Visit for more related articles at OMICS Journal of Radiology
The metal artifacts obscure or mimic pathologies and thus severely limit the diagnostic value of CT imaging in orthopedic applications. The fundamental root cause is the beam hardening effect due to the polychromatic Xray beam. Recently, dualenergy CT with fastkVp switching technique has been introduced that allows synthetic monochromatic energy images to be generated through material decomposition. These monochromatic energy images are not only free of metal artifacts, but also available at a broad energy range for optimal contrast at bonetissue interface. This article summarizes the principle of this advanced CT technology with emphasis in its capability in suppressing metal implantinduced CT artifacts.
Keywords 
CT; Metal artifacts; Dual energy; FastkVp switching 
Introduction 
Dualenergy CT is an imaging technique that has been known and extensively studied for many decades [18]. However, due to various technical challenges, only recently has dualenergy CT imaging become a reality. Recent advances in CT scanner technologies have generated a renewed interest in dualenergy CT [914], which has led to the commercialization of dualenergy CT systems available for routine clinical use [10,1214]. 
Dualenergy CT is a special CT imaging procedure in which two CT scans of a patient are acquired in different Xray tube potentials (and spectra) and used to perform energy and materialselective reconstruction of the patient. Relative to conventional singleenergy CT imaging, dualenergy CT imaging offers the capability to enhance material differentiation and reduce beam hardening effect. 
Conventional dualenergy CT exam is accomplished by socalled “rotaterotate” technique: A lowkVp CT exam is acquired, and then the patient is translated back to the origin, followed by the acquisition of a highkVp CT exam. Similar to Digital Subtraction Angiograph (DSA), this technique is very sensitive to patient motion because the time interval between the two kVps is in the order of seconds. As a result, poor spatialtemporal registration between high and lowkVp Xray beams is a major source of image artifacts in conventional dualenergy CT imaging. Compared to singleenergy CT, motioninduced artifacts, such as blurring of the edges and streaks centered on objects that are moving are more severe in dualenergy CT. Furthermore, patient motion can be falsely interpreted as a change in tissue composition, which, typically manifested as lightanddark edge effect around the moving objects, can cause inaccurate material densities and/or misdiagnosis (Figure 1). 
To address the motion issue, a fastkVp switching (FKS) dualenergy CT imaging method, where kVp is rapidly switched between low and highkVp in adjacent views, has recently been proposed [1214]. Compared to conventional dualenergy CT imaging, FKS dualenergy CT imaging has several benefits including fine temporal view registration, helical and axial acquisitions, and full field of view. It also presents several design challenges that warrant careful considerations. 
Image Acquisition 
Xray tube/generator 
The fundamental solution to avoid the motion issue is to acquire the low and highkVp projections on a viewbyview basis in a single gantry rotation. This enables precise spatialtemporal registration of two different kVps, thus freezing motion and significantly reducing artifact. 
The Xray generation system must enable the rapid kVp switching to achieve sufficient energy separation and view sampling speed. The generator and tube must be capable of reliably switching between 80 and 140 kVp, and have the capability to support sampling rate as quickly as every 150 microseconds. To ensure the signal fidelity, Xray generator with ultralow impedance of tens of microseconds is also necessary. 
Following the acquisition, correction calibrations are applied to the data. The rise and fall of the kVp waveform complicates the spectral calibration. Figure 2 illustrates the actual kVp waveform employed in FKS dualenergy acquisition. The nonideal kVp rise/fall makes it difficult to find a fixed kVp that matches the same spectral response as the fastkVp switching. 
