PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
IEEE Access. Author manuscript; available in PMC 2017 September 16.
Published in final edited form as:
Published online 2016 September 16. doi:  10.1109/ACCESS.2016.2602854
PMCID: PMC5119548
NIHMSID: NIHMS817150

High-kVp Assisted Metal Artifact Reduction for X-ray Computed Tomography

Yan Xi, Yannan Jin, Bruno De Man, Member, IEEE, and Ge Wang, Fellow, IEEE*

Abstract

In X-ray computed tomography (CT), the presence of metallic parts in patients causes serious artifacts and degrades image quality. Many algorithms were published for metal artifact reduction (MAR) over the past decades with various degrees of success but without a perfect solution. Some MAR algorithms are based on the assumption that metal artifacts are due only to strong beam hardening and may fail in the case of serious photon starvation. Iterative methods handle photon starvation by discarding or underweighting corrupted data, but the results are not always stable and they come with high computational cost. In this paper, we propose a high-kVp-assisted CT scan mode combining a standard CT scan with a few projection views at a high-kVp value to obtain critical projection information near the metal parts. This method only requires minor hardware modifications on a modern CT scanner. Two MAR algorithms are proposed: dual-energy normalized MAR (DNMAR) and high-energy embedded MAR (HEMAR), aiming at situations without and with photon starvation respectively. Simulation results obtained with the CT simulator CatSim demonstrate that the proposed DNMAR and HEMAR methods can eliminate metal artifacts effectively.

Index Terms: Computed tomography, metal artifact reduction, kVp switching, iterative reconstruction

I. Introduction

X-RAY computed tomography (CT) scanners are widely used in clinical imaging, thanks to its excellent performance with spatial resolution up to 2.5 lp/mm and accurate quantification keeping error within several Housfield units [1, 2]. However, high-density materials, such as metallic implants, will introduce strong artifacts and significantly degrade the diagnostic value of the CT images [3]. Since the introduction of CT technology, the metal artifact problem has been challenging and it still causes problems today.

Metal artifacts are clinically common, and caused by multiple mechanisms, including beam hardening, noise and photon starvation, scattering, non-linear partial volume and aliasing [4]. To address this problem, MAR techniques have been intensively studied for over 30 years [57]. Generally speaking, MAR algorithms can be classified into three categories, projection-domain methods [811], image-domain methods [1214], and hybrid methods [15, 16]. Projection-domain methods try to replace the projections in metal areas with values calculated from neighbors using linear interpolation, normalized interpolation or other advanced image inpainting and reprojection techniques. Within this category, the normalized MAR (NMAR) [9] is considered to be the state-of-the-art approach and therefore used as the baseline in this study. Projection-domain methods are computationally efficient but depend on certain assumptions or prior knowledge. For example, for the NMAR approach to work well, a relatively accurate prior image is crucial; and if not estimated correctly, the method is prone to new artifacts. On the other hand, image-domain method and hybrid method treat the MAR problem aided by physical and/or mathematical models, such as modeling the spectrum of incident polychromatic X-rays [17] or treating MAR as an exterior problem [15, 16]. However, these algorithms are computationally intensive, even with the rapid advancement of computer hardware.

All three categories of MAR algorithms discussed above apply to single kVp scanning. The development of dual-energy CT, such as dual source-detector [18] and fast kVp-switching [19], offers new opportunities for MAR [2023]. With the capability of material decomposition, dual-energy CT can generate monochromatic images that are free from beam hardening artifacts. Previous studies demonstrated impressive MAR results with dual-energy CT technique [2427]. However, the typical dual-energy scanning protocol (80/140 kVp) still does not resolve all the MAR challenges, especially in situations where the standard 80 kVp and 140 kVp voltages are not sufficiently penetrating. With the trend of using low-dose scanning protocols [28], photon starvation related metal artifacts may become even more common. Higher voltages (e.g.: 160kVp or megavoltage CT) could be used to improve penetration and reduce photon starvation effects, but they are generally not dose-efficient, especially in contrast exams, or only exist in radiotherapy equipment [29, 30].

