IEEE Transactions on Radiation and Plasma Medical Sciences

The feasibility of accurate stent visualization with photon-

counting detector CT and K-edge imaging

Journal: IEEE Transactions on Radiation and Plasma Medical Sciences

Manuscript ID TRPMS-2023-0002.R1

Manuscript Type: Radiation Imaging

Date Submitted by the

11-Apr-2023
Author:

Complete List of Authors: Richtsmeier, Devon; University of Victoria, Physics and Astronomy
Rodesch, Pierre-Antoine; University of Victoria, Physics and Astronomy
Iniewski, Kris; Redlen Technologies
Siu, William; Fraser Health Authority
Bazalova-Carter, Magdalena; University of Victoria, Physics and
Astronomy

Radiation Detectors for medical applications, x-ray CT devices < Pre-

Keywords: clinical imaging systems, Clinical/preclinical evaluation/application
studies

Photon-counting detector CT, photon-counting detectors, CT

Free Entry keywords:
angiography, stent, K-edge imaging

https://mc.manuscriptcentral.com/trpms
Page 1 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 1

1
2
3 The feasibility of accurate stent visualization with
4
5
6
photon-counting detector CT and K-edge imaging
7 Devon Richtsmeier, Pierre-Antoine Rodesch, Kris Iniewski, William Siu, and Magdalena Bazalova-Carter
8
9
10
11
12 Abstract—CT stent imaging suffers from blooming and metal be mitigated as much as possible. The artifacts obscure the
artifacts, reducing the diagnostic quality of images in the areas visualization of the stent lumen, which needs to be examined
13
around stents. Photon-counting detectors (PCDs) have been in order to assess whether or not there are any blockages
14 shown to reduce these artifacts. Two clinical scanners, a GE
15 Optima 580 and Discovery IQ, were compared to a bench-top which have developed within the stent. Stent imaging with
16 PCD-CT system for stent imaging with three stents: Medtronic CT is especially problematic with stents with lumen diameters
17 Protégé, Cordis Precise, and Cordis S.M.A.R.T. Control. The under 3 mm [5] , as the lumen covers a smaller area and
apparent strut thickness, lumen diameter, and lumen attenuation visualization is more readily impaired by artifacts. A number
18 of the stents were evaluated in reconstructed images. K-edge
19 of CT advances are currently approaching the point where they
images were also reconstructed to demonstrate more accurate
20 delineation of the tantalum radiopaque markers. PCD-CT offered could be used in the clinic to address some of the issues of
21 lower percent differences for strut thickness for all three stents stent imaging, including ultra-high-resolution (UHR) CT and
22 (p < 0.001) and for lumen diameter for the Protégé and Precise photon-counting detector (PCD) CT [6].
stents (p < 0.01). The lumen attenuation was more accurate
23 with PCD-CT as well (p < 0.01), excluding the comparisons
24 between the Optima 580 and PCD-CT for the Protégé and
25 Precise stents. The PCD-CT system was better able to delineate PCDs are direct-conversion x-ray detectors, which are able
26 stents, specifically strut thickness. The stents were more easily to distinguish the energy of incident x-ray and bin them within
27 distinguished in PCD-CT images and in 3D volume renderings specific energy ranges. As PCDs have been developed which
than the clinical systems. The tantalum radiopaque markers were are able to handle higher flux rates, research has been on-
28 clearly visible in K-edge images due to reduced metal artifacts.
29 going to investigate replacing conventional energy-integrating
