Dual-Energy Computed Tomography

Physical Principles, Approaches to Scanning, Usage, and Implementation: Part 1

      Keywords

      Key points

      • Spectral computed tomography (CT) material differentiation relies on differences in energy-dependency of the attenuation of different materials.
      • The photoelectric effect has a strong energy dependence, and the attenuation due to the photoelectric effect is highly dependent on the atomic number (Z) of the element.
      • Elements with a high atomic number, such as iodine, that have a strong energy dependence can be exploited for spectral CT scanning.
      • Current commercially available spectral CT scanners are dual-energy CT scanners that may consist of 1 or 2 tubes, or use specialized layered detectors for spectral separation.
      • Multienergy scanning systems, such as photon counting scanners, are under development but not yet commercially available for routine clinical use.

      Introduction

      Dual-energy computed tomography (DECT) or spectral computed tomography (CT) is an advanced form of CT that uses different X-ray spectra to enhance material differentiation and tissue characterization. Current commercially available clinical systems use 2 different photon spectra for scanning, and therefore, the term DECT is sometimes used interchangeably with spectral CT. However, it is noteworthy that the term spectral CT could also encompass more advanced systems capable of discrimination between more than 2 spectra, such as the photon counting scanners currently under investigation and development. Therefore, strictly speaking, the terms are not synonymous and DECT is a subset of spectral CT.
      Applications of DECT for clinical use were initially explored in the 1970s.
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • McCollough C.H.
      • Leng S.
      • Yu L.
      • et al.
      Dual- and multi-energy CT: principles, technical approaches, and clinical applications.
      • Chiro G.D.
      • Brooks R.A.
      • Kessler R.M.
      • et al.
      Tissue signatures with dual-energy computed tomography.
      • Genant H.K.
      • Boyd D.
      Quantitative bone mineral analysis using dual energy computed tomography.
      • Kelcz F.
      • Joseph P.M.
      • Hilal S.K.
      Noise considerations in dual energy CT scanning.
      • Millner M.R.
      • McDavid W.D.
      • Waggener R.G.
      • et al.
      Extraction of information from CT scans at different energies.
      • Macovski A.
      • Alvarez R.E.
      • Chan J.L.
      • et al.
      Energy dependent reconstruction in X-ray computerized tomography.
      However, the technological and computational advances necessary for implementation of DECT and successful introduction into the clinical arena were not yet made. Therefore, attempts for implementation were temporarily abandoned, just to be revived later with the introduction of the first DECT system in the clinical arena in 2006.
      • Flohr T.G.
      • McCollough C.H.
      • Bruder H.
      • et al.
      First performance evaluation of a dual-source CT (DSCT) system.
      • Johnson T.R.
      • Krauss B.
      • Sedlmair M.
      • et al.
      Material differentiation by dual energy CT: initial experience.
      DECT scanning, as the name implies, is based on image acquisition with 2 different energy spectra. The data obtained are then combined in order to generate images for routine clinical interpretation or for more advanced material characterization. The objectives of this 2-part review are to provide an overview of the (1) physical principles behind spectral CT scanning and material differentiation, (2) major spectral CT acquisition systems available clinically, and (3) basics of implementation and use of the technology in clinical practice.

      Fundamental principles of spectral computed tomographic scanning and material characterization

      Overview

      With conventional, single-energy computed tomography (SECT), a polychromatic beam is emitted by a single source (X-ray tube), passes through the patient resulting in attenuation of the beam, and is captured by an array of detector cells. The resulting projection data, after preprocessing and reconstruction via sophisticated computer algorithms, is rendered into CT slices that are used for diagnostic interpretation. With DECT, on the other hand, projection data are obtained at 2 different energy spectra instead of one, and the information acquired is then blended to create images for routine diagnostic interpretation. More advanced tissue analysis and material characterization are also feasible with these data sets.
      The key advantage of DECT over SECT is that by acquiring data at 2 different energy spectra, it is possible to use sophisticated computer algorithms to combine the different energy data in order to evaluate tissue attenuation at different energies rather than at a single effective energy. Because different types of materials and tissues may attenuate X-rays differently at different energies depending on their elemental composition, DECT may be used to perform tissue characterization beyond what is possible with conventional SECT. For successful DECT scanning and material characterization, the following fundamental considerations related to the scanner type and tissues being characterized must be taken into account.

      Fundamentals of Dual-Energy Computed Tomography Scanning: Factors Related to the Scanner

      For acquiring the projection data, the scanner must use different energy spectra and separately record the high- and low-energy measurements. This has been implemented in the clinical setting in multiple ways. One way is through simultaneous use of 2 different imaging chains, each with its own dedicated source and detector. Another way to accomplish this is via a single X-ray source that is capable of fast switching between 2 energy levels and a detector with very fast readout capability; technical innovations in modern X-ray source technology—taking advantage of the high-voltage controls for X-ray generation—enable this capability. In yet another design, it is possible to acquire high- and low-energy data sets by use of specialized detectors that have 2 different scintillator layers, one for high energy and the other for low energy, built into them. Various methods for acquiring dual-energy data sets are discussed in detail later in this article.
      For the purposes of material differentiation, it would be ideal for each of the 2 different energy X-ray beams to be composed of monochromatic energies. However, with current X-ray tube technology used in the clinical setting, it is not possible to generate monochromatic X-ray spectra. Therefore, clinical DECT scanners use polychromatic X-ray sources but attempt to have as little overlap between the different energy spectra as possible. Fig. 1 shows a typical X-ray spectrum at 80 kilovolt peak (kVp) and 140 kVp, including inherent X-ray tube filtration, generated with the XSpect simulator.
      • Dodge C.W.
      A rapid method for the simulation of filtered x-ray spectra in diagnostic imaging systems.
      For DECT systems relying on different tube voltages for spectral separation, the standard peak energies used for scan acquisition are typically at 80 and 140 kVp. For some dual-source scanner models, 90 or 100 kVp with a filter may be used instead of 80 kVp, especially for scanning heavier patients. Alternatively, energies lower than 80 kVp, for example, 70 kVp, may also be used with some models or for specialized applications such as in pediatric imaging. For the high-energy acquisitions, 150 kVp may be used instead of 140 kVp with some models. Protocols may also vary depending on the scanner vendor or the specific application under consideration.
      Figure thumbnail gr1
      Fig. 1Typical X-ray spectrum for X-ray tube voltages of 80 kVp and 140 kVp, including inherent X-ray tube filtration (generated with the XSpect simulator). The peaks represent characteristic lines of a tungsten anode, and the continuous spectrum is the result of bremsstrahlung. This example illustrates the polychromatic nature of the X-ray beams used for DECT scanning. For DECT scanning, it is desirable to achieve as much separation as possible between the low-energy and high-energy spectra. In addition, the tube voltage used must not be too low or too high. If it is too low, it will be excessively absorbed by the body, yielding little information; if it is too high, it will not yield useful information because of poor tissue contrast. In routine DECT, 80-100 kVp (low-energy spectrum) and 140-150 kVp (high-energy spectrum) are commonly used settings for this purpose, although there may be variations depending on the scanner model, filters used, or the specific clinical application at hand.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      The reason for the choice of energies is that typically, at peak energies less than approximately 80 kVp, too few photons are generated and, in addition, a large proportion of photons would be absorbed by the body, and therefore, not generate any clinically useful information.
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • Johnson T.R.
      Dual-energy CT: general principles.
      For the high-energy acquisitions, voltages higher than 140 kVp (or 150 kVp for some models) are typically not available on all DECT scanners. Furthermore, such high energies may result in higher dose and too little soft tissue contrast, limiting applications in the clinical setting, although they could still be useful for specialized applications and for discriminating between low Z and high Z materials (discussed in greater detail in the following section).

