Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 History

First attempts to use spectral information in computed tomography date back to the late 1970s (Millner et al. 1979; Avrin et al. 1978; Chiro et al. 1979; Genant and Boyd 1977). At that time, two separate scans were acquired and either projection data or reconstructed data were postprocessed. However, the lacking stability of the CT density values, the long scan times which often caused patient motion between scans, the limited spatial resolution, and the difficulty of postprocessing were the main reasons why the method never achieved broad clinical acceptance (Kelcz et al. 1979). With the necessity to acquire both scans separately, the use of contrast material and its differentiation by dual-energy or spectral analysis was impossible. Since 2006, dual-energy CT has experienced a revival and success in clinical application. The main reason was the introduction of dual-source CT that made it possible to acquire two spiral scans simultaneously with different X-ray spectra by running the two X-ray sources at different tube voltages (Flohr et al. 2006; Johnson et al. 2007a). Today, there are several technical approaches to dual-energy CT.

2 X-ray Spectra

Generally, there are three requirements for spectral CT imaging: First, X-ray sources providing X-ray quanta with different energies are necessary. Obviously, this can be achieved optimally when using two separate X-ray sources emitting different photon energies. However, this criterion is basically also met by single X-ray sources, because they generally provide polychromatic spectra. The X-ray spectrum of a normal tube with a tungsten anode consists of a continuous part, the Bremsstrahlung, and discrete peaks according to the energy levels of the electrons in the shell of the tungsten atoms. Figure 1 shows the X-ray spectra that are obtained from an X-ray tube operated at 140 and 80 kV. Obviously, the area under the curve for equivalent tube currents differs by a factor of about 4–5, and the tube current needs to be adapted to obtain a similar output of quanta from both tubes. The higher energy spectrum is dominated by the characteristic lines of the tungsten anode, while the lower energy spectrum mainly consists of Bremsstrahlung. In this case, the mean photon energies are 71 and 53 keV, respectively, for the primary spectra. It is desirable to have as little overlap as possible between the spectra; therefore, the lowest and highest potentials offered by the respective CT scanner are always used for dual-energy acquisitions. A tube voltage lower than 80 kV is not useful because too much of the quanta would be absorbed by the human body; values higher than 140 kV are generally not available and would result in very low soft tissue contrast that likely would contribute very little to tissue differentiation. Obviously, it is not possible to obtain monochromatic X-rays with today’s tube technology; monochromatic X-ray sources so far do not provide a sufficient output of quanta for clinical applications. An exception is monochromatic synchrotron radiation that can be used for experimental setups. As synchrotrons are far too large for use in a rotating gantry, the object has to be rotated between the source and the detector, i.e., it is not possible to use them for clinical imaging.

Fig. 1
figure 1

Spectra of the Straton tube at 140 and 80 kV potential. The peaks represent the characteristic lines of the tungsten anode and the continuous spectrum is a result of Bremsstrahlung. The mean photon energies are 53 and 71 keV, respectively

Full size image

3 Detector Technology

The second requirement is that the detector has to be able to differentiate quanta of different energies. With technologies in use in clinical CT today, this is not directly possible with a single detector. The detector integrates the fluorescent light intensities of all photons detected during a single readout interval, but does not give account of their energy. Current approaches either rely on entirely separate X-ray sources and corresponding separate detectors, on reading out the projection data at different time points, or on using a two-layer or “sandwich” detector with different spectral sensitivities. In the near future, cadmium-based materials such as CdZnTe may serve as semiconductors for photon-counting detectors that resolve the energy of each individual photon. However, this detector technology cannot yet cope with the high photon flux required for clinical CT.

These first two requirements define the spectra of the X-ray sources and the corresponding detectors, representing the sensitivity of the systems for photons, detectors of different energies. The more these spectra differ, the stronger the contrast-to-noise ratio of the resulting spectral information, as long as both sufficient transmission and attenuation can be achieved in the human body at these photon energies.