30 Index Terms—Photon-counting detector CT, photon-counting detectors (EIDs) in conventional CT scanners to investigate the
detectors, CT angiography, stent, K-edge subtraction imaging benefits PCDs could offer [7]–[11]. To date this has resulted
31
32 in one PCD system being approved for clinical use [12], with
33 I. I NTRODUCTION others currently under development [13]–[15]. The benefits
34 offered by PCDs include reduced metal and blooming artifacts
CT angiography (CTA) is a common imaging assessment as well as increased spatial resolution [10], [11], among others,
35 used to diagnose vascular diseases and monitor potential
36 which could offer gains for CTA for imaging stents. A number
complications before and after treatment [1]. Vascular issues of previous studies have investigated the comparison between
37 such as stenoses can often be treated with the implantation
38 PCD-CT and EID-CT for stent imaging including lumen visu-
of a stent, which can prevent the blood vessel from nar- alization, lumen attenuation, lumen diameter, and the apparent
39 rowing. However, monitoring of the area in which the stent
40 size of stent struts [16]–[22]. PCD-CT also offers the potential
is implanted can be hindered due to the stent itself. Stents for K-edge imaging of high-atomic number contrast agents
41 are usually manufactured from various metals, which cause
42 [23]–[25], such as iodine and gadolinium, as well stents
metal artifacts [2], [3], especially blooming artifacts, due to containing proportions of high-atomic number elements [19].
43 their high atomic number and density. These artifacts can
44 Additionally, the first comparisons with in-human data of the
reduce the diagnostic value of CTA scans [4] and need to two modalities have been published [5], [26].
45
46 This work did not involve human subjects or animals in its research.
This work was partly funded by NSERC Alliance and Engage Plus grants,
47 NSERC CGS-D, an NSERC Discovery grant, the Canada Foundation for
48 In this study we evaluate the performance of a prototype
Innovation, the British Columbia Knowledge Development Fund, and the
49 Canada Research Chair program. bench-top PCD-CT system against two conventional clinical
50 CT systems by imaging several larger (> 5mm diameter)
Devon Richtsmeier is with the Department of Physics and Astronomy at
the University of Victoria, Victoria, BC, V8P 5C2, Canada. Corresponding
carotid artery stents. Though the stents are larger than those
51 author (email: [email protected]).
52 that are currently the most difficult to image, it provides a
Pierre-Antoine Rodesch is with the Department of Physics and Astronomy
53 baseline comparison between our bench-top PCD-CT system
at the University of Victoria, Victoria, BC, V8P 5C2, Canada (email: pier-
[email protected]). and every-day clinical systems. We evaluate the apparent
54 Kris Iniewski is with Redlen Technologies, Saanichton, BC, V8M 1X6,
55 Canada (email: [email protected]). lumen diameter, lumen attenuation and noise, and visualize
56 and quantify the size of radiopaque markers with K-edge
William Siu is with Fraser Health Authority, New Westminster, BC, V3L
3W7, Canada (email: [email protected]). subtraction imaging. We also demonstrate that our bench-top
57 Magdalena Bazalova-Carter is with the Department of Physics and Astron-
58 system provides the smallest measured apparent strut thickness
omy at the University of Victoria, Victoria, BC, V8P 5C2, Canada (email:
59 [email protected]). compared with previous studies.
60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 2 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 2