      Fundamentals of Dual-Energy Computed Tomography Scanning: Factors Related to the Materials or Tissues Being Evaluated

      Beyond the specific technique used for image data acquisition, factors related to the materials or tissues of interest must be taken into account for successful clinical application of DECT (Box 1). It is important to understand the basis of DECT material characterization as well as the strengths and limitations of this technique. X-ray attenuation is governed by 2 main processes, with a third process only contributing a negligible amount. At the typical tube voltages used in CT, Compton scattering accounts for the greatest contribution to the overall attenuation. Compton scattering is a function of both the electron density of the tissue and the tube voltage and spectrum. The electron density of the tissue is the dominant factor, as Compton effect is only minimally dependent on photon energy.
      • Johnson T.R.
      • Krauss B.
      • Sedlmair M.
      • et al.
      Material differentiation by dual energy CT: initial experience.
      Physical basis of material characterization and differentiation using dual-energy computed tomography
      • DECT material differentiation relies on energy dependence of CT attenuation for different materials.
      • Compton scatter and photoelectric effect are the main processes accounting for attenuation during CT scanning.
      • Whereas the Compton effect is minimally dependent on photon energy, the photoelectric effect is strongly X-ray energy dependent.
      • Because of its strong energy dependency, the photoelectric effect is key for material differentiation based on spectral properties with DECT.
      • The photoelectric effect is also highly dependent on the atomic number (Z) of the element being imaged.
      • Elements with a high atomic number, such as iodine (Z = 53) and calcium (Z = 20), have strong spectral properties and can be readily differentiated from low-Z materials (such as H, C, N, and O); this property forms the basis of multiple clinical applications for DECT scanning.
      • To distinguish different materials or tissues based on their spectral properties, there must be a sufficient difference in their atomic number (Z) or effective atomic number.
      The other important physical process accounting for CT attenuation is the photoelectric effect. This effect, which is strongly dependent on the atomic number or Z (ie, the number of protons within the nucleus) of the elements that constitute the tissue under consideration, is particularly important for spectral CT.
      • Alvarez R.E.
      • Macovski A.
      Energy-selective reconstructions in X-ray computerized tomography.
      Photoelectric interactions are strongly energy dependent, and as a result, key for DECT material characterization. The third physical process, Rayleigh or coherent scatter (related to the electrons), only accounts for a very small percentage of interactions and attenuation and is typically considered negligible in conventional absorption-based CT.
      For elements or tissues to be distinguishable based on their spectral properties, there must be sufficient difference in their atomic number or Z. Understanding the impact of such elemental or tissue properties is critical when planning applications of DECT in any research or clinical setting. As an example, common elements found in the human body, such as hydrogen (Z = 1), carbon (Z = 6), nitrogen (Z = 7), and oxygen (Z = 8), have low, very similar atomic numbers. As a result, these materials do not have sufficient component of photoelectric interactions: their attenuation is relatively low and very similar to each other at different energies, precluding reliable differentiation based on their spectral properties (Fig. 2).
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • Michael G.J.
      Tissue analysis using dual energy CT.
      Figure thumbnail gr2
      Fig. 2Example of tissues with weak and strong spectral characteristics based on their elemental composition. Axial non-contrast-enhanced CT image of the neck acquired in dual-energy mode using a fast kVp switching scanner is shown (A). Region of interest analysis was performed comparing the spectral Hounsfield unit attenuation curves of muscle (green) to that of the thyroid gland (blue) (B). Most of the soft tissue in the human body, including muscle, is composed of low-Z materials such as oxygen (Z = 8), carbon (Z = 6), and hydrogen (Z = 1). As a result, there is little energy dependency of measured attenuation of muscle on an uninfused study (B; green curves). The thyroid gland, on the other hand, contains iodine (Z = 53) with strong energy dependence due to photoelectric effect. Note the marked energy-dependent increase in its attenuation at low energies approaching the K-edge of iodine (33.2 keV) (B; blue curves). Because most clinically used CT contrast agents are iodine based, iodine’s strong energy dependence can be exploited in a variety of clinical settings.
      Elements with high atomic numbers and large differences in their atomic numbers, on the other hand, have a stronger energy dependence and can be distinguished from lower-Z materials using DECT. Among these elements, one of the prime candidates that is of clinical interest is iodine (Z = 53) (see Fig. 2). Most clinically used CT contrast agents are iodine based and are widely used for a variety of indications that include oncologic imaging and angiography. Iodine’s strong energy dependence can therefore be exploited in a variety of settings in which contrast-enhanced CT scans are obtained for material characterization, iodine quantitation, and improving diagnostic evaluation using DECT. Among the elements that are intrinsic to the human body and have a relatively high atomic number, calcium (Z = 20) is another prime candidate. This element has been used in a variety of clinical applications in body imaging and head and neck imaging. These applications are covered in greater detail in subsequent articles in this series.

