducing their effects on CT images to enhance the diagnostic quality of the images significantly.
e16
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
preoperative assessment of implants and radiotherapy planning in patients who are unsuitable for MRI. In qualitative image analysis, there was a statistically significant increase in blurring and unnatural texture following O-MAR. A limitation of O-MAR might stem from the fact that metal projection data are discarded, resulting in a potential loss of spatial resolution near the metal implant. The cause of the unnatural texture on the O-MAR images could not be determined because detailed information about the algorithms is not available to users. Prevention of unnatural image texture might be the main technical challenge for O-MAR in the future. There are some limitations to the present study. The main limitation was that diagnostic accuracy was not evaluated; the focus was comparing the image quality of each protocol. Second, the possibility that the quantitative parameters did not really measure image quality cannot be rule out. Radiodensity (HU) and image noise (SD of radiodensity of the ROI within the oral cavity) as surrogate markers of image quality, which have been widely used for the assessment of image quality in previous studies (in general, lower image noise means better image quality). In the present study, O-MAR reduced image noise caused by metal objects; the radiodensity of the O-MAR images was more plausible than that seen on the uncorrected images. However, O-MAR images were not compared with a reference standard (e.g., MRI, surgery). Thus, further clinical studies might be required to demonstrate whether O-MAR accurately depicts reality. Other limitations to this pilot study include the small sample size and bias due to images from the same case being presented (once with, once without O-MAR) in qualitative analysis. Also O-MAR was only tested using a single CT system. In conclusion, as O-MAR enables the depiction of structures in areas in which this was not previously possible due to dental metallic artefacts, O-MAR images may have a supplementary role, in addition to uncorrected images, in oral diagnosis.
References 1. Miracle AC, Mukherji SK. Conebeam CT of the head and neck, part 2: clinical applications. AJNR Am J Neuroradiol 2009;30:1285e92.
2. Boas FE, Fleischmann D. Evaluation of two iterative techniques for reducing metal artifacts in computed tomography. Radiology 2011; 259:894e902. 3. Dong J, Hayakawa Y, Kannenberg S, et al. Metal-induced streak artifact reduction using iterative reconstruction algorithms in x-ray computed tomography image of the dentoalveolar region. Oral Surg Oral Med Oral Pathol Oral Radiol 2013;115:e63e73. 4. Abdoli M, Ay MR, Ahmadian A, et al. Reduction of dental filling metallic artifacts in CT-based attenuation correction of PET data using weighted virtual sinograms optimized by a genetic algorithm. Med Phys 2010;37:6166e77. 5. Barrett JF, Keat N. Artifacts in CT: recognition and avoidance. RadioGraphics 2004;24:1679e91. 6. Bamberg F, Dierks A, Nikolaou K, et al. Metal artifact reduction by dual energy computed tomography using monoenergetic extrapolation. Eur Radiol 2011;21:1424e9. 7. 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;38:769e75. 8. Glover GH, Pelc NJ. An algorithm for the reduction of metal clip artifacts in CT reconstructions. Med Phys. 1981;8:799e807. 9. Kalender WA, Hebel R, Ebersberger J. Reduction of CT artifacts caused by metallic implants. Radiology 1987;164:576e7. 10. Tohnak S, Mehnert AJ, Mahoney M, et al. Dental CT metal artefact reduction based on sequential substitution. Dentomaxillofac Radiol 2011;40:184e90. 11. Lemmens C, Faul D, Nuyts J. Suppression of metal artifacts in CT using a reconstruction procedure that combines MAP and projection completion. IEEE Trans Med Imaging 2009;28:250e60. 12. Imai E, Horio M, Nitta K, et al. Modification of the Modification of Diet in Renal Disease (MDRD) Study equation for Japan. Am J Kidney Dis 2007;50:927e37. 13. Matsuo S, Imai E, Horio M, et al. Revised equations for estimated GFR from serum creatinine in Japan. Am J Kidney Dis 2009; 53:982e92. 14. Yamashita Y, Komohara Y, Takahashi M, et al. Abdominal helical CT: evaluation of optimal doses of intravenous contrast materialda prospective randomized study. Radiology 2000;216:718e23. 15. Joemai RM, de Bruin PW, Veldkamp WJ, et al. Metal artifact reduction for CT: development, implementation, and clinical comparison of a generic and a scanner-specific technique. Med Phys. 2012;39:1125e32. 16. Xu C, Verhaegen F, Laurendeau D, et al. An algorithm for efficient metal artifact reductions in permanent seed. Med Phys. 2011;38:47e56. 17. Yu H, Zeng K, Bharkhada DK, et al. A segmentation-based method for metal artifact reduction. Acad Radiol 2007;14:495e504. 18. Meyer E, Raupach R, Lell M, et al. Normalized metal artifact reduction (NMAR) in computed tomography. Med Phys. 2010;37:5482e93. 19. Wang G, Snyder DL, O’Sullivan JA, et al. Iterative deblurring for CT metal artifact reduction. IEEE Trans Med Imaging 1996;15:657e64. 20. Wang G, Vannier MW, Cheng PC. Iterative x-ray cone-beam tomography for metal artifact reduction and local region reconstruction. Microsc Microanal 1999;5:58e65.