1
2 II. M ATERIALS AND M ETHODS
µ − µw
3 A. Imaging Systems HU = 1000 ∗ ( ) (1)
µw
4
Three different CT systems were used to image stents in in which µ is the attenuation values in the un-normalized
5
this study: a GE Optima 580 radiation therapy simulator, a image and µw is mean signal within the water-containing
6
GE Discovery IQ scanner (both GE Healthcare, Chicago, IL), Eppendorf tubes within the un-normalized image. PCD-CT
7
and a prototype bench-top PCD-CT scanner. All three scanners and HR PCD-CT were reconstructed from the same projection
8
can be seen in Fig. 1A–C. The bench-top PCD-CT system data set. PCD-CT images were reconstructed in order to mimic
9
consisted of a Comet MXR 160/22 x-ray tube (Comet Tech- the reconstructed clinical CT images as closely as possible in
10
nologies, San Jose, CA), rotation and motion stages (Newport terms of pixel size and slice thickness, while HR PCD-CT
11
Corporation, Irving, CA), and a state-of-the-art flat panel PCD offered the standard in-house reconstruction parameters.
12
(Redlen Technologies, Saanichton, BC, Canada). The source
13
to isocenter distance was set to 322 mm with a source to de-
14 E. K-edge Image Reconstruction
tector distance of 578 mm. The PCD consisted of 2 mm thick
15 K-edge subtraction images of tantalum were reconstructed
cadmium zinc telluride (CZT) crystal with a 330 µm pixel
16
pitch and an active area of 8×190 mm2 , which gives a 106 using the K-edge decomposition algorithm (KDA) described
17 by Zhang et al [25]. Briefly, Ta-specific sinograms were ac-
mm field of view with 4.5 mm Z-coverage at isocenter [27].
18 quired according to the following equation:
The PCD is capable of the energy discrimination with binning
19
of up to six energy bins and can operate without polarization
20 µ̂bg,U · TL − µ̂bg,L · TU
at counts rates up to 650 Mcps/mm2 [28], [29]. DK (r) = (2)
21 µ̂K,L · µ̂bg,U − µ̂K,U · µ̂bg,L
22
where DK refers to the material-specific sinogram, µ̂bg is
23 B. Stents and Imaging Phantom
average the mass attenuation coefficient of the background,
24 Three nitinol stents were imaged in this study: a Medtronic µ̂K is the average mass attenuation coefficient of the K-edge
25 Protégé stent (Medtronic plc, Minneapolis, MN), a Cordis Pre- material, TL is the sinogram of the lower bin, and TU is the
26 cise stent, and a Cordis S.M.A.R.T. Control stent (both Cordis, sinogram of the upper bin. The subscripts U and L refer to the
27 Santa Clara, CA). Strut thicknesses were measured using a mi- upper and lower bin, respectively. So µ̂bg,L would be the aver-
28 croscope with a 4X objective and found to be 0.186 mm, 0.238 age mass attenuation coefficient of the background within the
29 mm, and 0.177 mm, respectively. All other physical dimen- energy range of the lower bin. To reconstruct the Ta-specific
30 sions were measured with calipers. The lumen diameters of sinogram the 67-81 keV and the 81-97 keV ranges were used
31 the stents were 5.82 mm, 5.71 mm, and 5.84 mm, respectively. for the lower and upper bins, respectively. Once the Ta-specific
32 The two Cordis stents also had tantalum radiopaque markers. sinogram was found, the K-edge images were reconstructed
33 All stents were inserted into plastic straws to mimic being with the FDK algorithm and a Shepp-Logan filter.
34 compressed in vivo, inserted into a custom 100-mm diameter,
35 30-mm thick high-density polyethylene phantom (Fig. 1D, E),
36 and filled with water. In addition to the stents, the phantom F. Dose
37 held 6.3 mm and 5.5 mm diameter Eppendorf tubes filled with CTDIvol (CT Dose Index) values were collected from the
38 water. clinical CT scanners and measured on the PCD-CT system
39 with the CTDI head phantom and 100-mm CTDI probe.
40
C. Data Acquisition
41 G. Conventional CT Image Analysis
42 Acquisition and reconstruction parameters for all systems
43 can be found in Table I. PCD-CT and HR PCD-CT were CT images were analyzed in a number of ways. For both
44 reconstructed from the same data set, with differing recon- PCD-CT resolution image types, the 35-120 keV CT images
45 struction parameters. For the PCD-CT acquisitions, the energywere analyzed in order to provide the closest comparison to the
46 thresholds were set to 35, 52, 67, 81, 95, and 120 keV. This clinical images. First, 3-dimensional (3D) volume renderings
47 created energy bins with ranges between each threshold pair, of the stents were created in 3D Slicer and segmented using
48 with the addition of a bin summing counts from all other a lower threshold of 1750 HU. Second, the lumen attenuation
49 energy ranges, 35-120 keV. was determined. A circular region-of-interest (ROI) between
50 3.5 and 3.7 mm in diameter was delineated within the stent
51 and the mean and variance within the ROI was calculated for
D. Image Reconstruction a minimum of seven slices. The lumen diameter and the strut
52
53 Clinical CT images were reconstructed using filtered back thickness were evaluated by measuring their apparent values in
54 projection with the BONEPLUS kernel. PCD-CT images the CT images. The center of the stent was first found using
55 were reconstructed bin-wise using the Feldkamp-David-Kress the Hough Gradient method [32], and fine-tuned manually.
56 algorithm [30] with a Shepp-Logan filter implemented in the Line profiles were then interpolated from the center through
57 TIGRE package for Python [31]. Once reconstructed, the individual struts and the full-width-half-maximum (FWHM) of
58 PCD-CT images were normalized bin-wise to Hounsfield units the peak formed from the strut’s signal was found. The lumen
59 (HU) utilizing Eq. 1: radius was calculated as the mean distance from the center of
60 https://mc.manuscriptcentral.com/trpms
Page 3 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 3