Metal segmentation from a (metal-corrupted) first-pass reconstruction is often the first step for many existing MAR algorithms. However, with severe beam-hardening or even photon starvation, it is often impossible to perform an accurate segmentation especially near metal surfaces. As a result, MAR methods may introduce new artifacts. To resolve the MAR problem, we propose a high-kVp assisted scan protocol. It is based on the kVp-switching technique [31]: a standard CT scan is combined a few-view high-kVp CT scan to help with the removal of metal artifacts. In other words, a high-kVp scan is sparsely distributed during a normal CT scan. The high kVp beam can effectively penetrate metal objects and avoid photon starvation. That information can be used to help reconstruct the low kVp data. To incorporate the high-kVp assisted CT information, two kinds of MAR methods are proposed: dual-energy normalized MAR (DNMAR) and high-energy embedded MAR (HEMAR). They are designed to overcome beam-hardening and photon-starvation, respectively. To evaluate their effects, realistic simulation tests are conducted and the proposed methods are compared to an NMAR-type approach.

II. Materials and Methods

In this section, we introduce the high-kVp assisted CT scan protocol. We describe two new MAR methods (DNMAR and HEMAR) and simulation studies to investigate the image quality improvements with the new algorithms relative to the (slightly modified) NMAR method (mNMAR).

A. CT Scan Protocols

The X-ray image acquisition process can be written as

I=I0EψS(E) exp (rLμ(r,E)dr)dE,
(1)

where I0 is the total incident intensity of X-ray photons, S (E) is the X-ray spectrum that defines the differential fraction at each energy level, μ (r, E) is the linear attenuation coefficient at a position r along a transmission line L at an energy level E.

A typical 140kVp X-ray spectrum of a CT scanner is shown in Fig. 1(a). The attenuation coefficients of water, bone and titanium are plotted in Fig. 1(b). Because of the wide distribution of the X-ray spectrum and the non-uniform attenuation rates over the energy range, an X-ray beam becomes increasingly harder and less attenuated while it is penetrating an object, which is the so-called the beam-hardening effect. In a CT scanner, first or second order corrections are utilized before image reconstruction to minimize the beam hardening effect due to water and bone [32, 33]. However, there is a large attenuation gap between metallic and other materials in the human body as shown in Fig. 1(b), standard correction methods do not eliminate metal artifacts. An example is shown in Fig. 1(c). Moreover, if a metallic implant is so large or dense that it stops most X-rays (Fig. 1(d)), the missing data will cause most severe artifacts.

Fig. 1
Metal artifacts are cause by spectral or beam-hardening effects and and photon-starvation. (a) X-ray spectrum at the tube voltage of 140kVp; (b) attenuation coefficients of water, bone and titanium materials between 40 and 140keV; (c) Beam-hardening streak ...

In NMAR-type methods, the accuracy of the segmentation of metal parts in a reconstructed image is critical to the MAR performance. Thus, if an initial image is strongly corrupted by beam hardening or photon starvation as shown in Fig. 2, it is difficult to perform an accurate segmentation, especially near metallic implants. Advanced image segmentation algorithms may produce better results, but any mismatch will still introduce additional artifacts, and may not work well in all cases.

Fig. 2
Reconstruction results containing metallic implants. (a) The result with only the beam-hardening effect; and (b) the result including both beam-hardening and photon starvation.