Contents lists available at SciVerse ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Reduction of dental metallic artefacts in CT: Value of a newly developed algorithm for metal artefact reduction (O-MAR) M. Kidoh a, b, *, T. Nakaura a, b, S. Nakamura a, b, S. Tokuyasu c, H. Osakabe c, K. Harada d, Y. Yamashita b a
Diagnostic Radiology, Amakusa Medical Center, Amakusa, Kumamoto, Japan Department of Diagnostic Radiology, Graduate School of Medical Sciences, Kumamoto University, Honjo, Kumamoto, Japan c Philips Electronics, Tokyo, Japan d Department of Surgery, Amakusa Medical Center, Amakusa, Kumamoto, Japan b
art icl e i nformat ion Article history: Received 17 January 2013 Received in revised form 23 July 2013 Accepted 7 August 2013
AIM: To evaluate the image quality of O-MAR (Metal Artifact Reduction for Orthopedic Implants) for dental metal artefact reduction. MATERIALS AND METHODS: This prospective study received institutional review board approval and written informed consent was obtained. Thirty patients who had dental implants or dental fillings were included in this study. Computed tomography (CT) images were obtained through the oral cavity and neck during the portal venous phase. The system reconstructed the O-MAR-processed images in addition to the uncorrected images. CT attenuation and image noise of the soft tissue of the oral cavity were compared between the O-MAR and the uncorrected images. Qualitative analysis was undertaken between the two image groups. RESULTS: The image noise of the O-MAR images was significantly lower than that of the uncorrected images (p < 0.01). O-MAR offered plausible attenuations of soft tissue compared with non-O-MAR. Better qualitative scores were obtained in the streaking artefacts and the degree of depiction of the oral cavity with O-MAR compared with non-O-MAR. CONCLUSION: O-MAR enables the depiction of structures in areas in which this was not previously possible due to dental metallic artefacts in qualitative image analysis. O-MAR images may have a supplementary role in addition to uncorrected images in oral diagnosis. Ó 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Streak artefacts from high-attenuation objects are a common problem in computed tomography (CT). This type of artefact typically occurs due to metallic implants such as joint replacement, osteosynthesis, dental implants, or * Guarantor and correspondent: M. Kidoh, Diagnostic Radiology, Amakusa Medical Center, Kameba 854-1, Amakusa, Kumamoto 863-0046, Japan. Tel.: þ81 969244111. E-mail address: [email protected] (M. Kidoh).