1
2 TABLE I
CT IMAGE ACQUISITION AND RECONSTRUCTION PARAMETERS .
3
4 Parameter GE Optima GE PCD-CT HR
5 580 Discovery PCD-CT
IQ
6 Collimation 16 × 16 × 24 × 24 ×
7 0.625mm 0.625mm 0.184mm 0.184mm
8 Tube Voltage 120 120 120 120
(kV)
9 Tube Current 180 180 1 1
10 (mA)
11 Rotation Time 1.0 1.0 180 180
12 (s)
Helical Pitch 0.5625 0.5626 0.5626 0.5626
13 Beam Filter HEAD HEAD 6 mm Al 6 mm Al
14 FILTER* FILTER*
15 Focal Spot Size 0.7 0.7 ∼0.4** ∼0.4**
(mm)
16 Source to De- 1063 949 578 578
17 tector Distance
18 Fig. 1. Setup images of the (A) GE Optima 580, (B) GE Discovery IQ, and (mm)
(C) the bench-top PCD-CT system. (D) Phantom layout. (E) Image the Cordis Source to Axis 606 541 322 322
19 Precise stent inside the straw. Distance (mm)
20 Reconstruction BONEPLUS BONEPLUS Shepp- Shepp-
21 Filter/Kernel *** *** Logan Logan
Reconstructed 15.3 12.7 10.5 10.5
22 the stent to the closer FWHM point. The strut thickness was
FOV (cm)
23 calculated as the FWHM of the peak itself. Finally, the modu- Reconstructed 0.299 0.248 0.248 0.205
24 lation transfer function (MTF) was calculated in all image sets Pixel size (mm)
25 based on the method detailed by Takenaga et al [33]. Briefly, Slice Thickness 0.625 0.625 0.625 0.208
(mm)
26 the center and radius of the 100 mm diameter phantom was
MTF 10% 1.16 1.16 1.29 1.33
27 found using the Hough method and fine-tuned manually. A (lp/mm)
28 normalized image was then calculated based on the equation: *HEAD FILTER is the filtration used from head scans in GE CT systems.
**The PCD-CT x-ray tube focal spot was measured using the EN 12543
29 convention, the approximate IEC 60336 size is shown.
I(i, j) − IBG
30 IN (i, j) = (3) ***BONEPLUS is a GE proprietary reconstruction kernel.
31 Iphantom − IBG PCD-CT = photon-counting detector CT
HR PCD-CT = high-resolution PCD-CT
32 in which IN is the normalized image, I is the original im- MTF = modulation transfer function
33 age, i and j are the row and column numbers for individual FOV = field of view
34 pixels, IBG is the mean signal outside of the phantom, and
35 Iphantom is the mean signal within the phantom. The signal
36 struts in the conventional CT images. The diameter of the Ta
of each pixel within 5 mm inside and outside of the edge of
37 markers were measured using calipers at ∼0.75 mm, though
the phantom was plotted against its distance from the center
38 the tantalum is pressed into the nitinol disks that make up the
of the circle, creating an over-sampled edge spread function
39 marker structure, so that actual size of the tantalum itself will
(ESF). The over-sampled ESF was fitted using least-squares
40 be somewhat smaller.
to the equation:
41
a
42 ESF (x) = +d (4) I. Statistical Analysis
43 1 + exp[−b(x − c)] Normality of the quantitative variables were assessed via the
44 where x is the distance from the center of the circle. a, b, c, Shapiro-Wilk test. All data, excluding the Ta marker data, were
45 and d were initially set to 1, -1, rp hantom, and 0, respectively. found to be non-normal and thus displayed as the median and
46 rp hantom is the radius of the phantom. The ESF was then interquartile range (IQR). The Wilcoxon rank-sum test was
47 differentiated to find the line spread function (LSF), and the used to compare the variables. Given the Ta marker data had
48 Fast Fourier Transform applied to the LSF to find the MTF. limited data points, it was reported as the mean and minimum
49 and maximum values.
50 H. K-edge Image Analysis
51 III. R ESULTS
52 The Ta marker size for the Cordis Precise and S.M.A.R.T.
Control stents was compared between clinical system images, A. Qualitative Evaluation
53
54 the two PCD-CT method images, and the two resulting K- Fig. 2 demonstrates representative CT slices from the GE
55 edge images from both PCD-CT methods. The size of the Optima 580 (Fig. 2A, E, I), GE Discovery IQ (Fig. 2B, F, J),
56 markers was measured by drawing a single profile through PCD-CT (Fig. 2C, G, K), and HR PCD-CT (Fig. 2D, H, L) for
57 each Ta marker in the slice in which they appeared largest. the Medtronic Protégé, Cordis Precise, and Cordis S.M.A.R.T
58 The size of each marker was measured as the FWHM of the Control stents, respectively. The CT images for both PCD-
59 profile, in the same manner as the width of the individual CT options were reconstructed using the 35-120 keV bin. The
60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 4 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 4