To estimate the effective energy, the unknown spectra of the lowand highkVp are estimated as the weighted linear combination of single kVp spectra [15]. The unknown Xray spectra of the low and high kVp views are estimated as the weighted sum of several known single kVp spectra: 
(1) 
where S_{k}(E) are the basis spectra of the single kVps, N_{k} is the total number of the basis spectra, and α_{k} are the weights of the basis spectra. The selfnormalized detector response to this spectrum can be written as: 
(2) 
in which, μ_{d}(E) is the linear attenuation coefficient of the detectors, t_{d} is detector thickness, μ_{b}(E,d) and I_{b}(d) are respectively the linear attenuation coefficient and the thickness of bowtie material b corresponding to detector channel d. If one denotes 
(3) 
Equation 2 can be simplified to 
(4) 
R(d) can be measured through a fast switching air scan, and G_{k}(d) can be calculated based on the system geometry. Hereby the weighting coefficients α_{k} can be solved from Equation 4 by least square fitting. The overall spectrum is decomposed into a superposition of several basis spectra through the measurement of the detector response to the bowtie attenuation. 
The above spectrum estimation technique is graphically demonstrated in Figure 3. P_{c}(I) epresents the equivalent Xray attenuation of the estimated spectrum by Equation 4. P(V_{e},I) is the measured Xray attenuation of the fastkVp system. It is obvious that the estimated spectrum P_{c}(I) matches the actual spectrum P(V_{e},I) with minimal difference. 
Detector 
The detector is a key contributor to fastkVp switching acquisitions through its scintillator and data acquisition system. Detector primary decay and afterglow performance are critical to avoiding spectral blurring between views. Primary speed and afterglow refer to the decays of light emitting from the scintillator for several to tens of milliseconds after the Xray source is switched off. This phenomenon is analogous to the decay of light signal on the television screen after it is turned off. The residual signal from one view smears information contained in the next during a scan, thereby causing degradation of spatial resolution and undesirable crosscontamination of spectra 
Gemstone scintillator material (GE Healthcare, Waukesha, WI) is a complex rare earth based oxide, which has a chemically replicated garnet crystal structure. This lends itself to imaging that requires high light output, fast primary speed, very low afterglow. Gemstone has a primary speed of only 30 ns, or 100 times faster than GOS (Gd_{2}O_{2}S), while also having afterglow that is only 20% of GOS, making it ideal for fast sampling [16]. The capabilities of the scintillator are also paired with an ultrafast data acquisition system (up to 7 enabling simultaneous acquisition of low and highkVp data at customary rotation speeds. 
Flux 
Compared to singleenergy CT, the traditional flux issues are more challenging in FKS dualenergy CT imaging. There hereby needs to be a strategy for balancing the flux between the two spectra and a need for noise reduction processing (which will be discussed later in Section 4). 
In dualenergy imaging, the flux ratio between the low and highkVps is in general constrained in such a way that the ContrasttoNoise Ratio (CNR) is maximized. Multiple studies [1719] have suggested the CNR is maximized with ~30% flux allocation, defined as the percentage of entrance skin exposure of the lowkVp to that of the low and highkVp combined. That is, roughly speaking, the exposure ratio between the low and highkVp is 
(5) 
The Xray exposure is a function of Xray energy spectrum, beam filtration, geometry, and the tubecurrenttime product (mAs). The FKS dualenergy CT system employs identical geometry and beam filtration for the 140 and 80 kVp acquisitions, which leaves only mAs adjustable. For the Xray tubes used in CT, the Xray exposure of a 140 kVp beam is roughly threefold of that of an 80 kVp beam [20]. Therefore, according to Equation 5, the mAs ratio between low and highkVp views should be kept around 3/2. Based on above analysis, the view time ratio between the low and highkVp views should be 60%40%. Figure 5 demonstrates the importance of balancing flux through modulating the view time in FKS dualenergy CT imaging. A simulated low contrast lesion (2 cm in diameter) is embedded in an oval phantom to show the impact of various view time ratios on the image quality. Comparing the results, 60%40% view time ratio clearly offers the best CNR for the lesion detection. 
Dose 
FKS dualenergy CT has been designed to minimize the additional dose relative to single energy. In a recent dose and low contrast detectability (LCD) comparison [13,14], the effectiveness of this sampling scheme with respect to dose was demonstrated by matching the LCD at a slice thickness equal to 5 mm and object size of 3 mm. The LCD is a clinicalrelevant image quality metric that quantifies image noise performance, making it ideal as an image quality target when measuring dose. The lower the LCD is, the less contrast a lesion of certain size (in mm) must have in order to be detected at a given confidence level (usually 95%). 