      Overview of current and emerging dual-energy computed tomographic systems

      The first DECT scanner approved for clinical use, a dual-source CT scanner, was introduced into the market in 2006. This scanner was followed a few years later by a CT scanner with fast kVp switching and fast detector technology. Since then, multiple scanners based on different dual-energy technologies have become available for clinical use from different vendors. There has also been refinement of different scanner types for optimizing image quality, postprocessing capabilities, radiation exposure, and ease of use. As is the case with any type of high-end technology, this is an evolving area with continued refinements and iterations being introduced. In this section, an up-to-date overview of the major DECT scanner types currently available is provided. The focus is on the fundamentals, such as mode of acquisition, as well as the advantages and disadvantages of each system. Additional information that is specific and pertinent to each scanner type is also provided as needed. It should be noted that for commercially available scanners, depending on the vendor, more than one model or generation of that scanner type may be available. Therefore, even for the same overall design, there may be important variations in the capabilities offered. This article focuses on the fundamentals of each design. A complete description of the variations in different models of the same scanner type is beyond the scope of this article, and interested readers are referred to vendor-provided documentation and literature. This section concludes with a discussion of emerging spectral CT platforms such as photon counting scanners.

      Dual-Source Dual-Energy Computed Tomography

      As the name implies, a dual-source scanner consists of 2 source and detector combinations
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • McCollough C.H.
      • Leng S.
      • Yu L.
      • et al.
      Dual- and multi-energy CT: principles, technical approaches, and clinical applications.
      • Johnson T.R.
      • Krauss B.
      • Sedlmair M.
      • et al.
      Material differentiation by dual energy CT: initial experience.
      • Johnson T.R.
      Dual-energy CT: general principles.
      • Petersilka M.
      • Bruder H.
      • Krauss B.
      • et al.
      Technical principles of dual source CT.
      • Krauss B.
      • Schmidt B.
      • Flohr T.G.
      Dual source CT.
      (Siemens AG, Forchheim, Germany) (Fig. 3). The source-detector combinations are at a near perpendicular angle, allowing the same volume to be scanned simultaneously with high- and low-energy spectra. The obvious advantage of this system is that separate tubes are used for generating the high- and low-energy spectra. The use of separate tubes enables independent adjustment of the tube voltage and current and helps optimize separation of the low- and high-energy spectra. Furthermore, the use of separate tubes facilitates balancing the total amounts of quanta emitted from the 2 tubes.
      Figure thumbnail gr3
      Fig. 3Dual-source DECT (Siemens AG). Schematic illustration of dual-source detector combination scanners with the 2 imaging chains in a nearly orthogonal configuration, allowing the same slice to be scanned simultaneously at the 2 energies. Yellow is used to illustrate the low-energy spectrum, and blue, the high-energy spectrum. Typically, 80-100 kVp (low energy spectrum) and 140-150 kVp (high energy spectrum) are used depending on the model, but other combinations may be used for specific applications. Because there are 2 separate source-detector pairs, a filter can be placed to harden the high-energy spectrum. For some higher generation dual-source scanners, it is possible to use a higher kVp for the lower spectrum (90 or 100 instead of 80 kVp) in conjunction with a filter. Because of the limited space in the CT gantry, there is only sufficient space for a smaller second detector, which in turns places restrictions on the usable field of view of the dual-energy CT mode, as shown in the illustration.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      Because of the use of 2 X-ray tubes, a filter may be applied to the tube emitting the higher energy spectrum in order to harden the spectrum (ie, reduce associated low-energy quanta from the polychromatic emission). For some scanner models, an additional filter may be available for use with the tube emitting the lower energy spectrum as well. If the second filter is used, acquisitions will typically be performed at 100/140 kVp instead of 80/140 kVp. This enables acquisition of scans with lower image noise for larger patients, whereby 80 kVp may not yield optimal results. An additional advantage of 2 separate tubes is that the combined tubes can provide a higher total X-ray flux in a given amount of time, which is also beneficial for larger patients. The use of separate tubes also allows the application of established SECT technology and algorithms.
      Among the disadvantages, one pertains to challenges posed by the limited space in the CT gantry. Because of this, there is only sufficient space for a smaller second detector, which in turns places restrictions on the usable field of view of the DECT mode (eg, 27, 33, or 35 cm, depending on the model) as compared with 50 cm for single-energy mode acquisitions. There is also the problem of cross-scatter from the primary course of one source-detector combination hitting that of the detector from the second source-detector combination at a perpendicular angle, contaminating its data with additional bias and noise. Therefore, technical modifications and optimizations are performed to reduce cross-scatter and corrections applied during image reconstruction (using different methods depending on the scanner model) to mitigate, at least in part, these effects.
      This design also results in challenges for postprocessing of the projection data. Because the projection data are acquired at different angles, for a given z-axis position, there is an offset of 90°. As a result, there is at least 70 milliseconds (ie, one-quarter rotation time) delay or “temporal skew” between the high and the corresponding low projection measurements. This temporal skew makes it hard to perform material decomposition in the projection domain, especially if there is any patient motion. Hence, the high- and low-energy datasets are separately reconstructed by filtered back projection, and material decomposition is performed afterward in the image domain, which may lead to imperfect elimination of beam hardening and reduced material decomposition accuracy. Last, there are considerable additional hardware requirements for this system compared with the single-source scanners, posing technological challenges and imposing certain limitations on the longitudinal or detector z-coverage of the system because the detector cost could become prohibitive.

      Single-Source Dual-Energy Computed Tomography with Rapid kVp Switching: Gemstone Spectral Imaging

      This scanner consists of a single source and detector combination
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • Chandra N.
      • Langan D.A.
      Gemstone detector: dual energy imaging via fast kVp switching.
      • Kalender W.A.
      • Perman W.H.
      • Vetter J.R.
      • et al.
      Evaluation of a prototype dual-energy computed tomographic apparatus. I. Phantom studies.

      Xu D, Langan DA, Wu X, et al. Dual energy CT via fast kVp switching spectrum estimation. Paper presented at: Medical Imaging 2009: Physics of Medical Imaging. Lake Buena Vista (FL), March 14, 2009.