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 Fig. 2. CT images of the plastic phantom with close-ups of the stents. Scans
of (A–D) the Medtronic Protégé stent, (E–H) the Cordis Precise stent, and
21 (I–L) the Cordis S.M.A.R.T. Control stent with each of the four imaging
22 methods. W/L = [2100/450].
23
Fig. 4. K-edge subtraction of radiopaque markers of the Cordis S.M.A.R.T.
24 Control stent. CT image acquired with the (A) GE Discovery IQ and (B)
25 PCD-CT. (C) Ta K-edge decomposition image. (D) PCD-CT image with the
26 Ta K-edge image overlaid on top of it. W/L = [2100/450].
27
28 PCD-CT compared to clinical CT, radiopaque markers can also
29 be better defined through the use of the energy information that
30 is inherent to PCD-CT. Fig. 4 shows the radiopaque markers
31 on the Cordis S.M.A.R.T Control stent in images from both
32 the GE Discovery IQ (Fig. 4A) and PCD-CT (Fig. 4B). Metal
33 and blooming artifacts can be seen in both, although they are
34 more prevalent in the PCD-CT image than in the HR PCD-
35 CT image. Using K-edge subtraction imaging, a Ta-specific
36 image could be created (Fig. 4C), which mitigated many of
37 the metal and blooming artifacts. The Ta-specific image could
Fig. 3. 3D volume renderings of all three stents. Volume renderings created
38 from scans of (A–D) the Medtronic Protégé stent, (E–H) the Cordis Precise be viewed alone or as an overlay to the PCD-CT image (Fig.
39 stent, and (I–L) the Cordis S.M.A.R.T. Control stent with each of the four 4D) in order to better register the markers within the phantom.
40 imaging methods.
41
B. Quantitative Evaluation
42
43 lumen of the stents appears larger in PCD-CT and HR PCD- Fig. 5 shows boxplots of the median and IQR for strut
CT images compared to both GE scanners, and the struts are thickness and lumen diameter. For all three stents, PCD-CT
44
easily delineated in PCD-CT, whereas they are not as easily and HR PCD-CT demonstrated strut thicknesses closer to
45
separated in the GE CT images. the true thicknesses than either clinical scanner (p < 0.001)
46
47 The 3D stent volume renderings created for all four image (Fig. 5A–C). PCD-CT and HR PCD-CT also demonstrated
48 sets are shown in Fig. 3. The most accurate representation values closer to the true value for the lumen diameter than
49 of the stents was produced by HR PCD-CT (Fig. 3D, H, L). either clinical scanner for the Medtronic Protégé and Cordis
50 Individual struts and the greater pattern of the stents were Precise stents (p < 0.01). The lumen diameter results were not
51 visible and easily recognizable. PCD-CT (Fig. 3C, G, K) significantly different for the Cordis S.M.A.R.T. Control stent
52 offered less accurate representations, although individual struts when comparing the clinical scanners to the PCD-CT system
could be distinguished and the pattern could still be recognized (p > 0.01).
53
in some areas of the stents when comparing them to the HR Table II shows the absolute difference of lumen attenuation
54
PCD-CT renderings. Neither of the clinical scanners offered from water (0 HU). The median HU value fell closer to zero
55
the same strut delineation or greater pattern recognizability with PCD-CT and HR PCD-CT than the clinical scanners (p
56
(Fig. 3A–B, E–F, I–J) and appeared largely as solid cylinders < 0.01). The only comparisons between HU data that were
57
with the occasional window into the interior lumen. not significantly different were between the GE Optima 580
58
59 In addition to better visualization of the stent structure in and PCD-CT for the Protégé and Precise stents (p = 0.065 and
60 https://mc.manuscriptcentral.com/trpms
Page 5 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 5