dental fillings. When CT examinations are carried out in dental and maxillofacial regions in the presence of metallic prosthetic appliances in the oral cavity, the appearance of metal-induced streak artefacts is unavoidable.1e4 The metal artefacts actually comprise two main different components. One is photon starvation due to full absorption of x-ray quanta, causing zero-transmission projections. The other is beam hardening due to absorption of low-energy quanta.5 Although there is no remedy for these artefacts, in clinical practice, higher-energy quanta reduce the extent of artefacts to some degree and could constitute a potential
0009-9260/$ e see front matter Ó 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.crad.2013.08.008
e12
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
approach to improve diagnostic CT image quality in metallic implants. Different approaches towards metal artefact reduction (MAR) have been described, including physical prefiltering, water correction, and dual energy scanning. Many different post-processing algorithms for MAR have also been reported.6e9 O-MAR (Metal Artifact Reduction for Orthopedic Implants, Philips Healthcare, Cleveland, OH, USA) is a recently released commercial product that implements a robust and efficient algorithm to mitigate artefacts caused by metal objects in CT images; that is, O-MAR is the first iterative reconstruction method applied clinically. To the authors’ knowledge, there have been no published reports regarding the effectiveness of O-MAR. In addition, the effectiveness of O-MAR for non-orthopaedic metal, for example, dental fillings, has not been sufficiently analysed because the main purpose of O-MAR is to address artefacts arising from orthopaedic metal. When multiple metallic objects, such as dental fillings, are present in dental regions, the clinical application of previous MAR methods is limited.10,11 The purpose of the present study was to evaluate the image quality of O-MAR for dental metal artefact reduction in CT in oral diagnosis.
Materials and methods This prospective study received institutional review board approval; prior informed consent to participate was obtained from all patients.
Patients Between August and October 2012, 73 patients were consecutively enrolled. Serum creatinine levels were determined in all patients prior to contrast materialenhanced examinations within 3 months of the imaging examination, and the estimated glomerular filtration rate (eGFR) was calculated using the modification of diet in renal disease (MDRD) formula.12,13 Body weight (BW) was also obtained in all patients prior to the CT examination to tailor the amount of contrast medium. The inclusion criteria were a past history or suspicion of metastatic head and neck tumour or lymph node metastasis. The exclusion criteria were as follows: (1) renal failure (eGFR <60 ml/min/1.73 m2); (2) a history of a reaction to iodinated contrast media; and (3) proven or suspected pregnancy. Twenty-one patients with an eGFR <57 ml/ min/1.73 m2 were excluded and scanned with a low contrast medium dose CT protocol. Consequently, 52 patients were enrolled in the study. There were 16 men and 36 women ranging in age from 26e88 years (mean 66.2 years); their BW ranged from 33e86 kg (mean 54.4 kg). Thirty patients had dental implants or dental fillings (metal group). The other 22 patients did not have dental metal. To define “typical attenuations” of the soft tissue in the oral cavity, they were assigned to a control group (non-metal group).
CT scanning and contrast medium infusion protocols All patients were imaged using a 256-row MDCT system (Brilliance iCT, Philips Healthcare). All patients were scanned after contrast material (600 mg iodine/kg BW) had been delivered over 30 s.14 The acquisition was performed with Dose Right (Philips Healthcare) automatic current selection [reference volume CT dose index (CTDIvol), 18 mGy for 120 kVp]. The detailed imaging parameters of each protocol are shown in Table 1. Iohexol (300 mg iodine/ml; Omnipaque-300, DaiichiSankyo, Tokyo, Japan), iomeprol (350 mg iodine/ml; Iomeron-350, Eisai, Tokyo, Japan), or iopamidol (370 mg iodine/ml; Iopamiron-370, Nihon Schering, Osaka, Japan) was delivered via a 20 G catheter inserted into an antecubital vein using a power injector (DUAL SHOT GX; Nemoto-Kyorindo, Tokyo, Japan). CT images were obtained through the oral cavity and neck during the portal venous phase. An automatic bolustracking program (Bolus Pro Ultra, Philips Medical Systems) was used to time the start of the scan after contrast medium injection. Monitoring was performed at the level of the L1 vertebral body. The region of interest (ROI) cursor was placed in the abdominal aorta. Real-time (120 kVp, 15 mAs) serial monitoring studies began 10 s after the start of contrast medium injection. The scan was started 55 s after the trigger (150 HU). The patients were instructed to hold their breath with tidal inspiration during scanning. Image reconstruction was performed with 5 mm thickness in a 21 cm display field of view (FOV). The images were reconstructed using a standard filtered back projection (FBP) algorithm with a standard soft-tissue kernel (kernel B; uncorrected image). In the metal group, the system reconstructed the O-MAR-processed images in addition to the uncorrected images. In this study, the images were not reconstructed using the O-MAR algorithm in the non-metal group because O-MAR has no impact on non-metal images theoretically. The CTDIvol provided by the CT machine was recorded for each patient.