Figure 6 compares the measured CTDI_{vol} (in mGy) between a FKS dualenergy and routine (i.e., singleenergy) abdominal CT exams as a function of monochromatic energy (in keV, dualenergy only) and tube current (in mA, singleenergy only). The SingleEnergy (SE) measurements are denoted as “SE xxx mA”, where “xxx” describes the tube current, while the dualenergy measurements are denoted as “Mono xx keV”, where “xx” represents the monochromatic energy (monochromatic energy will be discussed later in Section 2.3.2). All other acquisition protocols were the same between the two systems compared: kVp=120 kVp, gantry rotate speed=1.0s, bowtie=large body, slice thickness=5 mm, display field of view=22.5 cm). 
From Figure 6, one can see that, with a tube current of 360 mA, the singleenergy abdominal CT exam yields a LCD of 0.426% and a CTDI_{vol} of 29.18 mGy for a 3 mm object. The FKS dualenergy abdominal CT exam (65 keV) produces a nearly identical LCD of 0.422% (0.01% = 0.1 HU) with a CTDI_{vol} of 33.43 mGy for the same object size, or just 14% higher than that of a routine singleenergy abdominal CT exam. These results were obtained using the uniform section of Catphan 600 phantom, which represents a patient with ~20 cm waterequivalent diameter. 
Image Reconstruction 
Projectionspace material decomposition 
The mass attenuation coefficient across the Xray spectrum is a function of two independent variables: photoelectric effect and Compton scatter [1]. Based on this principle, the low and highkVp projection data can be retrospectively transformed into a pair of basis materials (such as water and iodine). 
Through a mathematical change of basis one can express the energy dependent attenuation observed in two kVp measurements in terms of two basis materials [14]: 
(6) 
where is an arbitrary ray path, and denote the highand lowkVp projection data, represents the density of the basis material i, α, β, χ, δ, ε are polynomial coefficients, and the subscript H and L refer to the high and low kVp, respectively. 
There are a couple of important facts in Equation 6 that should be noted here. First of all, and must be measured along the same ray path . This can be easily satisfied by FKS dualenergy CT imaging. The fastkVp switching mechanism ensures the low and highkVp projection data are spatially and temporally coregistered (Note: Strictly speaking, the low and highkVp views incur a small angular offset relative to each other. To obtain truly coregistered projection pair, these views are interpolated to the same angular positions prior to material decomposition). 
Secondly, the highorder terms in Equation 6 are critically important to account for spectral variation over the field of view due to source spectrum, bowtie filter, detector performance, and multimaterial beam hardening effects [1]. As a consequence, projectionspace material decomposition provides the opportunity for more quantitative precision than may be achieved with singleenergy imaging. This is the key difference between FKS dualenergy CT imaging and imagespace dualenergy CT imaging, which usually require additional beam hardening correction to recover the spectral information [10,21]. 
To solve Equation 6, dualenergy projection data corresponding to different thicknesses of water and iodine are acquired. Since modern CT scanners contain a large number of detector elements, Equation 6 is overdetermined and the resolution for polynomial coefficients α, β, χ, δ, ε, etc. can be found easily through least square fitting: 
(7) 
Image reconstruction 
Through material decomposition, the energy dependent attenuation measurements contained in kVp projections are transformed into energy independent basis material projection data corresponding to the two basis material pair (e.g., water and iodine). 
Although the pair of basis material projection data (sinogram) essentially contains all useful information about the material being imaged, they are difficult to understand and interpreted by clinicians. A more useful form, which clinicians are familiar with, is the reconstructed images. 
Having the identical geometry, the same reconstruction algorithm to reconstruct the singleenergy CT images can therefore be applied to the first basis material projection data to obtain the corresponding basis material density image. The step is repeated for the second basis material as well. 