1
2
3
4
5
6
7
8
9
10
11
12 Fig. 5. Measured stent dimensions. The shaded red bands indicate the true
dimensions. Orange lines indicate median values, and the boxes indicate
13 interquartile range. (A–C) The measured strut thickness for the (A) Medtronic
14 Protégé, (B) Cordis Precise, and (C) Cordis S.M.A.R.T Control stents with the
15 four image sets. (D–F) The measured lumen diameter for the (D) Medtronic
Protégé, (E) Cordis Precise, and (F) Cordis S.M.A.R.T Control stents with
16 the four image sets.
17
18 TABLE II
19 A BSOLUTE DIFFERENCE OF LUMEN ATTENUATION .
20
Medtronic Cordis Precise Cordis
21 Protégé S.M.A.R.T.
22 Control
23 (median; (median; (median;
IQR) IQR) IQR)
24 Optima 580 -14 HU; -10 HU; -5 HU; Fig. 6. Measured tantalum diameter for (A) the Cordis Precise stent and (B)
25 -61–41HU -66–44 HU -66–43 HU the Cordis S.M.A.R.T. Control stent.
26 Discovery -15 HU; -23 HU; -22 HU;
IQ -56–25 HU -71–21 HU -64–21 HU
27 PCD-CT -4 HU; -5 HU; 4 HU;
28 -63–64 HU -86–88 HU the strut thickness with PCD-CT can be largely attributed to
-59–63 HU
29 HR 2 HU; 5 HU; 12 HU; the higher inherent spatial resolution of PCDs compared with
PCD-CT -52–65 HU -82–104 HU -49–79 HU
the EIDs on the clinical scanners. The PCD has a detector
30 Lumen attenuation is presented as the median absolute difference
31 and IQR from 0 HU, the expected mean attenuation of water.
pixel pitch of 0.33 mm, which is likely smaller than the
32 EID detectors, allowing for less partial volume effects in the
33 projections. This allowed for reduced blooming artifacts in
34 p = 0.042, respectively). However, the IQRs for all the stents the reconstructions, even though the PCD-CT and clinical CT
35 were smaller with the clinical scanners than either PCD-CT scans were reconstructed with nearly identical voxel sizes,
36 method. which should ensure fewer differences in terms of partial
37 Fig. 6 shows the Ta marker size tangential to the edge of the volume effects in the reconstructed images. With HR PCD-CT,
38 circle made by the stent; perpendicular to the most prevalent which had a smaller voxel size compared to the other three
39 metal artifacts in the PCD-CT image (Fig. 4B). Both the PCD- image sets, there were even better results in terms of apparent
40 CT and HR PCD-CT demonstrate smaller diameters than the strut thickness, demonstrating that partial volume effects do
41 measured diameter of the nitinol and tantalum together, while have some effect on the blooming artifacts. With smaller stents
42 the clinical systems show larger values. The K-edge measure- it would be crucial to reconstruct at a higher resolution. The
43 ments are similar for the Precise stent, but smaller for the two PCD-CT methods also benefited from a smaller focal spot
44 S.M.A.R.T. Control. size, which would cause less focal spot blurring, as well as
45 The 10% value of the MTF for each system can be seen smaller source to axis and source to detector distances, leading
46 in Table I. The reported CTDIvol values for the Optima 580 to less object magnification over multiple pixels.
47 and the Discovery IQ were 28.60 mGy and 62.16 mGy, For lumen diameter (Fig. 5D-F), there was little variation in
48 respectively. The measured CTDIvol for the PCD-CT setup the measurements, due in part to the large lumen diameter. The
49 was 93.30 mGy. lumen edges were more visible in the PCD-CT and HR PCD-
50 CT images, though this was likely due to the BONEPLUS
51 kernel creating dark undershoots around the individual struts.
IV. D ISCUSSION Median lumen attenuation (Table II) was more accurate
52
53 We evaluated the performance of our bench-top PCD-CT for PCD-CT and HR PCD-CT compared to the two clinical
54 scanner with respect to two clinical EID-CT scanners for scanners, though again the differences were small. A more
55 stent imaging. Quantitatively, we found that the apparent accurate representation of what would be seen in vivo would
56 strut thickness was closest to the physical measurement using be if there was contrast within the stent lumen. This would
57 PCD-CT and HR PCD-CT for all stents. To our knowledge, make the stent structures less visible as the attenuation would
58 we also demonstrated the smallest measured apparent strut be higher within the lumen and could have some effect on the
59 thickness. The significant improvement in the measurement of noise within the lumen. However, in this simpler case, we were
60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 6 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 6