Quantitative image analysis A radiologist performed quantitative image analysis on reconstructed 5 mm thick transverse images. To evaluate Table 1 Imaging parameters. Beam collimation (mm) Slice thickness and intervals (mm) Helical pitch Table movement (mm/s) Reference CTDIvol (mGy) for ACS Rotation time (s) Effective tube current (mAs/section) Tube voltage (kVp) Total amount of contrast medium (mg iodine/kg) Injection duration (s) Bolus tracking trigger (HU) Scan delay (s)
128 0.625 5 0.8 128 18 0.5 155e490 120 600 30 150 55
CTDIvol, volume CT dose index; ACS, automatic current selection.
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
e13
the image contrast, the radiodensity values of the oral cavity were measured in a circular ROI. In addition, image noise was defined as the standard deviation (SD) in Hounsfield units (HU) of the oral cavity, which was bounded by hard palate and floor of mouth. Attempts were made to select the ROI in the soft tissue of the centre of the oral cavity of approximately 400 mm2, as this size was not so small as to be affected by pixel variability and not so large as to include tooth or air. The ROI on the section including dental artefact was measured in the metal group. To minimize bias from single measurements, the ROI was measured twice on the same image; the measurements were averaged.
Table 2 Quantitative image analysis
Qualitative image analysis
were considered statistically significant. The Kappa coefficients were calculated based on both the uncorrected and O-MAR results. The scale for the Kappa coefficients for interobserver agreement was as follows: less than 0.20 ¼ poor, 0.21e0.40 ¼ fair, 0.41e0.60 ¼ moderate, 0.61e0.80 ¼ substantial, and 0.81e1.00 ¼ near-perfect. Statistical analyses were performed with the free statistical software “R” (R, version 2.6.1; The R Project for Statistical Computing; http://www.r-project.org/).
To evaluate the image quality and the effect of O-MAR, qualitative image analysis of axial images on a picture archiving and communication system (PACS) viewer (Synapse, Fuji Film Medicals, Tokyo, Japan) was performed. Two radiologists with 9 and 5 years of experience in head and neck radiology independently graded the following four parameters: streaking artefacts, image sharpness, texture naturalness, and degree of depiction of oral cavity. The CT datasets were randomized, and the radiologists were blinded to the acquisition parameters. The presence of streaking artefacts at mediastinal window settings (window level and width of 40 and 280 HU) was examined qualitatively on a four-point subjective scale (1 ¼ extensive streak artefacts; 2 ¼ moderate streak artefacts present and interfering with the depiction of adjacent structures; 3 ¼ minimal streak artefacts present, but not interfering with the depiction of adjacent structures; 4 ¼ no streak artefacts). Image sharpness was graded with regard to the edge sharpness of the muscle tissue (1 ¼ marked blurring without definable margins; 2 ¼ blurring, but with definable margins; 3 ¼ minimal blurring; 4 ¼ sharp definition). Texture naturalness was graded (1 ¼ unnatural image texture present and unacceptable; 2 ¼ unnatural image texture present and interfering with the depiction of adjacent structures; 3 ¼ unnatural image texture present without interfering with the depiction of adjacent structures; 4 ¼ natural image texture). The degree of depiction of the oral cavity was graded (1 ¼ no depiction; 2 ¼ faint depiction; 3 ¼ good depiction; 4 ¼ excellent depiction). In cases of interobserver disagreement, final decisions were reached by consensus.
Statistical analysis All numeric values are reported as the mean SD. Wilcoxon signed rank test was used to compare the radiodensity, image noise, and qualitative data, because the KolomogoroveSmirnov test revealed a distribution of data (radiodensity in uncorrected images, and radiodensity and image noise in O-MAR images) that was different from normality (p < 0.05 for all). Levene’s test was used to compare the variation (standard deviation) of the attenuation values. McNemar’s test was used to compare the success rate of typical attenuations. Differences of p < 0.05
Uncorrected images
O-MAR images
p-Value p-Value for enhancement variability
Radiodensity e7.6 114.7a 80.2 26.7a <0.01 (HU) Image noise 222.8 222.1a 33.0 16.9a <0.01 Success rate 23.3% (7/30) 80.0% (24/30) <0.01 of “typical attenuations”b a b
<0.01
d
Data are mean standard deviation. “Typical attenuations”, 36e105 HU.