4 Tissue Properties

The third requirement is a sufficient difference in spectral properties of the materials under investigation. The attenuation that is quantified in CT to characterize different body tissues is caused by three physical processes. Compton scatter is the largest component of attenuation. It is related to the electron density and not to the protons in the atomic core that would allow a material differentiation (McCullough 1975). Rayleigh scatter also is related to the electrons and not to the atomic core, but contributes only a negligible amount. The third process, the photoelectric effect, is strongly related to the atomic number, Z, of the material, i.e., to the number of protons of the atomic core (Fig. 2). Therefore, only elements with a considerable difference in Z values will be distinguishable by their spectral properties. This difference can be characterized with the so-called dual-energy index (DEI):

$$ {DEI}=\frac{{x}_{80}-{x}_{140}}{{x}_{80}+{x}_{140}+2000},$$

where x80 is the CT value in HU at 80 kV tube potential and x140 the value of the respective voxel at 140 kV (Johnson et al. 2007a).

Fig. 2
figure 2

Sketch of a dual-source CT system. Two tubes and detectors are mounted orthogonally. To obtain dual-energy datasets, the tubes are operated at different tube voltages, e.g., 80 and 140 kV (green and violet). Additionally, a filter (indicated in red) can be applied to harden the high-energy spectrum

Full size image

According to Alvarez and Macovski (1976), the photoelectric interaction with the K shell is proportional to the third power of the atomic number (Z). Therefore, high values apply for Z values of 53 (iodine) or 54 (xenon). The elements of which the human body consists, i.e., hydrogen (Z = 1), oxygen (Z = 8), carbon (Z = 6) and nitrogen (7), have low Z numbers and hence do not show a sufficient photo effect and spectral behavior that would allow a differentiation (Michael 1992). Bone with its high content in calcium (Z = 20) and fat, which only consists of hydrogen and carbon, represent tissues that differ from other body tissues significantly, which also explains their good differentiation in standard CT. This feature has also been employed in approaches for the quantification of obesity or for the identification of calcifications in pulmonary nodules (Cann et al. 1982; Svendsen et al. 1993). The most clinically useful application of dual-energy CT can be expected for the differentiation of iodine (Kruger et al. 1977; Riederer and Mistretta 1977; Nakayama et al. 2005), which is generally used in CT as a contrast agent and the distribution of which can be masked by the underlying tissue.

5 Dual-Source CT

At present, there are three different approaches to clinical dual-energy CT commercially available or under investigation. The most straightforward approach is dual-source CT with two X-ray sources running at different voltages with two corresponding detectors (Fig. 2). Additionally, a filter can be used to further harden the higher energy spectrum, i.e., to rid the spectrum of low-energy quanta. This setup has several advantages. First, established X-ray sources and detector materials can be used. Also, the image reconstruction can rely on established methods. But the most important benefit of this technology is that the tube voltage and current can be adjusted freely to obtain the largest possible difference in photon energies with similar total amounts of quanta from both tubes. A disadvantage of this approach is the considerable hardware effort, making the system expensive. Also, the space in the CT gantry is only sufficient for a smaller second detector so that the field of view of the dual-energy scans is restricted. The acquired projection data primarily have to be reconstructed by standard filtered backprojection, separately for the two simultaneously acquired spiral datasets. The fact that the acquired projection data have an offset of 90° at equal z-axis positions means that a primary postprocessing of projection data is impossible because there are no equivalent projections. Still, data at equivalent z-positions are sampled simultaneously, so the angular offset does not imply a temporal offset between the acquisitions. A problem is posed by the cross-scatter that is oriented perpendicular to the primary course of beam, thus hitting the second detector and contaminating its data. However, comparative Monte Carlo simulations show a good spectral separation with dual-source CT (Fig. 3) (Kappler et al. 2009).

Fig. 3
figure 3

X-ray spectra of the two tubes running at 80 and 140 kV with 0.9 mm titanium and 3.5 mm aluminum filters on both and an additional 0.4 mm tin filter on the high-energy tube

Full size image

6 Rapid Voltage Switching

Another approach that requires less hardware effort is rapid voltage switching. With this method, the tube voltage follows a pulsed curve, and projection data is collected twice for every projection, one at high and one at low tube voltage (Fig. 4). Optimally, the tube current should be modulated at least inversely to the voltage so that the difference in the number of photons is kept as small as possible (Grasruck et al. 2009). The major advantage of this approach is that it can be realized at considerably lower cost, because no major additional hardware is required. Limitations include the slow acquisition, because the rotation time has to be reduced to less than half to allow the collection of the additional projections. Also, the course of the tube current and voltage retain a non-rectangular, curved shape so that the resulting spectral difference does not correspond to the nominal tube voltages. The adaptation of the current also has limitations so that there is generally a far lower signal at low energy than at high energy. Figure 5 shows the resulting spectra with a rather large difference in size.