      (GE Healthcare, Waukesha, WI, USA) (Fig. 4). With this scanner, the tube voltage follows a square wave form, and projection data are collected for twice the number of view angles, half at high- and the other half at low-tube voltage. This approach is made possible by the fast sampling capabilities of a proprietary, garnet-based scintillator detector with very low afterglow, referred to as Gemstone. During a single rotation, the image acquisition using this scanner is on a view-by-view basis, alternating between low and high kVp. Since the temporal skew or the delay between the high and low projections is as low as 50 microseconds, there is excellent spatiotemporal registration, and material decomposition is performed directly in the projection domain, making it quantitatively accurate and robust against any possible patient motion. With current Gemstone spectral imaging (GSI) scanners, axial or helical acquisitions are obtained for the full field of view of 50 cm.
      Figure thumbnail gr4
      Fig. 4Single-source DECT with rapid kVp switching: GSI (GE Healthcare). Schematic illustration of this type of a single source-detector combination system. Yellow is used to illustrate the low-energy spectrum, and blue, the high energy spectrum, typically 80/140 kVp. DECT projection data are acquired by very fast switching between low- and high-energy spectra combined with fast sampling capabilities of a proprietary, garnet-based scintillator detector with low afterglow for spectral separation at each successive axial or spiral view.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      On the source side, the generator and tube used for these scanners are capable of reliably switching between the 80 and 140 kVp voltages, with sampling periods as fast as 50 microseconds. For this approach to be successful, a highly specialized generator capable of very rapid transitions in tube potential as well as very fast electronics and detector materials are required.
      Overall, the advantages of rapid kVp switching or GSI DECT systems are excellent temporal registration and a cost-efficient design offering an opportunity to increase the longitudinal detector coverage up to 160 mm. These systems also enable scanning over the entire full field of view of 50 cm. In addition, material decomposition can be performed in the projection space, which helps to reduce beam-hardening artifacts in virtual monochromatic images (VMIs) generated in this manner (VMIs and other DECT reconstructions are discussed in part 2 of this review). If needed, with these systems, additional spectral analysis or reconstruction of different VMIs or maps can also be performed retrospectively in the image space based on the data generated from the original material decomposition and reconstruction from the projection data.
      One challenge in the design and implementation of such a system is the issues related to the relative photon flux between the 2 energies. In order to address this challenge, the scanners are designed to balance the flux, with allocation of additional time for the low kVp relative to the high kVp acquisition. Not only does this approach serve to reduce photon starvation conditions, but when complemented by the appropriate choice of rotation speed, it also achieves a more balanced flux state between the 2 energies and neutralizes patient radiation exposure.
      Disadvantages of this system include those related to adaptation of the current that can result in a relative reduction of signal from the low-energy spectrum, but this is partly compensated for by the additional time allocated for the low kVp acquisition, as described earlier. Last, in practice, the time profile of the tube voltage has a slightly trapezoidal shape rather than a purely rectangular shape. As a result, the spectral difference is slightly lower than the nominal tube voltages, and this has to be accounted for during image reconstruction.

      Layered Detector Dual-Energy Computed Tomography

      In the systems discussed so far, spectral energy separation is mainly dependent on design or alterations at the X-ray source and generator level that result in generation of the 2 different energy spectra. With layered detector or “sandwich” detector DECT,
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • McCollough C.H.
      • Leng S.
      • Yu L.
      • et al.
      Dual- and multi-energy CT: principles, technical approaches, and clinical applications.
      • Vlassenbroek A.
      Dual layer CT.
      • Alvarez R.E.
      • Seibert J.A.
      • Thompson S.K.
      Comparison of dual energy detector system performance.
      on the other hand, spectral separation is achieved at the level of the detector (Philips Healthcare, Andover, MA, USA) (Fig. 5). These systems consist of a single source and detector combination. However, they use highly specialized detectors that consist of 2 scintillator layers that have maximal sensitivity for different X-ray photon energies. When seen in a cross-sectional view, the detectors consist of 2 layers of scintillators directly on top of one another, with an optional interlayer filter. The system takes advantage of the polychromatic nature of the X-ray beam, and a single scan is performed at a high energy (120 or 140 kVp). The first or top (inner) layer preferentially absorbs lower-energy photons, by design approximately 50% of the total incident photon flux. The second, bottom (or outer) layer absorbs the remaining photons, which are primarily higher-energy photons. Naturally, this scheme works only because of the polychromatic nature of the X-ray beam generated by a conventional bremsstrahlung tube.
      Figure thumbnail gr5
      Fig. 5Layered or sandwich detector DECT (Philips Healthcare). Schematic illustration of this type of a single source-detector combination system in which spectral separation is achieved at the level of the detector. This system takes advantage of the polychromatic nature of the beam produced at the source and highly specialized detectors that consist of 2 layers having maximal sensitivity for different energies. The first layer (yellow) preferentially absorbs the low-energy photons, by design approximately 50% of the total incident photon flux. The second layer (blue) absorbs the remaining high-energy photons.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      One advantage of this system is its excellent temporal registration. Because energy separation is at the level of the detector and does not rely on different energy spectra generated at the source, there is no time lag between acquisitions of the different energy spectra. Therefore, this system is well suited for material decomposition in the projection domain, which would make it quantitatively accurate and robust for possible patient motion. Along similar lines, there is also perfect spatial registration of the acquired data to create the complete spectral dataset, without the need to compensate for shifting or interpolation that may be seen with systems relying on energy separation at the X-ray source.
      The fact that the tube always operates at high kVp results in a high total X-ray power, which is advantageous for larger patients. Because of the detector design with most low-energy photons being absorbed by the first layer, there is in effect “filtration” of those photons and hardening of the spectrum absorbed by the second layer. Similar to the fast kVp switching systems, scanning is performed at the full field of view of 50 cm. Last, because spectral separation is at the level of the detector, these systems always acquire scans in DECT mode, allowing retrospective spectral evaluation for all scan acquisitions. Because there is perfect alignment of acquired spectral data, material decomposition can be performed in the projection and image domains. Noise correlation in the projection domain can be used to improve material separation and reduce noise on low-energy VMIs.
      The main disadvantage of this system is its lower energy separation, because the scintillator absorption properties do not offer a sharp distinction between lower- and higher-energy photons. As a result, the material differentiation contrast is decreased unless a higher radiation dose is used. An earlier design choice to mitigate this challenge is to use an interlayer filter between the 2 scintillator layers. The use of an interlayer filter improves the energy separation but also reduces the dose efficiency. The noise level can be balanced between low- and high-energy acquisitions by designing individual-layer thicknesses in order to try and achieve comparable noise levels at the 2 different energies. Because spectral separation is exclusively at the level of the detector, this system does not permit alterations at the source that may optimize the balance between low- and high-energy spectra emitted, unlike the fast kVp switching or dual-source systems. This system does not have the problems of cross-scatter discussed earlier for dual-source scanners, but is susceptible to a different type of cross-scatter between the 2 detector layers. The technical challenges and expense for these systems are related to their specialized detector hardware requirements. Because of the relatively recent introduction of this system in the market, at this time there are fewer studies available that use this DECT technique. As a result, the clinical efficacy of this design when compared with dual-source or fast kVp switching systems is largely unknown. However, this is likely to change with time with more widespread availability and use of this type of scanner.

