Vol.:(0123456789)1 3

Chinese Journal of Academic Radiology (2020) 3:19–34 https://doi.org/10.1007/s42058-020-00029-z

REVIEW

Basic principles and clinical potential of photon‑counting detector CT

Thomas Flohr1 · Stefan Ulzheimer1 · Martin Petersilka1 · Bernhard Schmidt1

Received: 25 November 2019 / Revised: 25 November 2019 / Accepted: 17 February 2020 / Published online: 2 March 2020 © Springer Nature Singapore Pte Ltd. 2020

AbstractPhoton-counting detectors are a new technology for future computed tomography (CT) systems, with the potential to over-come major limitations of conventional CT detectors. They provide energy-resolved CT data at very high spatial resolution without electronic noise. This review article gives an overview of the basic principles of photon-counting detector CT, of potential benefits and limitations, and of the clinical experience gained so far in pre-clinical installations.

Keywords Computed tomography (CT) · Photon counting CT · Spectral CT

Introduction

Since its clinical introduction in the early 1970s, computed tomography (CT) has become the backbone of radiological diagnosis. It is an indispensable tool in the hands of the radi-ologist, widely used for both diagnostic and interventional procedures. Technological progress such as the introduction of spiral CT in 1990 [18], the broad availability of multi-detector row CT in 1999 [24, 33], and the introduction of advanced CT system concepts such as wide detector CT in 2004 [35] or dual-source CT in 2005 [10] paved the way for new applications.

Today, CT is a mature modality, and its further technical development has entered a saturation phase. Yet, there are still limitations which are difficult to overcome with current CT technology.

The in-plane spatial resolution of routine clinical CT today is limited to about 10–15 lp/cm, with a minimum sec-tion width of the CT images of 0.5–0.6 mm. This is not enough for several clinical applications, e.g., CT angio-graphic (CTA) examinations of small vessels such as the coronary arteries. As a result of limited spatial resolution, calcified plaques in the vessels appear much larger in the CT image than they are (“Ca-blooming”), hampering reliable assessment of the coronary lumen and leading to over-esti-mation of coronary artery stenosis. Coronary calcifications

with an Agatston score > 1000 have been identified as the most relevant independent predictor of uninterpretable coro-nary segments in coronary CTA [61]. Assessment of stent patency and in-stent re-stenosis is challenging and typically limited to stents with more than 3 mm diameter, even though progress has been made with recent generations of CT sys-tems [12]. Lung examinations, e.g., for interstitial lung dis-ease, will benefit from improved spatial resolution as well [32]. For special applications, such as inner ear scanning, techniques are available to increase spatial resolution beyond 20 lp/cm, e.g., by positioning a moveable comb in front of the detector pixels to reduce their aperture [11]. These tech-niques, however, come at the expense of increased radiation dose to the patient. Recently, an ultra-high-resolution CT system providing 0.25-mm section thickness was introduced (Canon Precision). The system has been evaluated in the initial studies [37, 47, 64]; routine clinical experience, how-ever, is still limited.

As another drawback, the potential of low radiation dose scanning with conventional CT systems is limited by elec-tronic noise of the measurement system. Electronic noise has a negligible effect on image quality for CT protocols using standard radiation dose. In ultra-low-dose scans, however, and for obese patients, the electronic noise may become compa-rable to the low detector signal and prevail over the typical Poisson noise of the X-ray photons. This leads to severe noise streaks in the CT images, e.g., in the shoulders or in the pelvis, and drift of CT numbers, e.g., in low-dose lung scans. If the electronic noise of the CT detector can be reduced, further radiation dose reduction and quantitative CT imaging even at low radiation dose seem feasible. Meanwhile, a conventional

* Thomas Flohr thomas.flohr@siemens-healthineers.com

1 Siemens Healthcare GmbH, Computed Tomography, 91301 Forchheim, Germany

20 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

CT detector with integrated electronics and substantially reduced electronic noise has been introduced [7]. Yet, there is room for improvement.

Since its clinical re-introduction in 2005 with dual-source CT [10], dual-energy CT has gained momentum as a tech-nique to enhance the clinical value of CT by providing infor-mation beyond mere morphology. Dual-energy CT exploits the material-specific difference in X-ray attenuation at different X-ray energies. Dual-energy CT has been used for material differentiation (e.g., for the characterization of kidney stones or differential diagnosis of gout), for improved visualization of lesions and other structures with the help of virtual monoen-ergetic images (VMIs), and for quantitation. The local iodine uptake has been evaluated as a surrogate parameter for local perfusion, e.g., for the characterization of perfusion defects in the lung parenchyma in patients with pulmonary embolism, or for differential diagnosis of abdominal lesions. Reviews on clinical applications of dual-energy CT may be found in [1, 6, 29, 31, 36, 42, 48]. Several techniques have been developed to acquire dual-energy CT data with conventional CT detectors, such as dual-source CT systems [10, 17], CT systems with fast kV switching [67], or dual-layer detector CT systems [43]. However, each of these solutions has inherent limitations. Fur-thermore, current clinical  CT systems do not enable spectral data acquisition at more than two energy levels.

Photon-counting detectors are a new technology with the potential to overcome major limitations of conventional CT detectors, by providing energy-resolved CT data at very high spatial resolution without electronic noise. Photon-counting detectors and their potential benefits were already evaluated in experimental CT benchtop systems in the first decade of the 21st century (e.g., [8]). The performance of the detectors used in these early systems, however, was not adequate for clinical CT imaging, mainly because the detectors did not tolerate the high X-ray fluxes and high photon count rates needed in medi-cal CT. Meanwhile, a significant progress in detector material synthesis and detector electronics design has been made, and photon-counting detectors are ready for pre-clinical testing in human subjects. Despite all technical advances, significant development efforts are still needed before the detectors can be broadly released in commercial CT systems.

This review article gives an overview of the basic principles of photon-counting detector CT, and of the clinical experi-ence gained so far in pre-clinical installations. Other reviews of photon-counting detector CT may be found in [27, 59, 60, 63].

Basic principles of photon‑counting CT

Basic principles of energy‑integrating solid‑state scintillation detectors

All medical CT systems today are equipped with solid-state scintillation detectors. In a two-step detection process, the absorbed X-rays are first converted into visible light in the scintillation crystal [e.g., gadolinium oxide or gadolinium oxysulfide (GOS)]. The light is then converted into an elec-trical current by a photodiode attached to the backside of each detector cell, see Fig. 1. The intensity of the scintil-lation light is proportional to the energy E of the absorbed X-ray photon, and so is the amplitude of the current pulse induced in the photodiode. All current pulses produced dur-ing the time of one reading (projection) are integrated and read-out as the detector signal of the projection. Solid-state detectors do not provide energy-resolved signals. Because of their detection principle, they are also called “energy-integrating detectors”. X-ray photons with lower energy E contribute less to the detector signal than X-ray photons with higher energy, because they produce less scintillation light. This energy weighting degrades the contrast-to-noise ratio (CNR) in the CT images, because the lower energy X-ray photons carry most of the low contrast information. Most of

Fig. 1 Schematic drawing of an energy-integrating scintillation detec-tor. a Side view; b top view. Individual detector cells made of a scin-tillator such as gadolinium oxide or gadolinium oxysulfide (GOS) absorb the X-rays (red arrows) and convert their energy into visible light. This light is detected by photodiodes attached to the backside of each detector cell and converted into an electrical current. Collimator blades are needed to suppress scattered radiation. Furthermore, the individual detector cells must be separated by optically in-transparent layers (e.g., based on TiO2) to prevent optical crosstalk—these lay-ers are “dead zones” because X-rays absorbed there do not contribute to the measured signal

21Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

the iodine signal in contrast-enhanced CT scans is generated by absorption of X-ray photons directly above the K-edge of iodine at 33 keV. These low-energy photons are down-weighted in the detector signal, and the iodine contrast in the image is reduced.