Fig. 4
figure 4

Sketch of a rapid kV switching system. There is only one tube and detector, but the tube voltage is switched rapidly between two levels. Optimally, the current is adjusted at least inversely

Full size image
Fig. 5
figure 5

There is an obvious difference in the total number of photons in the spectra because the output of quanta is smaller at low tube voltages, and this difference cannot be compensated fully by adjusting the current

Full size image

7 Layer Detector

A third approach is not to generate different photon spectra but to work with two detector layers that have their maximum sensitivity for different photon energies (Fig. 6). The sensitivity is determined by the scintillator material, e.g., consisting of ZnSe or CsI in the top and Gd2O2S in the bottom layer. This setup has the advantage that only one standard tube is required. Disadvantages include the hardware effort for the layer detector and the lower dose efficiency of such a setup. Also, the obtainable spectral difference is rather low so that the contrast of the spectral information is limited or requires additional dose. Figure 7 shows the sensitivity profile of a layer detector with ZnSe and Gd2O2S scintillators.

Fig. 6
figure 6

Sketch of a layer detector system. There is only one X-ray tube running at constant voltage. The dual-energy information is derived from two layers of the detector with different sensitivity profiles. The top layer may, e.g., use CsI or ZnSe as the scintillator so that it is more sensitive to low-energy quanta, while the bottom layer may consist of Gd2O2S as the standard material

Full size image
Fig. 7
figure 7

The sensitivity spectra of the two layers indicate a similar overall sensitivity for high- and low-energy quanta, but a considerable overlap which limits spectral contrast

Full size image

8 Sequential Acquisition

Another approach that is not being implemented currently is sequential filtering or voltage switching, i.e., two subsequent rotations are acquired in a sequence mode at the same table position with different tube voltages and optionally an additional filter. This would imply a rather minor hardware effort and may be a viable option especially for systems with broad detectors and a rather high number of simultaneously acquired slices. An obvious disadvantage is the rather long delay between both acquisitions, which is too long to preclude artifacts from cardiac or respiratory motion or changes in contrast material opacification.

9 Radiation Exposure

Regarding radiation exposure, dual-energy CT based on dual-source CT acquisitions does not require a higher patient dose than a routine CT scan of the same body region. It is possible to tailor the tube current such that the dose from both tubes matches that of a routine single-source CT protocol (Johnson et al. 2007a). Recent comparative phantom studies with external validation have proven that dual-energy acquisitions can provide similar or even improved contrast-to-noise ratios at equivalent dose (Schenzle et al. 2010). This is in contrast to investigations by Ho et al. 2009 who observed two to three times higher doses for dual-energy CT. However, their setup was based on a single-source system using rapid voltage switching and contained neither a normalization of image noise nor of dose, so the lower energy spectrum was obtained with the same tube current time product as the single energy scan. In another study with normalization to equivalent low-contrast detectability (Li et al. 2010), Li et al. observed an additional dose of 14% in the body and 22% in the head with rapid kV switching. Specific clinical studies comparing the dose efficiency of different dual-energy CT systems are lacking, but it is evident that dual-energy CT does not necessarily imply an increased dose.

10 Clinical Applications

Meanwhile there are numerous clinical applications of dual-energy CT. Among these are bone removal from the carotid (Morhard et al. 2009) or peripheral runoff (Sommer et al. 2009) CT angiography datasets, the reconstruction of virtual noncontrast images or quantification of iodine enhancement in lesions of solid organs (Graser et al. 2009), the depiction of iodine (Thieme et al. 2008; 2009) or xenon gas (Chae et al. 2008) distribution in the lung and the differentiation of kidney stones (Graser et al. 2008), or the identification of gout tophi (Johnson et al. 2007b). Also, image quality and display can be optimized, e.g., by generating monoenergetic images (Voit et al. 2009), or images with optimized contrast (Holmes et al. 2008). As this more specific or functional additional information can be obtained without an additional dose, it should be exploited whenever diagnostically useful.

11 Summary

In summary, dual-energy CT offers the possibility to exploit spectral information for diagnostic purposes. There are different technical approaches that all have inherent advantages and disadvantages. Some systems are commercially available, and there is already quite a number of well-established clinical applications.