Key words photon-counting - computed tomography - diagnostic imaging - photon-counting detector
- spectral computed tomography - energy-integrating detector
Introduction
Computed tomography (CT) remains one of the most commonly used imaging modalities
in the examination of oncological, pneumological, skeletal, and cardiovascular diseases,
and is also specifically used in certain pediatric clinical pictures. However, a major
limitation of CT imaging remains the associated radiation exposure, which particularly
affects the following three subgroups: first, oncological patients who are repeatedly
referred for CT follow-up; second, cardiovascular patients at a younger age who receive
coronary CT angiography to screen for coronary heart disease, especially using retrospective
gating; and third, pediatric patients who are more sensitive to the effects of ionizing
radiation as their bodies undergo rapid cell division [1 ]
[2 ]
[3 ]. With the introduction of the photon counting detector (PCD) CT, it should now be
possible to significantly reduce the dose while maintaining comparable or even improved
image quality. [4 ]
[5 ] Other significant technical advantages over conventional energy-integrating detectors
(EID) are the improved spatial resolution and the ability to record the energy of
individual photons, thus performing a CT by means of a “multi-energy” technique. These
technical advantages result directly from the new detector technology. A PCD is based
on a semiconductor material (usually cadmium telluride) that directly converts each
incoming X-ray photon into an electrical signal, whereby the level of the current
pulse generated by the X-ray photon is proportional to the energy of the same [6 ]. Therefore, the PCD is also referred to as a direct-converting detector. This bypasses
the intermediate step of conversion required for EID, in which X-ray photons must
first be converted into visible light in a scintillator and then into an electrical
signal in a photodiode [6 ]. Therefore, the otherwise necessary reflective “septa” between the individual detector
cells are also obsolete, since no light photons need to be bundled onto a detector
cell. This achieves a higher resolution in the PCD CT because, on the one hand, a
higher cell density on the PCD is possible and, on the other hand, the energy proportion
that would otherwise reach the septa and thus lack information is lost [6 ]
[7 ]. In addition, a flatter design of the photon counting detector is also possible.
Another advantage of photon counting is the elimination of electron noise, as only
integers of photons are read. The resulting specific advantages are discussed below
in the subsections on oncology, cardiovascular radiology, and pediatric radiology.
Oncological imaging
Head and neck area
The challenge with CT of the head/neck region is to achieve an optimal balance between
the applied radiation dose (to minimize radiation exposure of the radiation-sensitive
organs in the midface) and the detailed image quality to be achieved. [8 ] Fine structures, such as the paranasal sinuses or the skull base, in particular,
benefit from the higher image resolution of the PCD CT, as this enables better visualization
of both traumatic complications of the inner/middle ear and subtle bone infiltration
caused by inflammation or tumors (see [Fig. 1 ]) [8 ]. This high image resolution is also crucial for tumors of the larynx and (nasal)
pharynx, which are sometimes difficult to visualize. In particular, the assessment
of whether there is infiltration of the laryngeal cartilage, tongue bone, or even
the skull base is crucial here, and is made easier by the higher spatial resolution
of the PCD CT [9 ]. Also, the combination of the higher resolution of the PCD CT with already validated
bone subtraction algorithms allows for the assessment of the possible involvement
of the skull base [10 ]. In addition, the potential connection to the cervical vessels, which are superimposed,
for example, by high-contrast artifacts in the form of dental implants, can be minimized.
As with the Dual-Energy EID CT, the calculation of virtually monoenergetic (VMI) image
datasets allows either a high tissue contrast due to an increased iodine signal (low
keV) or the reduction of beam hardening artifacts (high keV) [11 ]. Furthermore, the PCD CT compared to the EID CT allows a higher tissue contrast
between gray and white matter, which, in combination with the reduced image noise,
can be crucial, especially in the diagnosis of acute stroke (1–3 hours) [12 ].
Fig. 1 76-year-old patient, preoperative status prior to implantation of a cochlear implant.
a Ultra-high-resolution single-source examination of the petrous bone (Siemens SOMATOM
Definition AS + 120 kV); b Ultra-high-resolution PCD-CT scan (Siemens Naeotom Alpha 120 kV) of the petrous bone.