An example is shown in Figure 8. Basis material density images represent the effective density for the anatomies necessary to create the observed low and highkVp attenuation measurements. For instance, pure water appears as 1,000 mg/cc in a water density image, 50 mg/cc of diluted iodine is labeled as such in an iodine density image, etc. Any nonbasis material is mapped to both basis materials. For this reason, basis material density images are sometimes called “material density map”. 
Monochromatic energy image 
Given the material basis density images, one can compute attenuation data that would be measured with a monoenergetic Xray source by combining the material density images to create a monochromatic image at any specific energy level (in keV), E [14]: 
(8) 
where (μ_{i}/ρ_{i}) represents the mass attenuation coefficient for material i. For consistency with the Hounsfield Unit, one can normalize the attenuation measurement with respect to water. A clinical example is shown in Figure 9. 
From Equation 8, the expected image noise (variance) in the monochromatic image can be expressed based on linear system theory and noise propagation principle: 
(9) 
Equation 9 implies that the noise in the generated monochromatic energy image is also energy dependent. This has been confirmed through phantom experiment. Figure 10 plots the measured noise (in HU) as a function of monochromatic energy level, in the range of 40 to 140 keV, for a 20 cm water phantom. The noise curve exhibits a global minimum near 65 keV for the particular acquisition protocol. 
Effective atomic number 
The Xray linear attenuation coefficient of a periodic element, μ, can be expressed as a function of the element’s material properties and E: 
(10) 
where σ, N_{A}, and A are the total effective cross section, Avogadro’s number and the mass number. 
The total effective cross section of Xray radiation in function of the chemical composition of the materials can be modeled as a combination of the Compton and photoelectric effects [22]: 
(11) 
where Z_{eff} is the effective atomic number. The coefficients a, b, and c depend only on the energy and their values can be obtained from NIST [23]. 
By plugging (11) into (10), the linear attenuation can be further expressed as 
(12) 
The simultaneous equations are formed using Equation 12 for two monochromatic energy images μ(E_{1} ) and μ(E_{2}). Solving the simultaneous equations, Z and ρ can be obtained as follows, 
(13) 
Z_{eff} describes the periodic elements most closely representing its energy dependent attenuation behavior, hence it often provides insight regarding the material’s chemical composition. Knowledge of the effective atomic number of critical tissue types or specific contrast agents may be leveraged to define and import custom basis materials, allowing for the enhancement of specific anatomical structures. 
Figure 11 shows graphically the distribution of several materials commonly encountered in diagnostic radiology in the basis material space (water and bone). Materials are clearly separated based their chemical decomposition. The effective atomic number of each material is reflected by the angular slope. Noise is primarily responsible for the scatter of points in each material and the overlap between different materials. 
Noise Suppression 
It is well known that the basis material density images have a much lower SignaltoNoise Ratio (SNR) than singleenergy CT images. This can be easily demonstrated by the following simple analysis. Let’s define the SNR of Iodine in a lowkVp image as 
(14) 
Then the SNR of Iodine in a basis material density image is 
(15) 
By comparing (15) with (14), and using the fact that and are very close for the most of relevant energy levels, it can be concluded that 
(16) 
A reduction of noise can be achieved by increased exposure. But based on Xray physics, a noise reduction by a factor of n requires an exposure increase by a factor of n^{2}, which leads to unacceptable dose levels for most patients. Therefore, noise suppression has to be automatically applied to FKS dualenergy CT imaging and on the basis material density images in order to enhance the quality of the image while preserving the density values. This allows for a quantitative basis material density image with good image quality. 
Noise reduction techniques that attempt to directly reduce noise in basis material density images have also been sought after. To this date, none of them have had success in reducing noise to satisfactory levels while minimally affecting iodine contrast without introducing artifacts. 
A more significant improvement in SNR, however, results from the understanding the physical property of the noises that exist in basis material density images. It has been proven that the noises in the two resulting basis material density images are negatively correlated [2,8]. Taking advantage of this property, several noise suppression algorithms have been developed [8,24]. These algorithms subtract a weighted highpass filtered version of the first basis material density image (e.g., water) to noise reduce the complimentary basis material density image (e.g., iodine). 