In this paper, we propose to supplement a normal CT scan with a high-kVp sparse view sampling so that we have additional information for correction of beam hardening and photon starvation. The proposed approach is based on the fast kVp-switching technique which allows the X-ray tube to switch between different tube voltages from view to view, and almost simultaneously collects projection data at mixed kVp settings [34, 35]. In the proposed high-kVp assisted MAR, a normal CT scan is combined with a low-current and few-view high-kVp scan. In a current fast-kVp switching CT scanner, the low-kVp and high-kVp X-ray are generated for alternative view angles. In our proposed scheme, a few-view even higher kVp scan is added without adding radiation dose significantly. For the purpose of this study we assume low-kVp is available at all view angles, including those view angles at which the high-kVp data is acquired. In practice, each view angle would result in data at only one kVp and view interpolation is required to generate low and high kVp at the same view angle. According to the photon flux through the metal object, two MAR algorithms are designed, DNMAR (dual-energy normalized MAR) and HEMAR (high-energy embedded MAR), suppressing metal artifacts caused by beam hardening and photon starvation respectively.

B. mNMAR

A large number of MAR algorithms were published to combat metal artifacts. Projection completion methods are widely adopted for their simplicity. One straight-forward approach is to perform interpolation in metal-blocked areas of the sinogram. A more sophisticated approach is the NMAR algorithm, which utilizes a prior image to normalize the original sinogram and then applies interpolation to the flattened data. Although this approach works well in many cases, it may also introduce new artifacts especially when there are fine structures near the metal. The prior image is based on the segmentations of air, tissue and bone, and should be free of artifacts and close to the true image. In some cases, the metal artifacts are so severe that an accurate estimation of a prior image is already a major challenge. In this work, we slightly modify the NMAR method, hence named mNMAR. The original image (Fig. 3(a)) is modeled by mNMAR aided by a histogram analysis, as shown in Fig. 3(b) where five materials are differentiated (air, lung tissue, soft tissue, bone, and metal implants). When the projection data is generated with the input X-ray spectrum plotted in Fig. 1(a), a reconstructed image after the first-order water-based beam hardening correction is shown in Fig. 3(a). Fig. 3(d) is the result of mNMAR.

Fig. 3
MAR with mNMAR. (a) An initially reconstructed image; (b) the histogram for material segmentation; (c) the prior image; and (d) the image reconstructed with mNMAR. All images are shown with a grayscale window of [0.01, 0.04] mm−1

For fair comparison with mNMAR, histogram guided thresholding is applied to all the algorithms studied here. When photon starvation happens, the missing data is first completed by linear interpolation. In our simulation, a threshold method is used to detect the photon starvation in which the threshold is set at 15 after −log operation. It should be noted that mNMAR is also employed as part of our proposed methods, as described in the following sections.

C. DNMAR

The main idea of DNMAR is to utilize the dual-energy CT technique to address the beam-hardening problem. Since detailed structures are not needed in the generation of a prior image in mNMAR, a few-view and low-current dual-energy CT scan is sufficient for DNMAR. In our implementation, a TV-based OS-SART iterative algorithm is used in the few-view image reconstruction.

The first step in DNMAR is the dual-energy beam-hardening correction. A projection-based empirical method is adopted [36]. A projection is corrected based on a reweighted dataset,

{pcor}={i=0Iaipnormi}+{j=0Jbjphighj},
(2)

where p*=ln (I*I0), a and b are empirically determined polynomial coefficients. In this experiment, the highest order of the polynomials are I = 3 and J = 3. A simple phantom experiment (such as water-aluminum phantom) is performed to calculate these empirical coefficients. With water and aluminum chosen as the basis materials, {ai, bj} are determined by solving

min{ai,bj}FBP(i=0Iaipnormi+j=0Jbjphighj)umono22,
(3)

where FBP (·) is a standard filtered-backprojection reconstruction operator, and umono is the ideal monochromatic image. The ideal monochromatic image is obtained by assigning known monochromatic linear attenuation coefficients to the respective regions in the phantom. Since FBP is a linear operator, all the norm-kVp images and high-kVp images can be reconstructed first and the unknown coefficients ai and bj can be moved in front and solved using the least squares method.