In comparison, the lower noise of the PCD-CT images and also the higher resolution
of the trabecular structures are particularly striking. Instead of a slightly blurred
spongiosa architecture, highly detailed contours, e.g., in the middle ear, are seen.
Window width: 3514 HU, center of the window: 927 HU.
Chest and breast
The spectral datasets generated by the PCD CT can be divided into different energy
areas by defining different energy threshold values. Through this subdivision, the
photons that reach the detector can be sorted into so-called “bins” depending on the
energy level [13 ]. In chest diagnostics, this bundling of low-energy photons after contrast media
administration allows high tissue contrast, both in the breast parenchyma [14 ]and in the pulmonary parenchyma [15 ]. This enables even small structures, such as pulmonary nodules, to be more clearly
displayed in the marginal area with the high-resolution PCD CT, and, in terms of their
size over time, to be more comparable [16 ]
[17 ]. The overall lower noise level as well as the higher spatial resolution of the native
PCD CT also lead to improved detailed visualization of peripheral lung structures,
such as thickened bronchial walls (see [Fig. 2 ]), ground-glass-like changes, and the “tree-in-bud”/mosaic pattern, which are all
crucial for the diagnosis of interstitial lung disease [18 ]. Milos et al. show that using a suitable reconstruction kernel (Bl64), bronchial
branches in the peripheral lung regions, the lobular fissures, and also the bronchial
walls can be better visualized in ultra-high-resolution (UHR) PCD CT without affecting
the vascular noise and the detail acuity of pulmonary round foci [19 ]. To further improve pulmonary image quality, it is possible to generate virtual
monoenergetic (VMI) image datasets. This, for example, makes pulmonary emphysema best
visible between 60 and 70 keV [20 ]. In addition, the vascular contrast can be significantly increased at less than
60 keV in the contrast-enhanced chest PCD CT, but this carries the disadvantage of
higher noise levels [20 ]
[21 ]. The generation of VMI image datasets also significantly reduces beam hardening
and metal artifacts at 90 keV [22 ]and thus allows a better assessment of tumor manifestations and metastases near the
artifacts. In addition, it is possible to reduce the radiation dose by 66 % compared
to the EID CT without compromising image quality and diagnostic certainty, which is
important, for example, in the assessment of interstitial lung disease (ILD) [23 ]
[24 ]. Another advantage is the ability to create virtual contrast media-free images (VNC),
which allows better quantification of pulmonary emphysema, as the higher parenchymal
density in contrast-enhanced CT can lead to underestimation of the emphysema [25 ].
Fig. 2 65-year-old patient with chronic bronchitis. a EID-CT (Siemens SOMATOM Definition AS+ 120 kV), b PCD-CT (Siemens Naeotom Alpha 120 kV). In particular, the bronchial walls are displayed
significantly more clearly (long arrow). The branching of the peripheral vessels directly
at the pleura (see arrowhead) is more detailed on PCD-CT than EID-CT. Window width:
1500 HU, center of the window: -500 HU, 1-mm slice thickness.
Imaging of the female breast also significantly benefits from PCD CT. Due to the higher
spatial resolution, it is possible to image focal findings [14 ] and vascular structures in good quality (see [Fig. 3 ]). In addition, by post-processing the image data, iodine maps could be reconstructed,
which allow conclusions to be drawn about contrast media uptake within the mammary
gland tissue. In this way, breast carcinomas can be better recognized, assessed, and
characterized in terms of their extent, similar to what has already been reported
for the Dual-Energy EID CT [14 ]
[26 ]
[27 ]. Finally, no compression of the breast is required compared to mammography and,
accordingly, a significantly higher level of patient comfort is provided [28 ]. Even though mammography is still recognized as the standard diagnosis in breast
imaging, PCD CT still offers promising new possibilities (e. g., in staging examinations)
that should be evaluated in further clinical trials.
Fig. 3 Initial diagnosis of a mass suspicious for malignancy in the upper inner and upper
outer quadrants of the right breast (arrows a-c ). MIP of an MRI scan showing the suspicious mass (a ). Corresponding 5-mm MIP of the PCD-CT scan (b ). Note the fine vascular structures that are visible both on the MRI scan and the
PCD-CT scan of the upper inner quadrant of the left breast (arrowhead).