1
2 also concerned with the accuracy of the signal to the expected which would serve to limit the noise in reconstructed images
3 value of 0 HU as well as the noise that was inherent to the and has a prototype detector, which is still being developed.
4 stent. In respect to the noise, the IQR of the lumen attenuation With that said, the bench-top PCD-CT system outperformed
5 is larger for both PCD-CT methods, indicating higher image the current, clinical EID scanners in terms measuring apparent
6 noise within the lumen diameter. The higher noise can be strut thickness.
7 attributed mostly to the high degree of non-conforming pixels In regard to the future clinical possibilities with this PCD
8 whose signal deviations are not fully mitigated by flat-field technology, the prospects are promising. The superior colli-
9 corrections. We also notice that these pixels, especially those mation and resolution of the PCD-CT methods is due almost
10 that create ring artifacts, can lead to additional metal artifact exclusively to the detector itself. Unlike with an EID, PCDs
11 streaks between the metal and artificially high signal areas require no septa between pixels in the detector; there is
12 within the ring artifacts [27]. And we see residual metal relatively little spread of the electrons produced in the PCD
13 artifact streaks within the PCD-CT images. For HR PCD-CT, crystal when compared with the potential light spread in EIDs.
14 the increase in noise over PCD-CT was also due to the smaller The minimal spread of electrons is due to the high bias
15 slice thickness, which resulted in reduced counts per voxel, and voltage applied and the fact that the anodes are pixelated. The
16 thus higher noise. The dose we saw with PCD-CT was also collimation in the PCD-CT methods is due to the size of the
17 higher than either clinical scanner, in fact it was three times as individual rows of the detector and the geometry of the setup.
18 high as the dose recorded by the Optima 580. This is due in Given that the detector specifications should remain the same
19 large part to the source to axis distance differences between when implemented in a clinical system, the collimation would
20 the scanners. If the dose was scaled accordingly, we would be approximately 24×0.188 mm with a geometry comparable
21 record CTDIvol values of 90.67 mGy and 175.47 mGy for the to the clinical systems studied here. Clinical results could even
22 Optima 580 and Discovery IQ, respectively, which compare be further improved with the use of proprietary kernels and
23 well to the 93.30 mGy for the PCD-CT system. other, more sophisticated reconstruction methods.
24 Qualitatively, the best results were seen with HR PCD-CT,
25 which offered the best 3D volume rendering of the stents and V. C ONCLUSION
26 subtly reduced blooming artifacts when compared to PCD- PCD-CT offered better stent visualization than current clin-
27 CT. However, both PCD-CT methods outperformed the EID- ical CT scanners including more accurate quantification of
28 CT scanners, offering much better delineation of individual apparent strut thickness due to the increased spatial resolution
29 struts in CT images as well as better 3D volume rendering of offered by photon-counting detectors. Additionally, the energy
30 the stents. In addition, PCD-CT offers the ability to visualize information of PCDs could be used to visualize radiopaque
31 the radiopaque markers with K-edge subtraction imaging, markers without metal artifacts in tantalum-specific K-edge
32 allowing the marker location and shape to be more accurately images.
33 determined (Fig. 4C, D). The measured Ta sizes showed