The workflow of DNMAR is shown in Fig. 4. First, the low kVp sinogram is extracted from the normal scan and combined with the high kVp sinogram to compute a corrected sinogram that has much less beam-hardening artifacts, according to Eq. (2). The weightings of dual-energy data are determined with the coefficients described above. Since the normal-kVp data and high-kVp data cannot be obtained simultaneously, linear interpolation is applied to the norm-kVp sinogram at the projection angles of the high-kVp sinogram. Next, a TV-based few-view iterative reconstruction is performed. As the reconstructed image is used as the basis for image segmentation, detailed texture information is not needed, thus the error introduced by interpolation will not affect the final result significantly. An example is shown in Fig. 5 in which (a) is the reconstruction result with ideal simultaneous data acquisition and (b) is the result with realistic data acquisition and interpolation processing. A TV-based reconstruction algorithm is applied to both, Fig. 5(c) shows the difference image. It demonstrates that the sinogram interpolation will not substantially degrade the segmentation in the prior image. Then, the cartoonized prior image is filled with the mean material attenuation values from a single-energy reconstruction (120kVp scan). Finally, similar to the mNMAR method, the low kVp sinogram is normalized by the prior image, and an interpolation step is taken to replace the metal-corrupted data in the projection domain.

Fig. 4
Workflow of the Dual-energy Normalized MAR (DNMAR) approach.
Fig. 5
Dual-energy few-view image reconstruction. (a) the reconstruction result with ideal simultaneous data acquisition, (b) the result with realistic data acquisition and interpolation processing and (c) their differences. (a–b) are displayed in [0.01, ...

D. HEMAR

With the prior image determined by both dual-energy and single-energy reconstructions, the DNMAR algorithm works well but it will fail if photon starvation is an issue. To address this problem, we propose a supplementary MAR algorithm in the higher-kVp assisted framework: High kVp Embedded MAR (HEMAR). The higher kVp X-ray beam has a stronger penetrating capability. The key of the HEMAR is to take a few-view, lower-current, higher-kVp scan as compensation for “blind spots” of the normal CT scan so that photon starvation can be effectively avoided.

From a normal kVp CT scan, it is difficult to segment a reconstructed image with the presence of photon starvation. In contrast, with HEMAR only high-kVp data will be utilized for identification of metal regions, based on a sparse view reconstruction. The diagram of HEMAR is shown in Fig. 6.

Fig. 6
Workflow of the High-kVp Embedded MAR (HEMAR) approach.

Similar to DNMAR, the HEMAR approach also begins with the reconstruction from few-view high-kVp data. The reconstructed image is used to generate a sinogram of the same projections sampling as that of the normal-kVp scan. Metal areas in the high-kVp image are segmented and used to track the metal positions (M) in the sinogram. Then, the neighborhood (MN) of a metallic implant is marked in the sinogram as shown in Fig. 6. In the sinogram-embedding step, only the metal projection data in the normal kVp sinogram is replaced by the re-projected high-energy sinogram after a proper conversion. The conversion of metal projections from high to normal kVp is realized using an empirical polynomial ({pi}). The empirical polynomial is calculated using the data in the neighborhood MN

minp(PMN)ipi(PHEMN)i2.
(4)

Then, the empirical parameters {pi} are applied to the metal projection area M:

(PM)ipi(PHEM)i.
(5)

While the proposed transformation may not be 100% accurate when the polynomial coefficients are obtained outside the metal object shadow and then applied inside the metal object shadow, we propose that this approach will still provide an adequate transformation. To ensure the smoothness of the translation between normal kVp data and high-energy data, a feathering polynomial function is included in Eq. 4. After the embedding step, the rest of the process is the same as that for mNMAR.

In the DNMAR and HEMAR methods, there is no need for higher energy data to provide a high quality image. Thus, a few-view and low-current scan is sufficient to outline highly attenuating materials accurately, which is dose efficiency.

E. CT Simulation

Realistic simulations were performed using the CatSim simulator [26], which includes all major contrast mechanisms, such as photoelectric, Compton scattering and Rayleigh scattering, and is in an excellent agreement with data from the GE LightSpeed VCT scanner [37].