      Single-Source Dual-Energy Computed Tomography with Beam Filtration at the Source: TwinBeam Dual-Energy Computed Tomography

      A relatively recent addition to clinically available DECT systems is the TwinBeam DECT scanner
      • Euler A.
      • Parakh A.
      • Falkowski A.L.
      • et al.
      Initial results of a single-source dual-energy computed tomography technique using a split-filter: assessment of image quality, radiation dose, and accuracy of dual-energy applications in an in vitro and in vivo study.
      • Kaemmerer N.
      • Brand M.
      • Hammon M.
      • et al.
      Dual-energy computed tomography angiography of the head and neck with single-source computed tomography: a new technical (Split Filter) approach for bone removal.
      (Siemens AG) (Fig. 6). This system consists of a single source and detector combination, and spectral separation is achieved at the level of the source. However, unlike the fast kVp approach, the beam is prefiltered using 2 different materials and split into high- and low-energy beams. A split filter consisting of gold and tin is placed at the output of the tube, resulting in separation of the beam into a side with a lower-energy spectrum and a side with a higher-energy spectrum. The corresponding halves of the detector are then used for detection of the low- and high-energy spectra (see Fig. 6).
      Figure thumbnail gr6
      Fig. 6Single-source DECT with beam filtration at the source: TwinBeam DECT (Siemens AG). Schematic illustration of this type of a single source-detector combination system in which a split filter consisting of gold and tin is placed at the output of the tube, resulting in separation of the beam into low- and high-energy spectra. The corresponding halves of the detector are then used for detection of the low- and high-energy spectra.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      Advantages of this system include the ability to image the full field of view, and lesser hardware complexity and lower cost compared with all above systems. TwinBeam may even be incorporated as an upgrade to some scanner models. The main disadvantage is that a different portion of the patient is irradiated by the low- and high-energy spectra. Hence, a helical scan is needed so that each voxel scanned at one energy is eventually also scanned with the other energy. The resulting temporal skew, however, is very high, and there is relatively poor temporal registration between high- and low-energy scans of any given voxel. Another important challenge is that a central 2- to 3-mm portion of the beam will have a mixed energy spectrum due to the finite focal spot size. As a result, that portion of the data cannot be used for material discrimination. There is also potential for cross-scatter originating from one side of the beam contaminating data at the other side of the beam. There are in addition limited ways of balancing the photon flux of the low- and high-energy spectra for optimal spectral differentiation beyond what the X-ray beam filters provide. These systems have become commercially available only recently, and therefore, at this time, there are relatively few clinical studies using this system design. This will undoubtedly change over time and more data on the performance of these systems is likely to emerge with an increase in their availability and use.

      Dual-Energy Computed Tomographic Scanning Using Sequential Acquisitions

      One of the earliest and technologically most straightforward ways to obtain DECT scans, at least from a hardware standpoint, is by acquiring 2 different scans sequentially.
      • Johnson T.R.C.
      • Kalender W.A.
      Physical background.
      • McCollough C.H.
      • Leng S.
      • Yu L.
      • et al.
      Dual- and multi-energy CT: principles, technical approaches, and clinical applications.
      • Millner M.R.
      • McDavid W.D.
      • Waggener R.G.
      • et al.
      Extraction of information from CT scans at different energies.
      • Alvarez R.E.
      • Macovski A.
      Energy-selective reconstructions in X-ray computerized tomography.
      With this approach, the spectral data at 2 different energies are acquired sequentially at the same table position or a range of table positions using different tube voltages (Fig. 7). The basic scheme can be optionally enhanced with the use of an additional reconfigurable filter similar to the dual-source or TwinBeam scanners.
      Figure thumbnail gr7
      Fig. 7Sequential scanning approaches to DECT scanning. This is one of the earliest and technologically most straightforward ways to obtain DECT scans. With this approach, the spectral data at 2 different energies are acquired sequentially at the same table position, or a range of table positions using different tube voltages. Although simpler to implement, this approach has significant limitations because of the delay between the acquisition of low- and high-energy data, as discussed in the text. One way to minimize the delay between the acquisition of low- and high-energy data is by alternating scanning of high and low kVp data for each gantry rotation, instead of scanning the entire volume with multiple rotations at one energy followed by the other.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      The obvious advantage with this approach is that there is little to no significant hardware modification required for scanning, and an established technology can be used for the acquisitions. The major and significant disadvantage is the temporal skew between the high- and low-energy acquisitions. The inherent delay or temporal skew can pose a limitation on evaluation of any process requiring a high temporal resolution, such as moving organs as may be seen in cardiac imaging. It would also pose a limitation on processes in which there is a change in contrast opacification, such as angiographic acquisitions or even routine studies evaluating tissue enhancement. Furthermore, any patient motion between the different energy acquisitions can result in significant distortion of spectral data. In its simplest form, sequential scans may be obtained with any CT scanner, and the acquired data combined afterward for spectral analysis.
      One way to partially mitigate the most significant disadvantage of this system is by minimizing the delay between the acquisition of low- and high-energy data. This can be achieved by alternating scanning of high and low kVp data for each gantry rotation, instead of scanning the entire volume with multiple rotations at one energy followed by the other (see Fig. 7). Partial scanning techniques may also help with temporal resolution for scanning of relatively static organs, but the delays are still too long, and motion misregistration of low- and high-energy data remains a significant problem. The alternating acquisition of different energy data at each gantry position is the approach used by some scanners with DECT capabilities, such as some Aquilion ONE models (Toshiba, Tochigi, Japan) and Revolution EVO (GE Healthcare). Other commercially available CT scanners may also have the option of sequential DECT acquisitions, some limited to that of the entire scanned volume at one energy followed by the other for very limited DECT applications. However, even with the more refined approaches, major limitations regarding effects of motion and temporal misregistration remain. These factors may limit successful application and use of this technology to certain niche areas.