The low-level analog electric signal of the photodiodes is susceptible to electronic noise which sets an ultimate limit to potential further radiation dose reduction. Optically in-transparent separation layers must be introduced between the individual detector cells to prevent optical crosstalk. They have a width of about 0.1 mm and reduce the geometric dose efficiency of the detector: X-ray photons absorbed in the separation layers do not contribute to the measured signal, even though they have passed through the patient—from a radiation dose perspective, they are wasted dose. Today’s CT detectors have detector cells with a size of about 1 × 1 mm2 and a geometric dose efficiency of about 80–90%. Quarter-ing the size of a detector element to about 0.5 × 0.5 mm2 to double the spatial resolution will reduce the geometric efficiency if the width of the separation layers is kept con-stant—therefore, it is at least problematic to significantly increase the spatial resolution of solid-state scintillation detectors beyond today’s performance levels.

Basic principles of photon‑counting detectors

Photon-counting detectors are made of semiconductors such as cadmium telluride (CdTe) or cadmium zinc tellu-ride (CZT). In a direct conversion process, the absorbed X-rays create electron–hole pairs in the semiconductor. The charges are separated in a strong electric field between cath-ode on top and pixelated anode electrodes at the bottom of the detector, see Fig. 2. The electrons drift to the anodes and induce short current pulses which last a few nanosec-onds (10−9 s). In a pulse-shaping circuit, the current pulses are transformed into voltage pulses with a full width at half maximum (FWHM) of 10–15 ns; the amount of charge in the current pulses is translated into the pulse height of the voltage pulses. The pulse height is, therefore, proportional to the energy E of the absorbed X-ray photons. The pulses are then individually counted as soon as they exceed a threshold, see Fig. 3.

Compared to solid-state scintillation detectors, photon-counting detectors have several advantages. The individual detector cells are defined by the strong electric field between common cathode and pixelated anodes (Fig. 2), and there is no need for additional separation layers. The geometrical dose efficiency is, therefore, better than that of scintillation detectors and only reduced by the unavoidable anti-scatter collimator blades or grids. Furthermore, each “macro” detec-tor pixel confined by collimator blades may be divided into smaller detector sub-pixels which are read-out separately to significantly increase spatial resolution (see Fig. 2b).

Fig. 2 Schematic drawing of a direct converting photon-counting detector. a Side view; b top view. A semiconductor such as cadmium telluride or cadmium zinc telluride absorbs the X-rays (red arrows). They create electron–hole pairs that are separated in a strong electric field between cathode and pixelated anodes. Collimator blades are needed to suppress scattered radiation. The individual detector pixels are formed by the pixelated anodes and the electric field; there are no separation layers between them. Each “macro pixel” between two collimator blades may be divided into smaller sub-pixels, as indicated for the three left detector cells. Of course, the pixelated anodes must then be finely structured as well, which is not shown here in order not to overload the drawing

Fig. 3 Electrons created by absorbed X-ray photons in a photon-counting detector induce fast signal pulses at the anode which are counted as soon as they exceed a threshold (dashed line). The pulse height is proportional to the X-ray energy (direct conversion). Low-amplitude baseline noise does not trigger counter thresholds, and the counting signal shows no typical electronic noise component

22 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

In a photon-counting detector, all current pulses produced by absorbed X-rays are counted if they exceed a threshold energy E. Typical minimum threshold energies are rather high, > 20 keV. Low-level electronic noise below the thresh-old does, therefore, not affect the count rates (see Fig. 3).

The absence of electronic noise is a major difference com-pared to solid-state scintillation detectors, resulting in less image noise, less streak artifacts, and more stable CT num-bers in CT scans at very low radiation dose or in CT scans of obese patients. Photon-counting detectors open the potential for further radiation dose reduction beyond today’s limits.

In a basic operation mode of a photon-counting detector, all current pulses produced during the time of one reading (projection) are simply counted. All X-ray photons contrib-ute equally to the measurement signal regardless of their energy E, as soon as it exceeds a threshold. There is no down-weighting of lower energy X-ray photons as in solid-state scintillation detectors. Photon-counting detectors can, therefore, provide CT images with potentially improved CNR, in particular in contrast-enhanced CT scans using iodinated contrast agent.

In a more advanced operation mode, more than one energy threshold may be introduced for energy discrimina-tion. With four different energy thresholds as an example, counter 1 counts all X-ray pulses with an energy exceed-ing E1, while counter 2 simultaneously counts all X-ray pulses with an energy exceeding E2, and so on—see Fig. 4. The photon-counting detector will simultaneously provide four CT raw data sets r1, r2, r3, and r4 with different lower energy thresholds E1, E2, E3, and E4 for spectrally resolved measurements, see Fig. 5. Up to six different threshold values have so far been realized in prototype settings [46]. Physically, the thresholds are realized by different volt-ages which are fed into pulse height comparator circuits.

Fig. 4 Pulse train in a photon-counting detector with four energy thresholds. In this example, five X-ray photons with an energy > 25 keV are detected (blue). Three of them are also detected in the energy band > 50 keV (green), but only two exceed a threshold of 75 keV (red) and only one exceeds the upper threshold of 90 keV (black). By subtracting CT raw data with adjacent low-energy thresh-olds, “energy bin” data are obtained. In this case, two detected X-ray photons fall into the energy bin 25–50 keV, 1 falls into 50–75 keV, 1 into 75–90 keV, and 1 into 90–140 keV (for an X-ray tube voltage of 140 kVp)

Fig. 5 Contrast-enhanced kidney scan of a 71-year-old female patient acquired with a pre-clinical hybrid photon-counting CT prototype. The X-ray tube voltage was 140 kVp. The photon-counting detec-tor simultaneously provided four CT raw data sets with low-energy thresholds of 25  keV, 50  keV, 75  keV, and 90  keV (see Fig.  4).

Images reconstructed from the four CT raw data sets demonstrate decreasing iodine contrast and increasing image noise with increasing low-energy threshold (top left to bottom right), because fewer low-energy X-ray photons contribute to the image. Courtesy of National Institute of Health NIH, Bethesda, MD, USA

23Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

By subtracting CT raw data with adjacent lower energy thresholds, “energy bin” data can be produced. Energy bin b1 = r2 − r1 as an example contains all detected X-ray photons in the energy range between E1 and E2.