Results Patient characteristics and radiation dose There were no significant differences between the metal and non-metal groups with respect to age, male : female ratio, BW, and CTDIvol (65.7 13.2 versus 67.0 14.9; 8:22 versus 8:14; 53.7 10.2 versus 55.4 12.7; 12.6 4 versus 13.6 4.2, respectively; p > 0.05 for all).
Quantitative image analysis Table 2 and Fig 1 show the results of quantitative image analysis. In the non-metal group (22 patients who did not have dental metal), attenuations of the soft tissue in the oral cavity ranged from 36e105 HU (mean 67.7 HU). These measured data were not affected by metal objects. Therefore, 36e105 HU was defined as “typical attenuations” in oral cavity. In the metal group (30 patients who had dental implants or dental fillings), the number and rate of patients in which "typical attenuation" were achieved were compared between the uncorrected images and the O-MAR images (the rate of “typical attenuation” ). patients ¼ Numberofpatients who achieved 30 As shown in Table 2, in the metal group, the success rate of “typical attenuations” of the O-MAR images was significantly higher than that of the uncorrected images (80% versus 23.3%, p < 0.01); that is, the O-MAR attenuation was more plausible than the attenuation seen on the uncorrected images. There were significant differences in the radiodensity between the uncorrected images and the O-MAR images (7.6 114.7 versus 80.2 26.7 HU, p < 0.01). The variation (standard deviation) of the radiodensity of the O-MAR images was significantly lower than that of the uncorrected images (26.7 versus 114.7, p < 0.01). The image noise of the
e14
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16 Table 3 Qualitative image analysis Uncorrected images Streaking artifacts Image sharpness Texture naturalness Degree of depiction
1.5 3.8 3.9 1.8
0.7 0.4 0.3 0.6
O-MAR images 3.5 3.4 3.5 3.2
0.5 0.5 0.5 0.6
p Value
Kappa
<0.01 <0.01 <0.01 <0.01
0.64 0.65 0.71 0.73
Note - Data are shown as the mean standard deviation.
Discussion
corrupted data in sinogram space are replaced via interpolation.17,18 Despite their popularity, however, the reliability of many pure interpolation-based approaches decreases considerably when considering large and/or multiple metal objects.18 Metal artefact reduction methods, such as the metal deletion technique (MDT) or the selective algebraic reconstruction technique (SART), which adopt an iterative reconstruction technique, have been developed.2 These methods might effectively reduce the appearance of metal artefacts in reconstructed images compared with previous projection completion methods; however, they have not been released as commercial products because they are computationally expensive for clinical CT machines and, thus, impractical for clinical use. Iterative reconstruction techniques produce better image quality than previous methods and might have the potential to generate artefactfree CT images.16 Boas et al. evaluated an iterative technique (MDT) for reducing metal artefacts in CT. They concluded that MDT effectively yielded reduced metal streak artefacts and had the potential to improve diagnostic accuracy.2 However, despite their excellent results, commercialization has been difficult. O-MAR is a commercial product that implements an iterative reconstruction algorithm to mitigate artefacts caused by metal objects in CT images; recent rapid growth in computer power has enabled a shortening of the calculation time; as a result, O-MAR, for which processing times are clinically acceptable, is the first to be released ahead of other iterative reconstruction methods. Iterative
O-MAR images may have a supplementary role in addition to uncorrected images in oral diagnosis. Although OMAR did not totally eliminate metal artefacts, it did enable the depiction of structures in areas in which this was not previously possible due to dental metallic artefacts in qualitative image analysis. There was minimal unnatural texture or blurring induced by O-MAR in the image; however, O-MAR offered plausible attenuations of soft tissue in quantitative image analysis and better qualitative score in streaking artefacts and degree of depiction of the oral cavity compared with non-O-MAR. O-MAR may provide additional diagnostic information to clinicians upon crossreferencing the uncorrected images with the O-MAR dataset in CT in dental regions. Several corrective methods have been previously studied to reduce streak artefacts caused by high-attenuation material15,16; the vast majority of existing MAR techniques fall into the projection completion category, whereby missing/
Figure 2 There were significant differences in streaking artefacts, image sharpness, texture naturalness and degree of depiction of the oral cavity between the uncorrected images and the O-MAR images (1.5 0.7 versus 3.5 0.5; 3.8 0.4 versus 3.4 0.5; 3.9 0.3 versus 3.5 0.5; 1.8 0.6 versus 3.2 0.6, respectively; p < 0.01).