      Emerging Spectral Computed Tomographic Systems

      One of the most advanced spectral CT systems currently under investigation and development is the photon counting scanner (Figs. 8 and 9).
      • McCollough C.H.
      • Leng S.
      • Yu L.
      • et al.
      Dual- and multi-energy CT: principles, technical approaches, and clinical applications.
      • Schlomka J.P.
      • Roessl E.
      • Dorscheid R.
      • et al.
      Experimental feasibility of multi-energy photon-counting K-edge imaging in pre-clinical computed tomography.
      • Shikhaliev P.M.
      Tilted angle CZT detector for photon counting/energy weighting x-ray and CT imaging.
      • Wang X.
      • Meier D.
      • Taguchi K.
      • et al.
      Material separation in x-ray CT with energy resolved photon-counting detectors.
      • Taguchi K.
      • Iwanczyk J.S.
      Vision 20/20: single photon counting x-ray detectors in medical imaging.
      • Herrmann C.
      • Engel K.J.
      • Wiegert J.
      Performance simulation of an x-ray detector for spectral CT with combined Si and Cd[Zn]Te detection layers.
      • Leng S.
      • Yu Z.
      • Halaweish A.
      • et al.
      A high-resolution imaging technique using a whole-body, research photon counting detector CT system.
      • Yu Z.
      • Leng S.
      • Jorgensen S.M.
      • et al.
      Initial results from a prototype whole-body photon-counting computed tomography system.
      • Gutjahr R.
      • Halaweish A.F.
      • Yu Z.
      • et al.
      Human imaging with photon counting-based computed tomography at clinical dose levels: contrast-to-noise ratio and cadaver studies.
      • Yu Z.
      • Leng S.
      • Jorgensen S.M.
      • et al.
      Evaluation of conventional imaging performance in a research whole-body CT system with a photon-counting detector array.

      Iwanczyk JS, Nygard E, Meirav O, et al. Photon counting energy dispersive detector arrays for x-ray imaging. Paper presented at: 2007 IEEE Nuclear Science Symposium Conference Record. Honolulu (HI), October 26, 2007–November 3, 2007.

      Romman Z, Benjaminov O, Levinson R, et al. Virtual non-contrast ct of the abdomen using a dual energy photon-counting CT scanner: assessment of performance. Paper presented at: Radiological Society of North America 2009 Scientific Assembly and Annual Meeting. Chicago (IL), November 29-December 4, 2009.

      Benjaminov O, Perlow E, Romman Z, et al. Novel, energy-discriminating photon counting CT system (EDCT): first clinical evaluation—CT angiography: Carotid Artery Stenosis. Paper presented at: Radiological Society of North America 2008 Scientific Assembly and Annual Meeting. Chicago (IL), November 30-December 5, 2008.