CT systems with photon-counting detector can pro-vide dual-energy or multi-energy information in any CT scan using a standard CT scanner without further modi-fications. Today’s established dual-energy applications—mainly based on decomposition into two base materi-als—are routinely feasible with two energy thresholds. Data acquisition with more than two thresholds opens the door to advanced material decomposition. Unfortu-nately, the availability of N energy thresholds does not imply potential differentiation of N base materials. There are only two relevant interaction mechanisms for elements without K-edge in the X-ray energy range accessible to CT (30–150 keV): Compton scattering and the photoe-lectric effect. Both have similar energy dependence for all materials. As soon as two base materials have been chosen, the energy-dependent attenuation of any other material can be described by a linear combination of the two base materials. It is, therefore, not possible to differ-entiate this other material from a mixture of the two base materials. Differentiation of two base materials requires two measurement values—dividing the energy range into more than two energy bins does, therefore, not provide relevant new information. The situation changes if a mate-rial with K-edge in the energy range accessible to CT is added to the two base materials. For a K-edge material, the energy dependence of X-ray attenuation is different, and CT measurements at three or more energies can be used for three-material decomposition (two base materials plus the K-edge material). Unfortunately, all elements naturally occurring in the human body do not have K-edges in the relevant energy range. Three- or more-material decom-position with CT data in three or more energy bins will, therefore, be limited to clinical scenarios in which two contrast agents (e.g., iodine and gadolinium, or iodine and bismuth) are simultaneously applied and need to be sepa-rated, or other heavy elements are introduced in the human body (e.g., iron, tungsten, or gold nanoparticles).

In addition to potential material decomposition, the CNR of the images can be further improved by optimized weighting of the different energy bins. Instead of just adding the bin data for the reconstruction of an image using all detected X-ray photons, higher weights may be assigned to the low-energy bin data. Giving the low-energy X-ray photons more weight will increase image contrasts, in particular in CT scans using iodinated contrast agent. Higher iodine CNR opens the potential for either radia-tion dose reduction or reduction of the amount of contrast agent in contrast-enhanced CT scans.

Challenges of photon‑counting detectors

The energy separation of a CdTe- or CZT-based photon-counting detector is reduced by undesired but unavoidable physical effects, such as signal splitting at detector pixel bor-ders which leads to erroneous counting of one high-energy X-ray photon as two or more lower energy X-ray photons in adjacent detector  pixels. Another interference factor is the energy loss of the X-rays due to K-escape, whereby the K-edges of the detector material cause preferential absorp-tion of some X-ray photon energies (Cd and Te have K-edges at 26.7 and 31.8 keV, respectively), and the corresponding release of characteristic X-rays at lower energies within the detector itself, see Fig. 6. Charge sharing, as well as fluo-rescence and the resulting K-escape events, lead to a dou-ble counting of X-ray photons at wrong X-ray energies and, therefore, to a reduction of spectral separation, see Fig. 7.

For a realistic detector model including charge sharing, fluorescence, K-escape, and other effects which cause sig-nificant spectral overlap, the energy discrimination potential with two energy bins is probably equivalent to that of a dual-kVp scan with optimized pre-filtration [19].

Energy separation may be improved by making the detec-tor pixels larger, because charge sharing at the boundaries then results in a smaller relative contribution to the total detector signal (see Fig. 7b, c).

Unfortunately, there is another physical effect which poses a limit to the size of the detectors: the voltage pulses after pulse shaping have an FWHM of 10 ns and more. At high X-ray flux rates commonly used in medical CT [up to 109 cts/(s mm2) in air at the isocenter], X-ray photons may hit a detector pixel too closely in time to be registered separately. Overlapping low-energy pulses may be incor-rectly registered as high-energy hits, and several overlapping pulses may be counted as one hit only, see Fig. 8.

Because of this so-called “pulse pile-up” the detec-tor saturates at higher X-ray flux rates, see also Fig. 9.

Fig. 6 Schematic illustration of the effects that reduce spectral resolu-tion in a photon-counting detector. Among them are charge sharing at pixel boundaries or energy loss due to fluorescence and K-escape, which lead to double counting of X-ray pulses at wrong energies

24 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

In typical clinical situations, patient attenuation leads to central flux rates well below 108 cts/(s mm2). At higher flux rates, even though the signal can be linearized in data pre-processing, pulse pile-up can lead to significant quan-tum losses and increased image noise. A way out of this dilemma is a reduction of the pixel size of the detector;

Fig. 7 Computer simulation of the detected X-ray spectra for a pho-ton-counting CT detector with two energy bins (bin1: 25–65  keV; bin2: 65–140  keV). a Ideal detector. b Detector using a realistic model, pixel size 0.225 × 0.225  mm2. Charge sharing and other effects such as K-escape lead to erroneous counting of high-energy X-ray photons in low-energy bins (see the characteristic high-energy tail of bin1  representing wrongly counted high-energy X-rays) and as a consequence to overlap between the energy bins and a reduction of spectral separation. c Detector using a realistic model, pixel size 0.45 × 0.45  mm2. Because of the larger pixel size, boundary effects such as charge sharing play a less dominant role, and spectral overlap is reduced

Fig. 8 Schematic illustration of pulse pile-up in a photon-counting detector with two energy thresholds (blue and green lines). The detected X-rays produce voltage pulses with a pulse width of 10 ns and more after pulse shaping. At high X-ray flux rates, the pulses overlap, see the dashed brown lines. Several pulses may then be counted as one hit only, see the blue and gree dots which indicate counting of a pulse. The larger the detector pixels are, the lower are the X-ray flux rates that can be registered without pulse pile-up

Fig. 9 Count rate as a function of the applied X-ray flux per energy threshold in a photon-counting detector with two energy thresholds (25 keV, 65 keV). Pulse pile-up leads to a non-linear detector count rate at higher X-ray flux and, finally, detector saturation. Lower energy thresholds are more affected

25Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

however, smaller pixels lead to more charge sharing and reduced spectral separation.

Another problem of photon-counting detectors is count-rate drift at higher X-ray flux rates. Non-homogeneously distributed crystal defects in the sensor material cause trap-ping of electrons and holes, a build-up of space charges, and a modification of the electric field distribution. This changes the characteristics of the signal pulses in the individual detector pixels and may lead to severe ring artifacts in the images at higher flux rates. However, a significant progress in material synthesis during the past years has resulted in reduced count-rate drift at clinically tolerable levels.

Pre‑clinical evaluation of photon‑counting CT

Photon-counting detectors are a very promising new tech-nology fur future CT systems. Yet, there are still technical challenges that need to be mastered before these devices can be broadly introduced into clinical CT systems. Currently, pre-clinical prototypes are used to evaluate the potential and limitations of photon-counting CT in clinical practice. We will focus on these pre-clinical installations and leave out other more experimental solutions, benchtop systems, and photon-counting micro CT systems.

A pre-clinical single-source CT system with photon-counting detector based on CZT (Philips Healthcare, Haifa, Israel) was evaluated both with phantoms and with animal scans. The system provides an in-plane field of view of 168 mm and a z coverage of 2.5 mm at the iso-center, with a rotation time of 1 s [23]. The size of the detector pixels is 0.5 × 0.5 mm2. The photon-counting detector has five energy thresholds. A lung phantom including nodules of different sizes and shapes was scanned with this system and a con-ventional CT in standard and high-resolution modes, dem-onstrating the potential of photon-counting CT to improve the assessment of lung structures due to higher resolution compared to conventional CT [23]. Improved visualization of the in-stent lumen and in-stent re-stenosis in ten differ-ent coronary stents placed in a vessel phantom and filled with contrast agent was demonstrated in [3]. Differentia-tion between two contrast agents (tantalum and iodine) by means of tantalum K-edge imaging was shown in a phan-tom experiment [44], as well as differentiation between gold nanoparticles and iodinated contrast agent in a New Zealand rabbit model [4]. In a phantom experiment on an abdominal aortic aneurysm phantom, Dangelmaier et al. [5] evaluated the separation of iodine, gadolinium, and calcium, and concluded that photon-counting CT might be able to capture endoleak dynamics and allow reliable distinction from intra-aneurysmatic calcifications in a single scan. Using a custom-made colon phantom filled with iodine and

a gadolinium-filled capsule representing a contrast-enhanced polyp, Muenzel et al. [34] demonstrated that differentiation between gadolinium-tagged polyps and iodine-tagged fecal material should be possible. Mixtures of iodine, gadolinium, and gold nanoparticles could be differentiated, as well [49]. The authors concluded that it might be possible to perform the imaging of multiple uptake phases in an organ with a single scan by injecting several contrast agents sequentially. The separation of iodine and gadolinium contrast agent for dual-phase liver imaging in a single acquisition was shown in an animal experiment with New Zealand rabbits [50]. Using the single-source prototype with CZT detector, dif-ferentiation between blood and iodine in a bovine brain was demonstrated by computing iodine maps and virtual non-contrast images [45].