These algorithms are effective in suppressing noise. However, they are at the risk to introduce a detrimental artifact. Although the highpass filtered image is smoothed, edge structures and blood vessels full of contrast medium are added to the complimentary basis material density image, causing “crosscontamination” that changes the accuracy of the density values. 
The solution, however, exists if the correlated anisotropic diffusion algorithm is employed [25]. The correlated anisotropic diffusion algorithm diffuses both basis material density images (e.g., water and iodine) at the same time. The correlated diffusion strength consists of components corresponding to the image gradients from both images, and the ratios between image gradients. After the diffusion, a noise mask can be computed for each basis material density image as the difference between the original image and its filtered version. Finally, a water (or iodine) noise mask is weighted and then added to the iodine (or water) density image to cancel the correlated noises. 
The effectiveness of the correlated anisotropic diffusion algorithm is demonstrated in Figure 12. In this clinical example, the original water and iodine density images are very noisy. Noise reduced water density image using a singlepass algorithm, where the crosscontaminated structures are clearly visible throughout the liver area. By comparing the complimentary noise reduced iodine density image, one can correlate the contaminations in the water density image with the iodinated hepatic vessels in iodine density images. Using the twopass algorithm, noise reduced water density image is free of contamination. 
Clinical Applications in Orthopedics 
FKS dualenergy CT imaging provides diagnostic information beyond that have been found in conventional singleenergy CT imaging, in a manner consistent with workflow, and that increases the efficacy of clinical diagnosis. The clinical application and visualization of dualenergy data is presently an area of active research. 
Research into FKS dualenergy CT imaging has been conducted in multiple clinical applications. A discussion of abdominal applications may be found in [15,26]. More recently, attention has been given to apply FKS dualenergy CT imaging to orthopedics applications, due to its excellent temporalspatial resolution and potential for reduced beam hardening and reduced metalrelated artifacts. 
One research team compared the extent of metal artifacts between conventional singleenergy CT and FKS dualenergy in an animal model with two 5.5 mm diameter stainless steel scoliosis rods inserted into the paraspinal thoracolumbar regions. They found that the monochromatic energy image from FKS dualenergy CT shows noticeable less metal artifacts and significantly better interpretability than singleenergy CT (p<0.05) [27,28]. 
In another study, the researchers conducted first a phantom study of size and CT number of titanium and stainless steel plates, then a clinical study involving 26 patients with metallic hardware. They concluded that FKS dualenergy CT imaging can improve the delineation of prosthesis and periprothetic regions which would otherwise be corrupted by artifacts. 
The example displayed in Figure 13 demonstrates the increased accuracy in hip implant assessment from FKS dualenergy CT imaging, through improved differentiation of hip bone and metallic implant, enabling accurate measurement of joint spacing. 
Another excellent example of metal artifacts from pedicle screws is shown in sagittal reformatted postoperative spinal CT (Figure 15a). The reconstructed FKS dualenergy CT image (Figure 15b) displays significantly less streak artifacts, enabling much improved visualization of the dural sac and spinal muscles. 
Conclusion 
Metal artifact reduction still remains a clinical challenge for orthopedics applications. The fundamental root cause is the beam hardening effect due to the polychromatic Xray beam. Dualenergy CT with fastkVp switching technique allows synthetic monochromatic energy images to be generated through material decomposition. The monochromatic energy images are inherently free of beam hardening, and thus metal artifacts are remarkably reduced. In addition, Dualenergy CT with fastkVp switching also provides a broad energy range for optimal contrast at bonetissue interface. Preliminary clinical studies have demonstrated this new technology’s potential in improving image quality and reducing metal artifacts for a broad range of orthopedics applications. 
References 

Figure 1  Figure 2  Figure 3  Figure 4  Figure 5 
Figure 6  Figure 7  Figure 8  Figure 9  Figure 10 
Figure 11  Figure 12  Figure 13  Figure 14  Figure 15 
Make the best use of Scientific Research and information from our 700 + peer reviewed, Open Access Journals