A numerical phantom with metallic rods was used in the simulation. Materials and diameters of metallic rods were adjusted to emulate different degrees of metal artifacts. As shown in Fig. 7, six experimental settings were studied with Ti-3mm, Ti-6mm, Ti-10mm, Fe-3mm, Fe-6mm, and Fe-10mm respectively (the combination denotes the metal material and the rod diameter).

Fig. 7
Numerical phantom. The metallic implants are marked in color: yellow for titanium, and red for Ferrum. Images are displayed in [0.01, 0.04] mm−1.

The CatSim simulator was configured as listed in Table I. The X-ray tube was set at 120kVp voltage and 300mA current, which is typical for a normal CT scan. For the high-kVp assisted scan, the X-ray tube was set at 160kVp and 50mA, and 90 views per rotation were acquired. It should be noted that in the experiment with double Fe-10mm metallic implants, the high-kVp scan tube current was elevated to 100mA to avoid photon starvation.

TABLE I
Imaging Parameters

III. Results

Two kinds of experiments were performed, with single and two metallic implants respectively. FBP reconstructions with a simple interpolation MAR method and the mNMAR method were performed, as shown in Fig. 8 and Fig. 9 respectively. These results indicate that the mNMAR method is effective in suppressing artifacts caused by small or weak metallic implants, as shown in Fig. 9(a, b, d, g). However, if the metal is dense and larger, there are severe residual artifacts (Fig. 9(i, k, l)). Reconstruction results from high-kVp assisted CT scans are shown in Figs. 10 and and11.11. The DNMAR and HEMAR methods were applied to remove the metal artifacts. DNMAR is very effective in eliminating metal artifacts in cases without photon starvation. Cases with photon starvation were not processed with DNMAR since it would lead to an inaccurate prior image determination. So with photon starvation cases, HEMAR is employed which shows a strong reduction of metal artifacts relative to mNMAR.

Fig. 8
Reconstruction results using the simple interpolation method. Images are displayed in [0.01, 0.04] mm−1
Fig. 9
Reconstruction results using the mNMAR method. Images are displayed in [0.01, 0.04] mm−1.
Fig. 10
Reconstruction results using the proposed DNMAR method. Since photon starvation happens in certain cases, their DNMAR results are not available, shown as blank areas. Images are displayed in [0.01, 0.04] mm−1.
Fig. 11
Reconstruction results using the proposed HEMAR method. Images are displayed in [0.01, 0.04] mm−1.

IV. Discussions and Conclusion

Metal artifacts can significantly degrade the quality of CT images, compromising diagnostic performance and therapeutic planning outcomes. In this paper, we have presented a high-kVp assisted CT scan protocol and two corresponding MAR algorithms (DNMAR and HEMAR), which are designed to address beam-hardening and photon starvations respectively. The high-kVp assisted CT scan relies on a fast-kVp switching mechanism that allows the tube voltage to switch view by view between the high-kVp and the normal kVp. Both DNMAR and HEMAR are based on the mNMAR method, but provide a more accurate prior image. According to our results, DNMAR and HEMAR can offer improved image quality in the presence of metallic implants. Generally speaking, collection of projection data at a high kVp setting will increase the radiation dose. However, with the proposed DNMAR and HEMAR methods, the increased dose is well under control due to the low-dose few-view nature of the associated protocols. The high-energy scan aims at providing a smooth artifact-free prior image, which is not necessarily of high image contrast. Hence, in our experiments the radiation dose increase is less than 10%. The tube current of a high-kVp scan is adjusted to the metal size and density to avoid photon starvation.

In conclusion, we have introduced additional critical high-energy x-ray measurements within the current CT architecture to effectively address the current challenges in the MAR area. Our proposed protocols and algorithms should have a high potential for clinical and oncological translation. In future work, we would like to evaluate the proposed methods on real patient datasets and perform task-based evaluation.

Acknowledgments

This work was supported in part by NIH under Grants R01 EB016977, R01 EB011785, and U01 EB017140

Footnotes

Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org.

Contributor Information

Yan Xi, Rensselaer Polytechnic Institute.