      The principle behind these scanners is the use of photon counting detectors that are used to resolve the energy of individual photons or photon bins. Theoretically, these highly specialized and efficient detectors would count each individual incident X-ray photon and measure its energy range. Narrow selectable subranges (or bins) of the spectrum can then be used to detect and classify materials based on their spectral response, enabling robust multienergy material characterization.
      Figure thumbnail gr8
      Fig. 8Schematic illustration of a photon counting scanner, one of the most advanced spectral CT systems currently under development. These scanners use photon counting detectors to resolve the energy of individual photons or photon bins. Theoretically, these highly specialized and efficient detectors would count each individual incident X-ray photon and measure its energy. Narrow selectable subranges (or bins) of the spectrum can then be used to detect and classify materials based on their spectral response, enabling robust multienergy material characterization.
      (Courtesy of Reza Forghani, MD, PhD, Montreal, Quebec, Canada and Bruno De Man, PhD, Niskayuna, NY.)
      Figure thumbnail gr9
      Fig. 9Examples of abdominal images obtained with a full field-of-view photon-counting CT scanner prototype developed by GE Healthcare and installed at Rabin Medical Center, Israel in 2008. Two different slices are shown in 2 different representations: (A, C) monochromatic images at 70 keV and (B, D) effective Z images. Different Z numbers are mapped to different colors. The images were obtained with a 32-slice helical scan, 1-second gantry rotation, 140 kVp and 140 mA.
      (Courtesy of Dr Ofer Benjaminov, Rabin Medical Center, Israel; with permission.)
      There are multiple potential advantages of such a system. Photon-counting detectors have a higher geometric efficiency than conventional energy-integrating detectors. However, this cannot fully be exploited in whole-body CT scanners because the geometric efficiency is limited by the antiscatter collimator. Furthermore, with these detectors, an energy threshold may be applied to enable the rejection of false counts that are solely due to measured electronic noise. Hence, the impact of detector electronic noise can be entirely eliminated, or at least significantly reduced. Although electronic noise will still affect the measured energy of true individual photons, it will not alter photon counts.
      These systems are conceptually interesting because they provide energy information for each individual photon, and they may have the potential to perform K-edge imaging and detect and classify materials of potential interest at very low concentrations. As such, photon counting has the potential to provide improved spectral material characterization compared with current clinically available approaches. The potential advantages of photon counting systems include improvements in characterization of energy dependency of material attenuation and improved distinction of a material based on its specific K-edge, also referred to as K-edge imaging.
      In practice, however, there are challenges that need to be overcome. There is still a substantial overlap between the different energy bins, and it is not yet clear whether the energy separation will be a significant advantage. Another advantage of photon-counting detectors is their smaller cell size, which could offer a substantial increase in spatial resolution. However, this would have to be coupled to a smaller focal spot size and a major X-ray flux, and a dose increase would be required to maintain acceptable image noise at the higher spatial resolution. For example, if all other factors are kept unchanged, a 2-fold improvement in spatial resolution may require a 16-fold increase in radiation dose in order to maintain the same image noise.
      The main limitations and challenges of photon counting spectral CT are currently technical in nature. Although photon counting detectors are used in other disciplines such as nuclear medicine, application of this technology for CT scanning faces significant technological hurdles because of the high exposure rates and photon flux required for CT. These hurdles include pulse pileup effects that can result in loss of counts, or potentially, even “paralyze” the detector. Pulse sharing across multiple detector pixels or K-escape (reemission of a characteristic X-ray) can also occur and result in degradation of the accuracy of the recorded energy. To be effective, photon counting detectors would have to be much faster than currently used CT detectors and avoid prohibitively long scan times. Currently, there are no commercially available photon-counting scanners for clinical use, but different prototypes are available that may one day enable clinical implementation of this exciting technology and open a new era of spectral CT imaging, potentially at a molecular level.
      In addition to active investigations and development of photon counting scanners, alternative methods aimed at expanding multienergy evaluation capabilities of spectral CT are also being explored. One example is the adaptation and combination of dual-source and TwinBeam technologies described earlier to perform triple or quadruple beam acquisitions.
      • Yu L.
      • Leng S.
      • McCollough C.H.
      Dual-source multi-energy CT with triple or quadruple X-ray beams.
      Another example is the use of multi-kVp imaging, switching the X-ray tube between multiple voltages. These are other examples of exciting opportunities and areas of active research aimed at producing more robust spectral CT platforms in the future. The different types of clinical DECT scanners and spectral CT scanners under investigation or development are summarized in Table 1.
      Table 1Different types of spectral computed tomography in clinical use or under development
      Commercially available clinical DECT approaches and scanners
       Tube voltage-based DECT scanners
      Dual-source DECT
      • Two-source detector combinations at nearly 90° angles
      • Allow the same volume to be scanned simultaneously at the 2 energies
      Single-source DECT with rapid kVp switching (GSI)
      • Single source and detector combination
      • Very fast (submillisecond) switching between low- and high-energy spectra using a single source at each successive axial or spiral view enables DECT acquisition
       Detector-based DECT scanners
      Layered or sandwich detector DECT
      • Single source and detector combination
      • Highly specialized dual-layer detectors, each layer having maximal sensitivity for different energies, used for DECT acquisition
       Filtration-based DECT scanners
      Single-source DECT with beam filtration at the source (TwinBeam DECT)
      • Single source and detector combination
      • A split filter at the output of the tube results in separation of the beam into low- and high-energy spectra, that in turn are detected by their respective halves of the detector enabling DECT acquisition
       Temporally sequential scanning
      Sequential scanning of entire scan volume
      • Scan the volumes sequentially at the 2 different energies
      • In theory possible with any scanner with the use of appropriate software for merging of different energy data
      Sequential scanning of each gantry rotation
      • Alternating scanning of high and low kVp data for each gantry rotation instead the entire volume
      • Reduces temporal and spatial misregistration compared with sequential scanning of the entire volume at each energy
      Spectral CT approaches and scanners under development
       Dual-source and filtration-based combinations
      • Combination of dual-source DECT and split filter placed at the output of one or both tubes
      • Each source with split filter results in separation of the beam into low- and high-energy spectra detected by their respective halves of the detector
      • Could potentially enable the acquisition of triple or quadruple energy acquisitions
       Photon counting
      • Detector-based material differentiation
      • Narrow selectable subranges (or bins) of the spectrum can potentially be used to detect and classify materials based on their spectral “K-edge” patterns
      • If successfully implemented, it could represent the most advanced spectral CT system to date, enabling robust multienergy material characterization

      Summary

      Spectral CT is an exciting and evolving technology that has the potential to change the traditional SECT approach currently used in most clinical settings. DECT provides a new layer of information, previously unavailable using SECT. This new information has the potential to be expanded to elemental or molecular composition analysis far beyond conventional SECT scanning. In this article, the first part of a 2-part review, the fundamental principles underlying spectral CT scanning and different spectral CT systems, both commercially available or under investigation, were reviewed. Such familiarity is essential for those using or planning to use this technology optimally.
      Since the advent of DECT systems, there has been a steady increase in the applications of this technology for the evaluation of brain or head and neck abnormality. Concomitantly, there has been a steady increase in the number of commercially available systems. Multiple workflow optimizations are also being implemented to efficiently incorporate this technology into the clinical routine. Various articles in this issue review different applications of DECT and how they are being incorporated into the clinical workflow. Even at this early stage in the development of this technology, it is clear that DECT will play an increasingly important role in neuroradiology and head and neck imaging and further improve the diagnostic evaluation of our patients.