Abdomen and pelvis
Abdominopelvic imaging is largely influenced by the higher fat percentage in this
area of the body, which is often constitution-related, as this requires an increased
radiation dose. The PCD scanner improves image quality while reducing image noise
and photon starvation artifacts, especially in obese patients. Several studies have
reported that with increasing BMI, the signal-to-noise ratio of various abdominal
structures decreases less strongly in PCD CT, and noise increases less strongly than
in EID CT (see [Fig. 4 ]) [4 ]
[29 ]
[30 ]. Further potential improvement with regard to image quality lies in the various
iterative reconstruction stages (quantum iterative reconstruction, QIR), which can
be selected in stages 1 to 4 and achieve increasing noise suppression with increasing
stages, used for optimal detection of liver lesions [31 ].
Fig. 4 69-year-old patient with a BMI of 43. a Single-source EID-CT with low contrast and reduced visualization of the mesenteric
veins (white arrows). b PCD-CT with high contrast and improved visualization of the mesenteric veins. Observe
the lower image noise of the PCD-CT images and the higher vascular contrast. Window
width: 342 HU, center of the window: 56 HU.
In addition to the absolute dose reduction per scan due to higher dose efficiency,
PCD CT offers the theoretical possibility of reducing the required number of total
scans per examination by simultaneous application of several contrast media [9 ]. Several promising studies have shown that it is possible to separate two fundamentally
different contrast media, which were administered with a certain delay or in different
ways, in one scan [32 ]
[33 ]
[34 ]. For example, in an ex vivo phantom, the simultaneous use of rectal and intravenous
contrast media enabled the differentiation between simple polyps and bowel contents
[35 ]. In addition, an animal model study using gadolinium-based and iodine-containing
contrast media to differentiate hepatic arteries and veins in a biphasic single scan
using PCD CT shows promising results and paves the way for correct differentiation
of hepatic lesions based on their characteristic contrast uptake and washout at each
phase [33 ]
[36 ]. Another technique for characterizing liver lesions is so-called volume perfusion
CT (VP CT). This is already used as an alternative to liver MRI on EID CT, as it enables
the degree of arterialization of hepatocellular carcinomas (HCCs) and other tumors
to be quantified [37 ]. Volume perfusion CT of the liver combined with the higher resolution and lower
radiation exposure of PCD CT promise encouraging results and may possibly lead to
changes in the guidelines (see [Fig. 5 ]). In addition, PCD CT makes it easier to assess the liver parenchyma, as the spectral
information can be used to generate VNC images that come very close to the true contrast
media-free (native) datasets. In this way, hepatic steatosis can be detected with
a sensitivity of 94 % from VNC images on the PCD CT [38 ].
Fig. 5 63-year-old man with an HCC. a Volume perfusion PCD-CT scan showing a hyperarterialized liver lesion in segment
8 (white arrows). b Dynamic perfusion PCD-CT scan with visualization of the arterial and venous vessels
in an MIP using 3D-angio post-processing software. c, d Corresponding MRI images show the liver lesion with less contrast enhancement (C
= T1 Vibe DIXON DCE arterial phase axial, d T2 Blade FS axial).
With regard to the kidneys, on the one hand, the energy discrimination between the
grouped photon energies provides information about the elementary composition of particularly
small (< 3 mm) renal stones. On the other hand, compared to EID CT, PCD CT also enables
better detection of such small concretions [39 ].