A pre-clinical hybrid dual-source CT scanner equipped with a conventional scintillation detector and a CdTe pho-ton-counting detector (Siemens Healthcare GmbH, Forch-heim, Germany) was described and evaluated in Kappler et al. [20–22]. The shortest rotation time of the system is 0.5 s. The X-ray tubes can be operated at voltages up to 140 kVp. The tube current can be set to values between 25 and 550 mA. The photon-counting detector consists of sub-pixels with a size of 0.225 × 0.225 mm2. The detector provides two energy thresholds per sub-pixel. 4 × 4 sub-pixels form a “macro pixel” (see also Fig. 2b) with a size of 0.9 × 0.9 mm2, comparable to today’s medical CT systems. The sub-pixels can be binned and read-out in different ways, see Fig. 10. The low-energy “sharp pixels” and “UHR pix-els” after 2 × 2 binning have a size of 0.45 × 0.45 mm2. By assigning alternating low-energy thresholds and alternating high-energy thresholds to adjacent detector sub-pixels in a “chess pattern mode”, the detector provides four energy thresholds in “macro pixels”. Considering geometrical mag-nification, the low-energy “sharp pixels” and the “UHR pix-els” correspond to a detector pixel size of 0.25 × 0.25 mm2 at the iso-center of the scanner.

The in-plane field of view of the photon-counting detec-tor is 275 mm; the z coverage at the iso-center is 8–16 mm, depending on the read-out mode (see Fig. 10). A completion scan with the energy-integrating sub-system can be used to extend the data of the photon-counting detector to the full field of view of 500 mm.

Yu et al. [65] evaluated the imaging performance of the pre-clinical hybrid dual-source CT scanner by means of phantom and cadaver scans, assessing typical image quality parameters such as CT number accuracy, spatial resolution, noise, and CNR. The authors found that the photon-counting CT system provided clinical image quality at clinically real-istic levels of X-ray photon flux. They demonstrated that pulse pile-up had a negligible effect on image quality. Only subtle high-flux effects were noticed for tube currents higher than 300 mA in a small phantom (neonate water phantom).

26 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

Further phantom and cadaver studies [14] confirmed the potential of the photon-counting CT system to achieve clini-cal levels of image quality at clinical radiation dose levels, see Fig. 11.

As a next step, the overall performance of the proto-type photon-counting CT scanner was assessed in scans of human volunteers. Pourmorteza et al. [38] evaluated contrast-enhanced abdominal CT scans in 15 asymptomatic

volunteers, in comparison with conventional CT. At matched radiation dose, photon-counting detector images showed similar qualitative and quantitative scores for image qual-ity, image noise, and artifacts, while additionally providing spectral information for material decomposition.

Gutjahr et al. [14] evaluated the improvement of iodine CNR by photon-counting CT, which is expected as a result of the missing down-weighting of low-energy X-ray photons.

Fig. 10 Read-out modes of the photon-counting detector in a pre-clinical hybrid prototype based on a dual-source CT gantry. “Macro pixels” (4 × 4 binning of the sub-pixels both for the low-energy threshold and for the high-energy threshold), “sharp pixels” (2 × 2 binning of the sub-pixels for the low-energy threshold, 4 × 4 binning of the sub-pixels for the high-energy threshold), “UHR pixels” (2 × 2 binning of the sub-pixels both for the low-energy threshold and for

the high-energy threshold), and “chess mode pixels” (2 alternating pairs of energy thresholds provide four energy bins in macro pixels). The actual dimensions of the detector cells are indicated, at the iso-center of the scanner the pixels are smaller because of geometrical magnification.  z coverage, however,  directly refers to the iso-center of the scanner

Fig. 11 Images of a cadaver head scanned on a pre-clinical hybrid dual-source CT prototype with energy-integrating detector and photon-counting detector. Left: energy-integrating detector. Center: photon-counting detector, low-energy bin 25–65 keV. Right: photon-

counting detector; high-energy bin 65–140  keV. The high-energy images of the posterior fossa acquired with the photon-counting detector (right) show significantly less beam-hardening artifacts (dark streaks between areas of dense bone). With permission from [14]

27Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

The authors measured iodine CNR in 4 anthropomorphic phantoms, simulating 4 patient sizes, at 4 X-ray tube volt-ages. For the same X-ray photon flux, the phantom measure-ments demonstrated a mean increase in iodine CNR of 11%, 23%, 31%, and 38% for the photon-counting detector system, relative to the scintillation detector system, at 80, 100, 120, and 140 kVp, respectively. Yu et al. [65] had already found an overall improvement of the iodine CNR by about 25% in their study. These improvements in iodine CNR can poten-tially be translated into reduced radiation dose, or reduced amount of contrast agent.

Improvement of soft-tissue contrasts by photon-counting CT was demonstrated by Pourmorteza et al. [39] in a brain CT study with 21 human volunteers. Photon-counting CT images received much higher reader scores for the differen-tiation of grey and white brain matter than conventional CT images. This was due to both higher soft-tissue contrasts (10.3 ± 1.9 HU grey–white brain matter contrast for photon-counting CT versus 8.9 ± 1.8 HU for conventional CT), and lower image noise for photon-counting CT, see Fig. 12.

These findings [14, 39, 65] confirm the predicted CNR improvements with photon-counting detectors because of the better weighting of low-energy X-rays. CNR is significantly improved in contrast-enhanced CT scans (iodinated contrast agent versus soft tissue), and to a lesser degree in non-con-trast brain CT scans (grey versus white brain matter).

Several authors evaluated the imaging performance of the pre-clinical hybrid photon-counting CT prototype at low radiation dose, to assess the impact of the potentially miss-ing or strongly reduced electronic noise on image quality in different clinical applications.

Using an anthropomorphic thorax phantom, Yu et al. [66] demonstrated both qualitatively and quantitatively that

electronic noise had a noticeably stronger degrading impact in shoulder images acquired with the conventional energy-integrating detector than in images acquired with the photon-counting detector, see Fig. 13.