      References

        • Johnson T.R.C.
        • Kalender W.A.
        Physical background.
        in: Johnson T. Fink C. Schönberg S.O. Dual energy CT in clinical practice. Springer-Verlag Berlin Heidelberg, Berlin2011: 3-9
        • McCollough C.H.
        • Leng S.
        • Yu L.
        • et al.
        Dual- and multi-energy CT: principles, technical approaches, and clinical applications.
        Radiology. 2015; 276: 637-653
        • Chiro G.D.
        • Brooks R.A.
        • Kessler R.M.
        • et al.
        Tissue signatures with dual-energy computed tomography.
        Radiology. 1979; 131: 521-523
        • Genant H.K.
        • Boyd D.
        Quantitative bone mineral analysis using dual energy computed tomography.
        Invest Radiol. 1977; 12: 545-551
        • Kelcz F.
        • Joseph P.M.
        • Hilal S.K.
        Noise considerations in dual energy CT scanning.
        Med Phys. 1979; 6: 418-425
        • Millner M.R.
        • McDavid W.D.
        • Waggener R.G.
        • et al.
        Extraction of information from CT scans at different energies.
        Med Phys. 1979; 6: 70-71
        • Macovski A.
        • Alvarez R.E.
        • Chan J.L.
        • et al.
        Energy dependent reconstruction in X-ray computerized tomography.
        Comput Biol Med. 1976; 6: 325-336
        • Flohr T.G.
        • McCollough C.H.
        • Bruder H.
        • et al.
        First performance evaluation of a dual-source CT (DSCT) system.
        Eur Radiol. 2006; 16: 256-268
        • Johnson T.R.
        • Krauss B.
        • Sedlmair M.
        • et al.
        Material differentiation by dual energy CT: initial experience.
        Eur Radiol. 2007; 17: 1510-1517
        • Dodge C.W.
        A rapid method for the simulation of filtered x-ray spectra in diagnostic imaging systems.
        Wayne State University, 2008 (Available at:)
        • Johnson T.R.
        Dual-energy CT: general principles.
        AJR Am J Roentgenol. 2012; 199: S3-S8
        • Alvarez R.E.
        • Macovski A.
        Energy-selective reconstructions in X-ray computerized tomography.
        Phys Med Biol. 1976; 21: 733-744
        • Michael G.J.
        Tissue analysis using dual energy CT.
        Australas Phys Eng Sci Med. 1992; 15: 75-87
        • Petersilka M.
        • Bruder H.
        • Krauss B.
        • et al.
        Technical principles of dual source CT.
        Eur J Radiol. 2008; 68: 362-368
        • Krauss B.
        • Schmidt B.
        • Flohr T.G.
        Dual source CT.
        in: Johnson T. Fink C. Schönberg S.O. Dual energy CT in clinical practice. Springer-Verlag Berlin Heidelberg, Berlin2011: 10-20
        • Chandra N.
        • Langan D.A.
        Gemstone detector: dual energy imaging via fast kVp switching.
        in: Johnson T. Fink C. Schönberg S.O. Dual energy CT in clinical practice. Springer-Verlag Berlin Heidelberg, Berlin2011: 35-41
        • Kalender W.A.
        • Perman W.H.
        • Vetter J.R.
        • et al.
        Evaluation of a prototype dual-energy computed tomographic apparatus. I. Phantom studies.
        Med Phys. 1986; 13: 334-339
      1. Xu D, Langan DA, Wu X, et al. Dual energy CT via fast kVp switching spectrum estimation. Paper presented at: Medical Imaging 2009: Physics of Medical Imaging. Lake Buena Vista (FL), March 14, 2009.

        • Vlassenbroek A.
        Dual layer CT.
        in: Johnson T. Fink C. Schönberg S.O. Dual energy CT in clinical practice. Springer-Verlag Berlin Heidelberg, Berlin2011: 21-34
        • Alvarez R.E.
        • Seibert J.A.
        • Thompson S.K.
        Comparison of dual energy detector system performance.
        Med Phys. 2004; 31: 556-565
        • Euler A.
        • Parakh A.
        • Falkowski A.L.
        • et al.
        Initial results of a single-source dual-energy computed tomography technique using a split-filter: assessment of image quality, radiation dose, and accuracy of dual-energy applications in an in vitro and in vivo study.
        Invest Radiol. 2016; 51: 491-498
        • Kaemmerer N.
        • Brand M.
        • Hammon M.
        • et al.
        Dual-energy computed tomography angiography of the head and neck with single-source computed tomography: a new technical (Split Filter) approach for bone removal.
        Invest Radiol. 2016; 51: 618-623
        • Schlomka J.P.
        • Roessl E.
        • Dorscheid R.
        • et al.
        Experimental feasibility of multi-energy photon-counting K-edge imaging in pre-clinical computed tomography.
        Phys Med Biol. 2008; 53: 4031-4047
        • Shikhaliev P.M.
        Tilted angle CZT detector for photon counting/energy weighting x-ray and CT imaging.
        Phys Med Biol. 2006; 51: 4267-4287
        • Wang X.
        • Meier D.
        • Taguchi K.
        • et al.
        Material separation in x-ray CT with energy resolved photon-counting detectors.
        Med Phys. 2011; 38: 1534-1546
        • Taguchi K.
        • Iwanczyk J.S.
        Vision 20/20: single photon counting x-ray detectors in medical imaging.
        Med Phys. 2013; 40: 100901
        • Herrmann C.
        • Engel K.J.
        • Wiegert J.
        Performance simulation of an x-ray detector for spectral CT with combined Si and Cd[Zn]Te detection layers.
        Phys Med Biol. 2010; 55: 7697-7713
        • Leng S.
        • Yu Z.
        • Halaweish A.
        • et al.
        A high-resolution imaging technique using a whole-body, research photon counting detector CT system.
        Proc SPIE Int Soc Opt Eng. 2016; 9783
        • Yu Z.
        • Leng S.
        • Jorgensen S.M.
        • et al.
        Initial results from a prototype whole-body photon-counting computed tomography system.
        Proc SPIE Int Soc Opt Eng. 2015; 9412
        • Gutjahr R.
        • Halaweish A.F.
        • Yu Z.
        • et al.
        Human imaging with photon counting-based computed tomography at clinical dose levels: contrast-to-noise ratio and cadaver studies.
        Invest Radiol. 2016; 51: 421-429
        • Yu Z.
        • Leng S.
        • Jorgensen S.M.
        • et al.
        Evaluation of conventional imaging performance in a research whole-body CT system with a photon-counting detector array.
        Phys Med Biol. 2016; 61: 1572-1595
      2. Iwanczyk JS, Nygard E, Meirav O, et al. Photon counting energy dispersive detector arrays for x-ray imaging. Paper presented at: 2007 IEEE Nuclear Science Symposium Conference Record. Honolulu (HI), October 26, 2007–November 3, 2007.

      3. Romman Z, Benjaminov O, Levinson R, et al. Virtual non-contrast ct of the abdomen using a dual energy photon-counting CT scanner: assessment of performance. Paper presented at: Radiological Society of North America 2009 Scientific Assembly and Annual Meeting. Chicago (IL), November 29-December 4, 2009.

      4. Benjaminov O, Perlow E, Romman Z, et al. Novel, energy-discriminating photon counting CT system (EDCT): first clinical evaluation—CT angiography: Carotid Artery Stenosis. Paper presented at: Radiological Society of North America 2008 Scientific Assembly and Annual Meeting. Chicago (IL), November 30-December 5, 2008.

        • Yu L.
        • Leng S.
        • McCollough C.H.
        Dual-source multi-energy CT with triple or quadruple X-ray beams.
        Proc SPIE Int Soc Opt Eng. 2016; 9783