Bone and bone marrow
PCD CT provides a much better representation of the trabecular bone microstructure,
which is also a primary effect of the improved spatial resolution. On the one hand,
osteoporotic patients benefit from this, since bone strength can be determined via
bone mineral density and fractures can also be better visualized [40 ]. On the other hand, cancer patients benefit from this, since osseous metastases
and their courses can be presented in much more detail (see [Fig. 6 ]). As in other parts of the body, noise reduction is achieved by using ultra-high-resolution
PCD CT image datasets [40 ], which enables more detailed visualization of bony structures and thus better detection
of metastases and fractures [41 ]
[42 ]
[43 ]. In addition, the improved visualization is also associated with a significant reduction
in radiation dose in PCD CT examinations of the extremities of up to 49 % in the wrist
structures [44 ]
[45 ]. Promising results have already been achieved with visualization of cartilage in
patients with knee osteoarthritis, which have the potential to improve the diagnostic
significance of PCT CT in joint diseases [46 ]. PCD CT also makes it possible to distinguish between calcium pyrophosphate and
hydroxylapatite deposits in the joint cartilage. This could provide new insights into
the pathogenesis of rheumatic diseases with crystal deposits [47 ]. In addition, multi-energy post-processing allows the quantification of gout deposits
and the creation of virtual non-calcium images (VNCa) to visualize bone marrow edema
[48 ]. Similar to Dual-Energy EID CT, visualization of bone marrow edema is intended to
make it easier to detect subtle fractures and to assess the fracture value [49 ].
Fig. 6 73-year-old patient with stable multiple myeloma with pronounced lytic changes of
the entire axial skeleton. a DSCT Siemens Somatom Definition Flash. b ) PCD-CT Siemens Naeotom Alpha. Note the significantly lower noise on the PCD-CT scan
and the significantly better edge definition of the lytic bone structure in L5. The
PCD-CT scan allowed a dose reduction of 27% (CTDIvol 11,37mGy vs. 8,19mGy). Window
width: 1500 HU, center of the window: 450 HU.
Cardiovascular imaging
Cardiovascular imaging benefits from PCD CT due to three major technical advantages
over traditional EID CT: Firstly, an improved spatial resolution, secondly, an improved
signal-to-noise ratio, and thirdly, the ability to record all the spectral information
of the photons.
Improved spatial resolution of cardiovascular structures
The segmentation and characterization of coronary plaques and the vascular lumen is
limited by the spatial triggering of the current conventional EID CT [50 ]
[51 ]
[52 ]
[53 ]. In particular, the quantification of coronary stenosis in calcified plaques is
limited by blooming artifacts of coronary alkaline deposits, which depend primarily
on spatial resolution [54 ]. For this reason, coronary stenosis is often overestimated in clinical practice
[52 ]. In phantom studies, it has already been confirmed that the improved spatial resolution
of PCD CT compared to EID CT particularly depicts non-calcified and lipid-rich plaques
more accurately [55 ]
[56 ]. Similarly, compared to EID scanners, the high-resolution PCD CT improves the visualization
of the vascular lumen in the stent and thus the assessment of relevant in-stent stenoses
[57 ]
[58 ]
[59 ].
In this context, it is worth highlighting the “ultra-high-resolution” CT coronary
angiography on PCD CTs that has recently become available, as this also allows the
reduction of blooming artifacts in calcified plaques [60 ]. An initial study showed very good imaging of the vascular lumen and coronary plaques
in UHR mode, especially in the case of reconstruction with hard pliable nuclei [61 ]. [Fig. 7a–d ] show a CT coronary angiography in UHR mode with very good visualization of the lumen
on a PCD scanner. It should be emphasized here that UHR acquisition with the same
target acuity (same convolution kernel, etc.) as “normal” acquisition does not a priori
require an additional dose, but nevertheless allows the benefits of intoxication reduction.
Only the combined use of very sharp kernels and noise minimization is associated with
an increased radiation dose. However, the gain in information justifies the use of
a higher radiation dose, particularly in patients with a higher pre-test probability
of stenosing coronary disease.
Fig. 7 Ultra-high-resolution (UHR) PDC-CT scan (Siemens Naeotom Alpha 120 kV) of an 88-year-old
patient for planning prior to transcatheter aortic valve implantation (TAVI). In addition
to visualization of the aortic valve, diagnostic visualization of the coronary arteries
and thus exclusion of an obstructive coronary artery disease is possible in spite
of pronounced coronary calcifications with an Agatston score of 1281. a Three-dimensional visualization with the cinematic rendering technique; b Stretched MPR of the left anterior descending artery; c Non-contrast coronary calcification CT, axial 2.0-mm slices and reconstruction using
soft-tissue kernel (Qr36); d UHR-CT with contrast agent, axial 0.2-mm slices and reconstruction using vascular
kernel (Bv56, Q4).