Symons et  al. [51] evaluated the performance of the photon-counting CT prototype for potential low-dose lung cancer screening. Scanning a lung phantom at low radiation dose, the authors found a better Hounsfield unit stability for lung, ground-glass, and emphysema-equivalent foams for the photon-counting detector than for the energy-integrating detector, with a better reproducibility. Stability of Houns-field units is an important pre-requisite for quantitative CT. Additionally, photon-counting CT showed up to 10% less noise, and 11% higher CNR at a CTDIvol of 0.75 mGy, cor-responding to a dose-length product for an average length thorax (30 cm) of 22.5 mGy cm. The better performance of photon-counting CT at very low radiation dose was attrib-uted to the effective elimination of electronic noise and bet-ter weighting of low-energy X-ray photons. In a study with 30 human subjects [52], the authors evaluated whether pho-ton-counting detectors can improve dose-reduced chest CT image quality in vivo. Compared to energy-integrating CT, photon-counting CT demonstrated higher diagnostic quality with significantly better image quality scores for lung, soft tissue, and bone and with fewer beam-hardening artifacts, lower image noise, and higher CNR for lung nodule detec-tion (see Fig. 14).

Coronary artery calcium (CAC) scoring is another CT application that demands low radiation dose to the patient, in particular when performed in a screening context. Symons et al. [57] hypothesized that the absence of electronic noise in combination with improved calcium contrast due to missing down-weighting of low-energy X-ray photons may

Fig. 12 Brain image of a 59-year-old woman, acquired with the standard energy-integrating scintillation detector (left) and the pho-ton-counting detector (right) of a pre-clinical hybrid dual-source CT prototype. The same tube voltage (kVp) and tube current–time prod-

uct (mAs) settings were used to obtain the same X-ray photon flux for both images. Window width 80 HU, window center 45 HU. The photon-counting CT images show higher grey–white brain matter contrast. With permission from [39]

28 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

improve the quality of CAC scoring at low radiation dose. In a combined phantom, ex vivo and in vivo study, the authors evaluated the potential of photon-counting CT for stand-ard and reduced-dose CAC scoring. Agreement between standard-dose (average CTDIvol = 5.4 mGy) and low-dose (average CTDIvol = 1.6 mGy) CAC score in ten volunteers was significantly better for photon-counting CT than for conventional CT. The authors found higher calcium-soft-tissue contrasts for photon-counting CAC scans with better low-dose CAC score reproducibility and a smaller number of false positive voxels above the Agatston threshold of 130

HU. They concluded that photon-counting CT technology may play a role in further reducing the radiation dose of CAC scoring.

The photon-counting detector of the pre-clinical hybrid dual-source CT scanner is expected to provide increased spa-tial resolution because of its smaller effective detector pixel size in “sharp mode” and in “UHR mode” (see Fig. 10). Several studies were performed to evaluate spatial resolu-tion both in phantom scans and in scans of human subjects.

Leng et al. [26] measured a cut-off spatial frequency of 32.4 lp/cm in “UHR mode”, corresponding to 150 μm

Fig. 13 Image of a shoulder phantom at various dose levels (120 kVp with 20, 30, and 60 mAs, respectively), acquired with the standard energy-integrating scintillation detector (left) and the photon-counting detector (right) of a pre-clinical hybrid dual-source CT proto-type. The photon-counting CT images have less streak artifacts and more homogeneous image noise because of the absence of electronic noise. The improve-ment is the more obvious the lower the radiation dose is. With permission from [65]

29Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

in-plane spatial resolution. “UHR mode” and “sharp mode” provided similar values of the modulation transfer func-tion (MTF) in the low-threshold images due to the identical detector pixel sizes in these modes. The measured minimum slice widths (full width at half maximum) in the “UHR”, “sharp”, and “macro mode” were 0.41 mm, 0.44 mm, and 0.67 mm, respectively. At matched in-plane resolution, “UHR mode” and “sharp mode” images had up to 15% lower image noise than “macro mode” images because of the better MTF of the measurement system. The authors demonstrated improved spatial resolution in clinical images of the lung, shoulder, and temporal bone, see Fig. 15.

Pourmorteza et al. [40] assessed the clinical feasibil-ity, image quality, and radiation dose implications of the “sharp mode” and “UHR mode” in a cohort of eight humans who underwent scans of the brain, the thorax, and at the level of the upper left kidney. The authors observed

improved image quality in terms of spatial resolution and image noise compared with standard-resolution photon-counting CT. Substantially better delineation of temporal bone anatomy scanned with the “UHR mode” compared with the ultra-high-resolution mode of a commercial energy-integrating-detector CT scanner is shown in [69].

Coronary CTA is another clinical application poten-tially benefitting from increased spatial resolution. Fig-ure 16 shows a coronary stent scanned with the photon-counting CT prototype in the “sharp” and in the “macro” acquisition mode. The high-resolution capabilities of photon-counting CT for coronary stent imaging were confirmed in several studies. In a phantom experiment, Symons et al. [55] demonstrated significantly improved coronary stent lumen visibility with the “UHR mode”. Von Spiczak et al. [62] found superior qualitative and quantita-tive image characteristics for photon-counting coronary

Fig. 14 Example of a low-dose lung scan acquired with the pre-clinical hybrid dual-source CT prototype. Left: energy-integrating detector images. Right: photon-counting detector images, demonstrating less image noise (arrowheads) and improved CNR. Courtesy of R Symons, NIH, Bethesda, USA

Fig. 15 Example of a shoulder scan acquired with the pre-clinical hybrid dual-source CT prototype. Left: energy-inte-grating detector image. Right: photon-counting detector image, “sharp mode”, demonstrating higher spatial resolution and significantly improved visuali-zation of bony structures. With permission from [26]

30 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

stent imaging when using a dedicated sharp convolution kernel.

In an in vivo study with 22 adult patients referred for clinically indicated high-resolution chest CT, Bartlett et al. [2] demonstrated superior visualization of higher order bron-chi and third-/fourth-order bronchial walls at preserved lung nodule conspicuity compared with clinical reference images when “sharp mode” photon-counting CT was combined with image reconstruction at 1024 × 1024 matrix size using a ded-icated sharp convolution kernel (see Fig. 17).

According to the authors, the results of their study dem-onstrate the benefits of photon-counting CT for high-reso-lution imaging of airway diseases, which may potentially be extended to other pathologies, such as fibrosis, honeycomb-ing, and emphysema.

Improvement of lung nodule volume quantification with the “UHR mode” was shown in a phantom experiment with 20 synthetic lung nodules [68]. The overall volume estima-tion was more accurate using “UHR mode” and a very sharp kernel, having 4.8% error compared with 10.5–12.6% error in the “macro mode”. The improvement in volume meas-urements was more evident for small nodule sizes or star-shaped nodules.

Practically, spatial resolution is not only determined by the detector pixel size, but also by the focal spot of the X-ray tube which needs to be correspondingly small. The smaller the focal spot is, the less tube power is usually available, which may limit the clinical applicability of ultra-high-res-olution CT scanning. Furthermore, increasing the resolution of a CT scan comes at the expense of significantly increased

Fig. 16 Image of a coronary stent (left) scanned with the pre-clinical hybrid dual-source CT prototype with photon-counting detector. The “macro mode” (center, see also Fig. 10) cor-responds to the resolution level of today’s medical CT systems. The “sharp mode” (right, see also Fig. 10) provides a cut-off spatial resolution of 32 lp/cm [26]. Images courtesy of Clini-cal Innovation Center, Mayo Clinic Rochester, MN, USA

Fig. 17 Lung images of a 68-year-old man acquired with the pre-clinical hybrid dual-source CT prototype. Left: clinical reference image, medium sharp convolution kernel, 512 × 512 image matrix. Right: photon-counting detector image, “sharp mode”, sharp convolu-

tion kernel (10% value of the MTF 18.9 lp/cm) beyond the resolution limits of the energy-integrating detector, 1024 × 1024  image matrix. Improved visualization of honeycombing and fibrosis. With permis-sion from [2]

31Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

image noise at a given radiation dose level which may not be tolerable in all cases and require increased radiation dose to the patient to compensate for the higher noise. Non-linear data and image denoising techniques will play a key role in harnessing the high-resolution potential of photon-counting detectors in clinical practice (see, e.g., [16, 28, 41]).