Yannan Jin, GE Global Research.

Bruno De Man, GE Global Research.

Ge Wang, Rensselaer Polytechnic Institute.

References

1. Kalender WA. X-ray computed tomography. Physics in medicine and biology. 2006;51:R29. [PubMed]
2. wang g, Kalra M, Murugan V, Xi Y. Simultaneous CT-MRI – Next Chapter of Multi-modality Imaging. Medical Physics. 2015 [PubMed]
3. Barrett JF, Keat N. Artifacts in CT: Recognition and Avoidance 1. Radiographics. 2004;24:1679–1691. [PubMed]
4. Boas FE, Fleischmann D. CT artifacts: causes and reduction techniques. Imaging in Medicine. 2012;4:229–240.
5. De Man B. Iterative reconstruction for reduction of metal artifacts in computed tomography. 2001
6. Glover GH, Pelc NJ. An algorithm for the reduction of metal clip artifacts in CT reconstructions. Medical physics. 1981;8:799–807. [PubMed]
7. Schulze R, Heil U, Groß D, Bruellmann D, Dranischnikow E, Schwanecke U, et al. Artefacts in CBCT: a review. 2014
8. Kalender WA, Hebel R, Ebersberger J. Reduction of CT artifacts caused by metallic implants. Radiology. 1987;164:576–577. [PubMed]
9. Lell MM, Meyer E, Kuefner MA, May MS, Raupach R, Uder M, et al. Normalized metal artifact reduction in head and neck computed tomography. Investigative radiology. 2012;47:415–421. [PubMed]
10. Wang Q, Li L, Zhang L, Chen Z, Kang K. A novel metal artifact reducing method for cone-beam CT based on three approximately orthogonal projections. Physics in Medicine & Biology. 2013;58:1–17. [PubMed]
11. Veldkamp WJH, Joemai RMS, Molen AJVD, Geleijns J. Development and validation of segmentation and interpolation techniques in sinograms for metal artifact suppression in CT. Medical Physics. 2010;37:620–628. [PubMed]
12. De Man B, Nuyts J, Dupont P, Marchal G, Suetens P. Reduction of metal streak artifacts in x-ray computed tomography using a transmission maximum a posteriori algorithm. IEEE transactions on nuclear science. 2000;47:977–981.
13. Wang G, Snyder DL, O’Sullivan JA, Vannier MW. Iterative deblurring for CT metal artifact reduction. Medical Imaging, IEEE Transactions on. 1996;15:657–664. [PubMed]
14. Zhang X, Wang J, Xing L. Metal artifact reduction in x-ray computed tomography (CT) by constrained optimization. Medical physics. 2011;38:701–711. [PubMed]
15. Lemmens C, Faul D, Nuyts J. Suppression of metal artifacts in CT using a reconstruction procedure that combines MAP and projection completion. Medical Imaging, IEEE Transactions on. 2009;28:250–260. [PubMed]
16. Zhao S, Robeltson D, Wang G, Whiting B, Bae KT. X-ray CT metal artifact reduction using wavelets: an application for imaging total hip prostheses. Medical Imaging, IEEE Transactions on. 2000;19:1238–1247. [PubMed]
17. De Man B, Nuyts J, Dupont P, Marchal G, Suetens P. An iterative maximum-likelihood polychromatic algorithm for CT. Medical Imaging, IEEE Transactions on. 2001;20:999–1008. [PubMed]
18. Graser A, Johnson TR, Chandarana H, Macari M. Dual energy CT: preliminary observations and potential clinical applications in the abdomen. European radiology. 2009;19:13–23. [PubMed]
19. Zou Y, Silver MD. Analysis of fast kV-switching in dual energy CT using a pre-reconstruction decomposition technique. Medical Imaging. 2008:691313-691313-12.
20. Xi Y. United iterative reconstruction for spectral computed tomography. IEEE Transactions on Medical Imaging. 2015 [PubMed]