In theory, the CT-based fractional flow reserve (CT FFR) should also benefit from
the higher spatial resolution [62 ], as the accuracy depends on the precise segmentation of the coronary vessels.
Improved signal-to-noise ratio of cardiovascular structures
The optimized signal-to-noise (SNR) ratio leads to two major advantages in assessing
coronary stenoses: On the one hand, the detector-enhanced SNR can compensate for the
tougher convolution kernel, which is used to reduce blooming artifacts with a constant
radiation dose [63 ]. On the other hand, in the case of constant SNR, the radiation dose can be reduced,
which particularly benefits patients during follow-up (e. g., Marfan syndrome) [64 ]. [Fig. 8 ] shows a CT coronary angiography on a PCD CT using a high-pitch spiral technique
with a low radiation dose in a young patient. Detector-based elevated SNR can be used
to reduce the contrast media dose in patients with renal insufficiency. For example,
a study of 100 patients with an indication for follow-up PCD CT aortography showed
a 25 % reduction in contrast media compared to EID CT, while maintaining the same
image quality [65 ].
Fig. 8 Cardiac PCD-CT (Siemens Naeotom Alpha 120 kV) in a 41-year-old patient with atypical
thoracic symptoms for ruling out coronary heart disease. Due to the low pretest probability
for the presence of obstructive coronary heart disease and the young patient age,
a low-dose high-pitch spiral examination with 90 kV was performed (CTDIvol: 2.7 mGy;
DLP: 52 mGy*cm). a Three-dimensional visualization of the heart with the cinematic rendering technique;
b Axial 0.6-mm slices after contrast administration and reconstruction using a vascular
kernel (Bv40, Q3); c Stretched MPR of the right coronary artery without detection of stenoses.
Cardiovascular post-processing applications
The ability of PCD CT to register the energy of individual photons and to process
this as separate signals basically has the technical potential of a universal multi-energy
acquisition for all scans [66 ]. However, in contrast to the two-layer detectors of a manufacturer of current EID
scanners, which can only register two different effective photon energies, several
photon energies could in principle be registered here [66 ]. This makes it possible to differentiate the material, enabling VNCa datasets to
be generated, and thus avoiding the need for a previous native scan to quantify the
coronary calcification load [67 ]. In addition, PCD CT allows the calculation of virtual monoenergetic images and
the resulting advantages. On the one hand, it has been shown that greater weighting
of the high-energy bins is helpful in reducing high-contrast artifacts, such as those
from heart valves, pacemakers, or left ventricular support devices [68 ]. This could improve heart valve diagnostics by PCD CTs in the future. On the other
hand, higher weighting of the low-energy bins can increase the iodine contrast and
is therefore important for CT angiography, especially in patients with impaired renal
function.
In addition, spectral image datasets can also be used to create iodine maps in the
pulmonary parenchyma to visualize perfusion deficits and thus allow conclusions to
be drawn about pulmonary artery embolism [69 ]. To date, this has only been tested on the Dual-Source EID CT; however, in view
of the better intraluminal contrast of PCD CT, it should also help to better visualize
small subsegmental embolisms.
Another area of use of the multi-energy technique is the detection of late myocardial
enhancement in the CT for the diagnosis of myocardial infarction. Here, clinical trials
on EID CTs have already demonstrated that multi-energy imaging improved the detection
of late enhancement [70 ]. This has even been demonstrated on EID for late enhancement of non-ischemic causes
[71 ]. Using late enhancement scans, it was technically possible to calculate the myocardial
extracellular volume (ECV) without an additional native phase on PCD CTs in one study.
However, this has not yet been compared with that of MRI, so to date only initial
results have been available, and CT late enhancement cannot yet be considered an established
method. [72 ]
Consistent with abdominal applications, the time-staged application of iodine and
gadolinium – and their differentiation in phantom studies – showed promising results
for the detection of Endoleaks [73 ]. In this way, separate detection of iodine and gadolinium could also be used in
the future to simultaneously perform CT coronary angiography and detect myocardial
infarction in one scan.