A key benefit of photon-counting CT is the availability of spectral CT data in any scan. Spectral information can readily be added to the anatomical images for a better visu-alization of structures, for material classification and quan-titation, and to obtain quantitative information about local perfusion by means of iodine maps.

The spectral performance of the pre-clinical hybrid dual-source prototype with photon-counting detector was evalu-ated in phantom studies [25]. Using five phantoms with lat-eral widths of 25, 30, 35, 40, and 45 cm to represent slim to obese adult patients, the authors assessed the accuracy of iodine quantification in iodine maps and that of CT number accuracy in virtual monoenergetic images (VMIs) and found it to be comparable to that of clinical dual-source, dual-energy CT. According to the authors, photon-counting CT provides high accuracy for iodine quantification and accurate CT numbers in VMIs while offering other advantages such as perfect temporal and spatial alignment to avoid motion artifacts, high spatial resolution, and improved CNR.

Symons et al. [56] evaluated the performance of photon-counting CT for the examination of major arteries of the head and neck in 16 asymptomatic volunteers. The authors found significantly higher image quality scores than with the conventional energy-integrating detector CT, at lower

image noise and less artifacts. Furthermore, they assessed the use of iodine maps and VMIs in head and neck CTA. In an anthropomorphic head phantom containing tubes filled with aqueous solutions of iodine (0.1–50 mg/ml) excellent agreement between actual iodine concentrations and iodine concentrations measured in the iodine maps was observed. VMIs were proposed as a method to enhance plaque detec-tion and characterization as well as grading of stenosis by reconstructing images at different keV (see Fig. 18).

The routine availability of VMIs may pave the way to further standardization of CT protocols, in particular if CNR and image quality of the VMIs are enhanced by refined pro-cessing (see, e.g., [13]). With photon-counting CT, VMIs at pre-defined keV levels depending on the clinical ques-tion (e.g., 50–60 keV for contrast-enhanced examinations of parenchymal organs; 40–50 keV for CT angiographic studies) may serve as a standard output of any CT imag-ing procedure regardless of the acquisition protocol. Some authors [70] even recommend a standardized acquisition protocol with 140 kVp X-ray tube voltage for contrast-enhanced abdominal CT examinations in all patient sizes, with standardized VMI reconstruction at 50 keV. According to the authors, optimal or near optimal iodine CNR for all patient sizes is obtained with this protocol.

Several authors assessed the performance of spectral photon-counting CT for detection and characterization of kidney stones, which is another established dual-energy CT application [9, 15, 30]. They found comparable overall per-formance to state-of-the art energy-integrating detector CT in differentiating stone composition, while photon-counting

Fig. 18 Contrast-enhanced images of the internal carotid artery of a 73-year-old woman acquired with the pre-clinical hybrid dual-source CT prototype. All images are displayed with center 145 HU, width 800 HU. The high-energy bin image demonstrates less blooming of the calcified plaque than the low-energy bin image. Virtual monoen-

ergetic images at different keV (shown at the bottom) can be used to optimize the contrast between plaque and iodine-enhanced lumen to improve plaque characterization and grading of stenosis. With per-mission  from [56]

32 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

CT was better able to help characterize small renal stones [30]. Because of its higher spatial resolution, photon-count-ing CT can provide both, a high-resolution image of the stone structure and a material-map image of the stone com-position (see Fig. 19).

In addition to evaluating the performance of established dual-energy applications with photon-counting CT, multi-material decomposition enabled by data acquisition with more than two energy bins has been studied. Symons et al. [53] determined the feasibility of dual-contrast agent imag-ing of the heart to simultaneously assess both first-pass and late enhancement of the myocardium with the pre-clinical hybrid dual-source CT prototype. In a canine model of myo-cardial infarction, gadolinium was injected 10 min prior to CT, while iodinated contrast agent was given immediately before the CT scan. The authors concluded that combined first-pass iodine and late gadolinium maps allowed quan-titative separation of blood pool, infarct scar, and remote myocardium. The same authors also investigated the fea-sibility of simultaneous material decomposition of three contrast agents in vivo in a large animal model [54]. The authors could decompose bismuth, iodine, and gadolinium in a canine model and derive quantitative maps of the differ-ent contrast agent concentrations. In addition, they observed tissue enhancement at multiple phases in a single CT acqui-sition, opening the potential to replace multiphase CT scans by a single CT acquisition with multiple contrast agents (see Fig. 20).

In clinical practice, the use of multi-material maps may be hampered by the unavoidable increase of image noise

Fig. 19 Example of an abdominal CT scan of a volunteer acquired with the pre-clinical hybrid dual-source CT prototype in “sharp mode”. The high-resolution low-threshold images at 0.25-mm nomi-nal slice thickness show the internal structure of two kidney stones (top left), while a material map reveals their composition (top right). A volume rendered image puts both stones into their anatomical con-text (bottom: note the different color scheme for uric acid and cal-cium). Courtesy of S. Leng, Mayo Clinic Rochester, MN, USA

Fig. 20 Example for simultaneous imaging of multiple contrast agents by multi-material decomposition in a dog model. Scan data were acquired with the pre-clinical hybrid photon-counting CT pro-totype based on a dual-source gantry. Bismuth subsalicylate was administered more than 1 day prior to scanning. The scans were per-formed after both intravenous administration of gadolinium-based contrast agent and iodine-based contrast agent. The iodine-based con-trast agent was injected 3–4 min after the gadolinium-based contrast

agent to visualize different phases of renal enhancement at the same time point in a single CT acquisition. Scan data were read-out in four energy bins (25–50 keV, 50–75 keV, 75–90 keV, and 90–140 keV). Left: grayscale image using all detected X-ray photons by combin-ing the data of the four energy bins. The three contrast agents cannot be differentiated. Right: grayscale image with overlay of the colored material maps. Iodine is indicated in red, gadolinium in green, and bismuth in blue. With permission  from [54]

33Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

in a multi-material decomposition. Similar to ultra-high-resolution scanning, non-linear data and image denoising techniques will play a key role to fully exploit the potential of multi-material decomposition in clinical routine (see, e.g., [58]).

In this review article, we have outlined the basic princi-ples of photon-counting CT and its potential clinical appli-cations. Once remaining challenges of this technology have been mastered, photon-counting CT has the potential to bring clinical CT to a new level of performance.

References

1. Albrecht MH, De Cecco CN, Schoepf UJ, et al. Dual-energy CT of the heart current and future status. Eur J Radiol. 2018;105:110–8.

2. Bartlett DJ, Koo WC, Bartholmai BJ, et al. High-resolution chest computed tomography imaging of the lungs: impact of 1024 matrix reconstruction and photon-counting detector computed tomography. Invest Radiol. 2019;54(3):129–37.

3. Bratke G, Hickethier T, Bar-Ness D, et al. Spectral photon-count-ing computed tomography for coronary stent imaging: evalua-tion of the potential clinical impact for the delineation of in-stent restenosis. Invest Radiol. 2019. https ://doi.org/10.1097/rli.00000 00000 00061 0 (epub ahead of print).