Figure 1 The image noise of O-MAR images (33 16.9) was significantly lower than that of uncorrected images (222.8 222.1; p < 0.01). There were significant differences in the radiodensity between the uncorrected images and the O-MAR images (e7.6 114.7 versus 80.2 26.7 HU, p < 0.01).
O-MAR images (33 16.9) was significantly lower than that of the uncorrected images (222.8 222.1; p < 0.01).
Qualitative image analysis Table 3 and Fig 2 show the results of the qualitative image analysis. There were significant differences in streaking artefacts, image sharpness, texture naturalness, and degree of depiction of the oral cavity between the uncorrected images and the O-MAR images (1.5 0.7 versus 3.5 0.5; 3.8 0.4 versus 3.4 0.5; 3.9 0.3 versus 3.5 0.5; 1.8 0.6 versus 3.2 0.6, respectively; p < 0.01). There were substantial interobserver agreements with respect to streaking artefacts, image sharpness, texture naturalness, and degree of depiction of the oral cavity (Kappa ¼ 0.64, 0.65, 0.71 and 0.73, respectively). Representative cases are shown in Figs 3 and 4.
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
e15
Figure 3 (a) Uncorrected image and (b) O-MAR image of a 50-year-old man with suspicion of lymph node metastasis. (b) An example of O-MAR being applied to a patient with a dental filling. On the uncorrected image, the dental region was obscured by the dark shadow caused by the metal. On the O-MAR image, the dental region was visible.
reconstruction algorithms are specifically designed to reconstruct images from incomplete projections.19,20 However, this category of algorithm is much more computationally expensive than projection completion methods, and has only recently become available for clinical use. The rapid progress of computers has resulted in major technological developments in reconstruction techniques; attempts to minimize the execution time of applications were established by the rapid development of central processing unit (CPU) chips. Owing to these developments, recently, OMAR was released for clinical use. In fact, using a standard personal computer, O-MAR image reconstruction is possible within a few minutes in most cases, a time that is acceptable for clinical use.
The most important finding of the present study is that a better qualitative score was noted in terms of the degree of depiction of the oral cavity with O-MAR compared with that in non-O-MAR, i.e., O-MAR made it possible to depict structures in areas where this was not previously possible due to metal artefacts in qualitative image analysis. When multiple metallic objects, such as dental fillings, are present in the imaging field of view, the clinical application of projection completion methods is limited.10,11 However, in the present study, O-MAR removed metal artefacts originating from dental metals effectively. Magnetic resonance imaging (MRI) can quite often uncover lesions obscured by amalgam artefact. However, O-MAR might be able to render severely degraded images useful for oral diagnosis,
Figure 4 (a) Uncorrected image and (b) O-MAR image of a 50-year-old man with suspicion of lymph node metastasis. On the uncorrected image, the dental filling caused severe metal artefacts in the CT images. On the O-MAR image, both the streak and the darkening artefacts were mitigated. Although O-MAR does not totally eliminate metal artefacts, it is capable of reducing their effects on CT images to enhance the diagnostic quality of the images significantly.
e16
M. Kidoh et al. / Clinical Radiology 69 (2014) e11ee16
preoperative assessment of implants and radiotherapy planning in patients who are unsuitable for MRI. In qualitative image analysis, there was a statistically significant increase in blurring and unnatural texture following O-MAR. A limitation of O-MAR might stem from the fact that metal projection data are discarded, resulting in a potential loss of spatial resolution near the metal implant. The cause of the unnatural texture on the O-MAR images could not be determined because detailed information about the algorithms is not available to users. Prevention of unnatural image texture might be the main technical challenge for O-MAR in the future. There are some limitations to the present study. The main limitation was that diagnostic accuracy was not evaluated; the focus was comparing the image quality of each protocol. Second, the possibility that the quantitative parameters did not really measure image quality cannot be rule out. Radiodensity (HU) and image noise (SD of radiodensity of the ROI within the oral cavity) as surrogate markers of image quality, which have been widely used for the assessment of image quality in previous studies (in general, lower image noise means better image quality). In the present study, O-MAR reduced image noise caused by metal objects; the radiodensity of the O-MAR images was more plausible than that seen on the uncorrected images. However, O-MAR images were not compared with a reference standard (e.g., MRI, surgery). Thus, further clinical studies might be required to demonstrate whether O-MAR accurately depicts reality. Other limitations to this pilot study include the small sample size and bias due to images from the same case being presented (once with, once without O-MAR) in qualitative analysis. Also O-MAR was only tested using a single CT system. In conclusion, as O-MAR enables the depiction of structures in areas in which this was not previously possible due to dental metallic artefacts, O-MAR images may have a supplementary role, in addition to uncorrected images, in oral diagnosis.