Pediatric imaging
When using ionizing radiation on children, radiation exposure considerations are particularly
significant due to the increased radiation sensitivity of children, as the risk of
developing malignant disease is thought to be increased after radiation exposure in
childhood [74 ]
[75 ]. Computed tomography in particular plays a crucial role here, as it accounts for
the majority of radiation exposure in medicine. At the same time, CT, provided that
the indication is correct, is of relevant importance for therapy planning and decision-making
in many pediatric diseases [76 ].
Therefore, the greatest advantage of PCD CT in pediatric radiology is the reduction
of radiation exposure [77 ]. The so-called “low dose” PCD CT significantly reduces the radiation dose on the
one hand, but also maintains the accuracy of relevant anatomical structures and, in
addition, significantly reduces image noise [78 ]. The effect of dose reduction occurs especially in the case of repetitive examinations
(see [Fig. 9 ]).
Fig. 9 16-year-old boy with infiltrates after stem cell transplantation in AML. a, b Third-generation dual-source CT scan (Somatom Force, Siemens Healthineers) - low-dose
protocol with spectral filtering; c PCD-CT scan (Siemens Naeotom Alpha 120 kV) - low-dose protocol with spectral filtering.
Comparable dose values and image quality for all examinations with a CTDI/DLP of 0.31mGy/10.1
mGy*cm (a ), 0.32 mGy/9.8 mGy*cm (b ), and 0.38mGy/12.4 mGy*cm (c ). a Finding prior to treatment; b) Infiltrates in the right upper lobe; c Regressive infiltrates under antibiotic and antiviral therapy. Due to respiratory
problems, various breathing positions were used during the examinations.
Another advantage lies in the extraction of spectral information through PCD. The
full spectral evaluation of the detector data requires a tube voltage of 120 kV or
140 kV, which is rarely used in pediatric patients due to the associated higher radiation
dose. Therefore, in pediatric CT, the tube voltage is often reduced to 70 to 90 kV,
resulting in reduced image quality due to higher noise. However, due to the spectral
data of PCD CT, monoenergetic reconstruction is possible. This can, for example, increase
the contrast and thus reduce the amount of contrast media (see [Fig. 10 ]).
Fig. 10 17-month-old girl with possible pulmonary sequestration in the case of known CPAM
(congenital pulmonary airway malformation). a, b Comparison of contrast between the standard reconstruction at 90 kV (a ) and the monoenergetic reconstruction at 60 kV (b ). The monoenergetic images show an increase in the HU value of the contrast agent,
which contributes to better detectability of vascular structures.
Limitations and outlook
The concept of PCD CT promises numerous advances in oncological, cardiovascular, and
pediatric imaging. However, CT manufacturers have faced a wide range of technical
challenges to achieve excellent clinical applicability compared to EID CTs and others.
In addition to improved image quality and the possibility of radiation dose reduction,
the future clinical focus will presumably be on spectral image information. The different
X-ray attenuating properties of different tissue types or novel contrast media make
it possible to perform density quantification, which could serve as a reference for
secondary diagnostic and therapeutic applications [79 ]. However, a few steps still need to be taken before the entire clinical spectrum
is fully usable:
Firstly in oncological imaging, in which not all post-processing applications, such
as the separation of different types of contrast media or bone marrow imaging in contrasting
image datasets, are currently ready for use.
Furthermore, although a spatial resolution of 0.2 mm and a temporal resolution of
66 ms is currently possible in cardiac imaging, which corresponds to the currently
highest temporal resolution [80 ], spectral information is currently not available in UHR mode.
In pediatric imaging it is hoped that spectral separation of different photon energy
groups may allow multiphase images to be obtained with a single imaging technique
[81 ]. For this purpose, in addition to the necessary contrast media, the corresponding
post-processing applications are still required.
In general, increased spatial resolution and permanent spectral imaging create new
practical challenges, primarily due to significantly larger image datasets, which
places very high demands on the storage capacity of PACS systems and the performance
of post-processing applications [66 ].
Funding
Ministerium für Wirtschaft, Arbeit und Tourismus, Baden Württemberg (35-4223.10/20)