4. Cormode DP, Si-Mohamed S, Bar-Ness D, et al. Multicolor spec-tral photon-counting computed tomography: in vivo dual contrast imaging with a high count rate scanner. Sci Rep. 2017;7(1):4784.

5. Dangelmaier J, Bar-Ness D, Daerr H, et al. Experimental feasibil-ity of spectral photon-counting computed tomography with two contrast agents for the detection of endoleaks following endovas-cular aortic repair. Eur Radiol. 2018;28(8):3318–25.

6. De Santis D, Eid M, De Cecco CN, et al. Dual-energy computed tomography in cardiothoracic vascular imaging. Radiol Clin N Am. 2018;56(4):521–34.

7. Duan X, Wang J, Leng S, et al. Electronic noise in CT detec-tors: impact on image noise and artifacts. AJR Am J Roentgenol. 2013;201(4):W626–32.

8. Feuerlein S, Roessl E, Proksa R, et al. Multienergy photon-count-ing K-edge imaging: potential for improved luminal depiction in vascular imaging. Radiology. 2008;249(3):1010–6.

9. Ferrero A, Gutjahr R, Halaweish AF, Leng S, McCollough CH. Characterization of urinary stone composition by use of whole-body, photon-counting detector CT. Acad Radiol. 2018;25(10):1270–6.

10. Flohr TG, McCollough CH, Bruder H, Petersilka M, Gruber K, Süß C, et al. First performance evaluation of a dual-source CT (DSCT) system. Eur Radiol. 2006;16:256–68.

11. Flohr TG, Stierstorfer K, Süss C, Schmidt B, Primak AN, McCollough CH. Novel ultrahigh resolution data acquisition and image reconstruction for multi-detector row CT. Med Phys. 2007;34(5):1712–23.

12. Gassenmaier T, Petri N, Allmendinger T, et al. Next generation coronary CT angiography: in vitro evaluation of 27 coronary stents. Eur Radiol. 2014;24(11):2953–61.

13. Grant KL, Flohr TG, Krauss B, et al. Assessment of an advanced image-based technique to calculate virtual monoenergetic com-puted tomographic images from a dual-energy examination to improve contrast-to-noise ratio in examinations using iodinated contrast media. Invest Radiol. 2014;49(9):586–92.

14. Gutjahr R, Halaweish AF, 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(7):421–9.

15. Gutjahr R, Polster C, Henning A, Kappler S, Leng S, McCol-lough CH, Sedlmair MU, Schmidt B, Krauss B, Flohr TG. Dual energy CT kidney stone differentiation in photon counting com-puted tomography. In: Proc SPIE int soc opt eng, vol 10132. 2017.

16. Harrison AP, Xu Z, Pourmorteza A, Bluemke DA, Mollura DJ. A multichannel block-matching denoising algorithm for spectral photon-counting CT images. Med Phys. 2017;44(6):2447–52.

17. Johnson TRC, Krauß B, Sedlmair M, et al. Material differ-entiation by dual energy CT: initial experience. Eur Radiol. 2007;17(6):1510–7.

18. Kalender W, Seissler W, Klotz E, Vock P. Spiral volumetric CT with single-breath-hold technique, continuous transport and continuous scanner rotation. Radiology. 1990;176:181–3.

19. Kappler S, Niederlöhner D, Stierstorfer K, Flohr T. Contrast-enhancement, image noise and dual-energy simulations for quantum-counting clinical CT. Proc SPIE Med Imaging Conf. 2010;7622:76223H.

20. Kappler S, Hannemann T, Kraft E, et al. First results from a hybrid prototype CT scanner for exploring benefits of quantum-counting in clinical CT. In: Medical imaging 2012: physics of medical imaging. San Diego: International Society for Optics and Photonics; 2012. p. 83130X.

21. Kappler S, Henning A, Krauss B, et al. Multi-energy perfor-mance of a research prototype CT scanner with small-pixel counting detector. In: Medical imaging 2013: physics of medi-cal imaging. Lake Buena Vista: International Society for Optics and Photonics; 2013. p. 86680O.

22. Kappler S, Henning A, Kreisler B, et al. Photon counting CT at elevated X-ray tube currents: contrast stability, image noise and multi-energy performance. In: Medical imaging 2014: physics of medical imaging. San Diego: International Society for Optics and Photonics; 2014. p. 90331C.

23. Kopp FA, Daerr H, Si-Mohamed S, et al. Evaluation of a pre-clinical photon-counting CT prototype for pulmonary imaging. Sci Rep. 2018;8(1):17386.

24. Klingenbeck-Regn K, Schaller S, Flohr T, Ohnesorge B, Kopp AF, Baum U. Subsecond multi-slice computed tomography: basics and applications. Eur J Radiol. 1999;31:110–24.

25. Leng S, Zhou W, Yu Z, et al. Spectral performance of a whole-body research photon counting detector CT: quantitative accuracy in derived image sets. Phys Med Biol. 2017;62(17):7216–32.

26. Leng S, Rajendran K, Gong H, et al. 150-μm spatial resolution using photon-counting detector computed tomography technol-ogy: technical performance and first patient images. Invest Radiol. 2018;53(11):655–62.

27. Leng S, Bruesewitz M, Tao S, et al. Photon-counting detector CT: system design and clinical applications of an emerging technol-ogy. Radiographics. 2019;39(3):729–43.

28. Li Z, Leng S, Yu L, et al. An effective noise reduction method for multi-energy CT images that exploit spatio-spectral features. Med Phys. 2017;44(5):1610–23.

29. Lu GM, Zhao Y, Zhang LJ, Schoepf UJ. Dual-energy CT of the lung. AJR Am J Roentgenol. 2012;199(5 Suppl):S40–53.

30. Marcus RP, Fletcher JG, Ferrero A, et al. Detection and char-acterization of renal stones by using photon-counting-based CT. Radiology. 2018;289(2):436–42.

31. Marin D, Boll DT, Mileto A, Nelson RC. State of the art: dual-energy CT of the abdomen. Radiology. 2014;271(2):327–42.

32. Martinez FJ, Chisholm A, Collard HR. The diagnosis of idiopathic pulmonary fibrosis: current and future approaches. Lancet Respir Med. 2017;5(1):61–71.

34 Chinese Journal of Academic Radiology (2020) 3:19–34

1 3

33. McCollough CH, Zink FE. Performance evaluation of a multi-slice CT system. Med Phys. 1999;26:2223–30.

34. Muenzel D, Bar-Ness D, Roessl E, et al. Spectral photon-counting CT: initial experience with dual-contrast agent K-edge colonogra-phy. Radiology. 2017;283(3):723–8.

35. Mori S, Obata T, Nakajima N, Ichihara N, Endo M. Volumetric perfusion CT using prototype 256-detector row CT scanner: prelimi-nary study with healthy porcine model. AJNR Am J Neuroradiol. 2005;26(10):2536–41.

36. Odisio EG, Truong MT, Duran C, de Groot PM, Godoy MC. Role of dual-energy computed tomography in thoracic oncology. Radiol Clin N Am. 2018;56(4):535–48.

37. Onishi H, Hori M, Ota T, Nakamoto A, Osuga K, Tatsumi M, Fukui H, Tsukagoshi S, Uranishi A, Saito Y, Taniguchi A, Enchi Y, Sato K, Tomiyama N. Phantom study of in-stent restenosis at high-spatial-resolution CT. Radiology. 2018;289(1):255–60.