References 1. Miracle AC, Mukherji SK. Conebeam CT of the head and neck, part 2: clinical applications. AJNR Am J Neuroradiol 2009;30:1285e92.
2. Boas FE, Fleischmann D. Evaluation of two iterative techniques for reducing metal artifacts in computed tomography. Radiology 2011; 259:894e902. 3. Dong J, Hayakawa Y, Kannenberg S, et al. Metal-induced streak artifact reduction using iterative reconstruction algorithms in x-ray computed tomography image of the dentoalveolar region. Oral Surg Oral Med Oral Pathol Oral Radiol 2013;115:e63e73. 4. Abdoli M, Ay MR, Ahmadian A, et al. Reduction of dental filling metallic artifacts in CT-based attenuation correction of PET data using weighted virtual sinograms optimized by a genetic algorithm. Med Phys 2010;37:6166e77. 5. Barrett JF, Keat N. Artifacts in CT: recognition and avoidance. RadioGraphics 2004;24:1679e91. 6. Bamberg F, Dierks A, Nikolaou K, et al. Metal artifact reduction by dual energy computed tomography using monoenergetic extrapolation. Eur Radiol 2011;21:1424e9. 7. 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;38:769e75. 8. Glover GH, Pelc NJ. An algorithm for the reduction of metal clip artifacts in CT reconstructions. Med Phys. 1981;8:799e807. 9. Kalender WA, Hebel R, Ebersberger J. Reduction of CT artifacts caused by metallic implants. Radiology 1987;164:576e7. 10. Tohnak S, Mehnert AJ, Mahoney M, et al. Dental CT metal artefact reduction based on sequential substitution. Dentomaxillofac Radiol 2011;40:184e90. 11. Lemmens C, Faul D, Nuyts J. Suppression of metal artifacts in CT using a reconstruction procedure that combines MAP and projection completion. IEEE Trans Med Imaging 2009;28:250e60. 12. Imai E, Horio M, Nitta K, et al. Modification of the Modification of Diet in Renal Disease (MDRD) Study equation for Japan. Am J Kidney Dis 2007;50:927e37. 13. Matsuo S, Imai E, Horio M, et al. Revised equations for estimated GFR from serum creatinine in Japan. Am J Kidney Dis 2009; 53:982e92. 14. Yamashita Y, Komohara Y, Takahashi M, et al. Abdominal helical CT: evaluation of optimal doses of intravenous contrast materialda prospective randomized study. Radiology 2000;216:718e23. 15. Joemai RM, de Bruin PW, Veldkamp WJ, et al. Metal artifact reduction for CT: development, implementation, and clinical comparison of a generic and a scanner-specific technique. Med Phys. 2012;39:1125e32. 16. Xu C, Verhaegen F, Laurendeau D, et al. An algorithm for efficient metal artifact reductions in permanent seed. Med Phys. 2011;38:47e56. 17. Yu H, Zeng K, Bharkhada DK, et al. A segmentation-based method for metal artifact reduction. Acad Radiol 2007;14:495e504. 18. Meyer E, Raupach R, Lell M, et al. Normalized metal artifact reduction (NMAR) in computed tomography. Med Phys. 2010;37:5482e93. 19. Wang G, Snyder DL, O’Sullivan JA, et al. Iterative deblurring for CT metal artifact reduction. IEEE Trans Med Imaging 1996;15:657e64. 20. Wang G, Vannier MW, Cheng PC. Iterative x-ray cone-beam tomography for metal artifact reduction and local region reconstruction. Microsc Microanal 1999;5:58e65.