21. Yang Q, Cong W, Xi Y, Wang G. Spectral X-Ray CT Image Reconstruction with a Combination of Energy-Integrating and Photon-Counting Detectors. PloS one. 2016;11:e0155374. [PMC free article] [PubMed]
22. Zhang Y, Xi Y, Yang Q, Cong W, Zhou J, Wang G. Spectral CT reconstruction using image sparsity and spectral correlation; 2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI); 2015. pp. 1600–1603.
23. Bo M, Cong w, xi y. Energy Window Optimization for X-ray K-edge Tomographic Imaging. Biomedical Engineering. 2015 [PMC free article] [PubMed]
24. Bamberg F, Dierks A, Nikolaou K, Reiser MF, Becker CR, Johnson TR. Metal artifact reduction by dual energy computed tomography using monoenergetic extrapolation. European radiology. 2011;21:1424–1429. [PubMed]
25. Lee YH, Park KK, Song H-T, Kim S, Suh J-S. Metal artefact reduction in gemstone spectral imaging dual-energy CT with and without metal artefact reduction software. European radiology. 2012;22:1331–1340. [PubMed]
26. Zhou C, Zhao YE, Luo S, Shi H, Zheng L, Zhang LJ, et al. Monoenergetic imaging of dual-energy CT reduces artifacts from implanted metal orthopedic devices in patients with factures. Academic radiology. 2011;18:1252–1257. [PubMed]
27. Meinel FG, Bischoff B, Zhang Q, Bamberg F, Reiser MF, Johnson TR. Metal artifact reduction by dual-energy computed tomography using energetic extrapolation: a systematically optimized protocol. Investigative radiology. 2012;47:406–414. [PubMed]
28. Poletti P-A, Platon A, Rutschmann OT, Schmidlin FR, Iselin CE, Becker CD. Low-dose versus standard-dose CT protocol in patients with clinically suspected renal colic. American Journal of Roentgenology. 2007;188:927–933. [PubMed]
29. Paudel MR, Mackenzie M, Fallone BG, Rathee S. Clinical Evaluation of Normalized Metal Artifact Reduction in kVCT Using MVCT Prior Images (MVCT-NMAR) for Radiation Therapy Treatment Planning. [7/1/ 2014];International Journal of Radiation Oncology*Biology*Physics. 89:682–689. [PubMed]
30. Wu M, Keil A, Constantin D, Star-Lack J, Zhu L, Fahrig R. Metal artifact correction for x-ray computed tomography using kV and selective MV imaging. Medical Physics. 2014;41:121910. [PubMed]
31. Xu D, Langan DA, Wu X, Pack JD, Benson TM, Tkaczky JE, et al. Dual energy CT via fast kVp switching spectrum estimation. SPIE Medical Imaging. 2009:72583T-72583T-10.
32. Herman GT. Correction for beam hardening in computed tomography. Physics in Medicine and Biology. 1979;24:81. [PubMed]
33. Kachelrieß M, Sourbelle K, Kalender WA. Empirical cupping correction: a first-order raw data precorrection for cone-beam computed tomography. Medical physics. 2006;33:1269–1274. [PubMed]
34. Lin XZ, Miao F, Li JY, Dong HP, Shen Y, Chen KM. High-definition CT Gemstone spectral imaging of the brain: initial results of selecting optimal monochromatic image for beam-hardening artifacts and image noise reduction. Journal of computer assisted tomography. 2011;35:294–297. [PubMed]
35. Zhang D, Li X, Liu B. Objective characterization of GE discovery CT750 HD scanner: gemstone spectral imaging mode. Medical physics. 2011;38:1178–1188. [PubMed]
36. Yu L, Leng S, McCollough CH. Dual-energy CT–based monochromatic imaging. American Journal of Roentgenology. 2012;199:S9–S15. [PubMed]
37. De Man B, Basu S, Chandra N, Dunham B, Edic P, Iatrou M, et al. CatSim: a new computer assisted tomography simulation environment. Medical Imaging. 2007:65102G-65102G-8.