38. Pourmorteza A, Symons R, Sandfort V, et al. Abdominal imaging with contrast-enhanced photon-counting CT: first human experi-ence. Radiology. 2016;279(1):239–45.

39. Pourmorteza A, Symons R, Reich DS, Bagheri M, Cork TE, Kappler S, Ulzheimer S, Bluemke DA. Photon-counting CT of the brain: in vivo human results and image-quality assessment. AJNR Am J Neuroradiol. 2017;38(12):2257–63.

40. Pourmorteza A, Symons R, Henning A, Ulzheimer S, Bluemke DA. Dose efficiency of quarter-millimeter photon-counting computed tomography: first-in-human results. Invest Radiol. 2018;53(6):365–72.

41. Rajendran K, Tao S, Abdurakhimova D, Leng S, McCollough C. Ultra-high resolution photon-counting detector CT reconstruc-tion using spectral prior image constrained compressed-sensing (UHR-SPICCS). Proc SPIE Int Soc Opt Eng. 2018. https ://doi.org/10.1117/12.22946 28.

42. Rajiah P, Sundaram M, Subhas N. Dual-energy CT in musculoskel-etal imaging: what is the role beyond gout? AJR Am J Roentgenol. 2019;213(3):493–505.

43. Rassouli N, Etesami M, Dhanantwari A, Rajiah P. Detector-based spectral CT with a novel dual-layer technology: principles and appli-cations. Insights Imaging. 2017;8(6):589–98.

44. Riederer I, Bar-Ness D, Kimm MA. Liquid embolics agents in spec-tral X-ray photon-counting computed tomography using tantalum K-edge imaging. Sci Rep. 2019;9:5268.

45. Riederer I, Si-Mohamed S, Ehn S, et al. Differentiation between blood and iodine in a bovine brain—initial experience with spec-tral photon-counting computed tomography (SPCCT). PLoS One. 2019;14(2):e0212679.

46. Schlomka JP, Roessl E, Dorscheid R, et al. Experimental feasibil-ity of multi-energy photon-counting K-edge imaging in pre-clinical computed tomography. Phys Med Biol. 2008;53(15):4031–47.

47. Shanbhag SM, Schuzer JL, Steveson C, Rollison S, Bronson KC, Stagliano MS, Rogalla P, Blum A, Prokop M, Chen MY. Prototype ultrahigh-resolution computed tomography for chest imaging: initial human experience. J Comput Assist Tomogr. 2019;43(5):805–10.

48. Siegel MJ, Ramirez-Giraldo JC. Dual-energy CT in chil-dren: imaging algorithms and clinical applications. Radiology. 2019;291(2):286–97.

49. Si-Mohamed S, Bar-Ness D, Sigovan M, et al. Multicolour imaging with spectral photon-counting CT: a phantom study. Eur Rad Exp. 2018;2:34.

50. Si-Mohamed S, Tatard-Leitman V, Laugerette A, et al. Spectral photon-counting computed tomography (SPCCT): in vivo single-acquisition multi-phase liver imaging with a dual contrast agent protocol. Sci Rep. 2019;9(1):8458.

51. Symons R, Cork T, Sahbaee P, et al. Low-dose lung cancer screen-ing with photon-counting CT: a feasibility study. Phys Med Biol. 2017;62(1):202–13.

52. Symons R, Pourmorteza A, Sandfort V, et al. Feasibility of dose-reduced chest ct with photon-counting detectors: initial results in humans. Radiology. 2017;285(3):980–9.

53. Symons R, Cork TE, Lakshmanan MN, et al. Dual-contrast agent photon-counting computed tomography of the heart: initial experi-ence. Int J Cardiovasc Imaging. 2017;33:1253–61.

54. Symons R, Krauss B, Sahbaee P, et al. Photon-counting CT for simultaneous imaging of multiple contrast agents in the abdomen: an in vivo study. Med Phys. 2017;44(10):5120–7.

55. Symons R, de Bruecker Y, Roosen J, et al. Quarter-millimeter spec-tral coronary stent imaging with photon-counting CT: initial experi-ence. J Cardiovasc Comput Tomogr. 2018;12:509–15.

56. Symons R, Reich DS, Bagheri M, et al. Photon-counting computed tomography for vascular imaging of the head and neck: first in vivo human results. Invest Radiol. 2018;53(3):135–42.

57. Symons R, Sandfort V, Mallek M, Ulzheimer S, Pourmorteza A. Coronary artery calcium scoring with photon-counting CT: first in vivo human experience. Int J Cardiovasc Imaging. 2019;35(4):733–9.

58. Tao S, Rajendran K, McCollough CH, Leng S. Material decomposi-tion with prior knowledge aware iterative denoising (MD-PKAID). Phys Med Biol. 2018;63(19):195003.

59. Taguchi K, Iwanczyk JS. Vision 20/20: single photon counting X-ray detectors in medical imaging. Med Phys. 2013;40(10):100901.

60. Taguchi K. Energy-sensitive photon counting detector-based X-ray computed tomography. Radiol Phys Technol. 2017;10(1):8–22.

61. Vanhecke TE, Madder RD, Weber JE, et al. Development and valida-tion of a predictive screening tool for uninterpretable coronary CT angiography results. Circ Cardiovasc Imaging. 2011;4(5):490–7.

62. von Spiczak J, Mannil M, Peters B, et al. Photon counting computed Tomography with dedicated sharp convolution kernels—tapping the potential of a new technology for stent imaging. Invest Radiol. 2018;53(8):486–94.

63. Willemink MJ, Persson M, Pourmorteza A, Pelc NJ, Fleischmann D. Photon-counting CT: technical principles and clinical prospects. Radiology. 2018;289(2):293–312.

64. Yanagawa M, Hata A, Honda O, et al. Subjective and objective com-parisons of image quality between ultra-high-resolution CT and con-ventional area detector CT in phantoms and cadaveric human lungs. Eur Radiol. 2018;28(12):5060–8.

65. Yu Z, Leng S, Jorgensen SM, et al. Evaluation of conventional imag-ing performance in a research CT system with a photon-counting detector array. Phys Med Biol. 2016;61:1572–95.

66. Yu Z, Leng S, Kappler S, et al. Noise performance of low-dose CT-comparison between an energy integrating detector and a photon counting detector using a whole-body research photon counting CT scanner. J Med Imaging. 2016;3(4):043503.

67. Zhang D, Li X, Liu B. Objective characterization of GE discovery CT750 HD scanner: gemstone spectral imaging mode. Med Phys. 2011;38(3):1178–88.

68. Zhou W, Montoya J, Gutjahr R, et al. Lung nodule volume quan-tification and shape differentiation with an ultra-high resolution technique on a photon-counting detector computed tomography system. J Med Imaging (Bellingham). 2017;4(4):043502. https ://doi.org/10.1117/1.jmi.4.4.04350 2 (epub 2017 Nov 16).

69. Zhou W, Lane JI, Carlson ML, et al. Comparison of a photon-counting-detector CT with an energy-integrating-detector CT for temporal bone imaging: a cadaveric study. AJNR Am J Neuroradiol. 2018;39(9):1733–8.

70. Zhou W, Abdurakhimova D, Bruesewitz M, et al. Impact of pho-ton counting detector technology on kV selection and diagnostic workflow in CT. Proc SPIE Int Soc Opt Eng. 2018. https ://doi.org/10.1117/12.22949 52.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.