34 the best results with the K-edge images, as they mitigated ACKNOWLEDGMENT
35 the majority of the metal artifacts seen in the conventional The authors would like to thank Wilson Lo and Clay
36 images (Fig. 4A, B). A large difference was seen between Lindsay for their assistance in collecting the clinical data,
37 the sizes measured in the K-edge images compared to the along with Mina Bechai and the staff at Initio Medical, as
38 clinical systems, while less of a difference was seen between well as Kyle Bromma for acquiring microscope images of the
39 the the K-edge and conventional PCD-CT images. The main stents. All authors declare that they have no known conflicts of
40 reason for this being that the thickness of the Ta markers was interest in terms of competing financial interests or personal
41 measured along the short axis of the markers (Fig. 4C) in order relationships that could have an influence or are relevant to
42 to avoid the excessive metal artifacts which occurred along the the work reported in this paper.
43 longer axis in the conventional PCD-CT images (Fig. 4B). One
44 benefit of Ta-specific images for clinical imaging would be the R EFERENCES
45 reduction of metal artifacts around the markers, leading to less
[1] K. K. Kumamaru, B. E. Hoppel, R. T. Mather, and F. J. Rybicki, “CT
46 obfuscation of the stent lumen. Angiography: Current Technology and Clinical Use,” Radiol. Clin. N.
47 There are a number of limitations to our study. First, neither Am., vol. 48, no. 2, pp. 213–235, Mar 2010.
clinical scanner would be utilized for coronary CT angiogra- [2] W. A. Kalender, R. Hebel, and J. Ebersberger, “Reduction of CT artifacts
48 caused by metallic implants,” Radiology, vol. 164, no. 2, pp. 576–577,
49 phy as they would not supply the necessary resolution to image Aug 1987.
50 the small stents utilized in coronary arteries. However, the [3] F. E. Boas and D. Fleischmann, “CT artifacts: Causes and reduction
stents used in this study are carotid artery stents and were used techniques,” Imaging Med., vol. 4, no. 2, pp. 229–240, 2012.
51 [4] O. Ghekiere, R. Salgado, N. Buls, T. Leiner, I. Mancini, P. Vanhoenacker,
52 mainly as a comparison to evaluate the performance of our P. Dendale, and A. Nchimi, “Image quality in coronary CT angiography:
53 prototype bench-top PCD-CT system. The clinical scanners Challenges and technical solutions,” Brit. J. Radiol., vol. 90, no. 1072,
also used the BONEPLUS kernel which had some dark under- 2017.
54 [5] S. Boccalini, S. A. Si-Mohamed, H. Lacombe, A. Diaw, M. Varasteh,
55 shoots are the stent struts, which affected the lumen visibility P. A. Rodesch, M. Villien, M. Sigovan, R. Dessouky, P. Coulon,
56 in images. Both clinical scanners had x-ray focal spots which Y. Yagil, E. Lahoud, K. Erhard, G. Rioufol, G. Finet, E. Bonnefoy-
were larger than the PCD-CT system’s focal spot and had Cudraz, C. Bergerot, L. Boussel, and P. C. Douek, “First In-Human
57 Results of Computed Tomography Angiography for Coronary Stent
58 larger source to detector and source to axis distances. However, Assessment with a Spectral Photon Counting Computed Tomography,”
59 the PCD-CT system is not equipped with an anti-scatter grid, Invest. Radiol., vol. 57, no. 4, pp. 212–221, Aug 2022.

60 https://mc.manuscriptcentral.com/trpms
Page 7 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 7

1
2 [6] A. C. Kwan, A. Pourmorteza, D. Stutman, D. A. Bluemke, and J. A. [24] C. A. S. Dunning, J. O’Connell, S. M. Robinson, K. J. Murphy, A. L.
Lima, “Next-generation hardware advances in CT: Cardiac applications,” Frencken, F. C. J. M. van Veggel, K. Iniewski, and M. Bazalova-Carter,
3 Radiology, vol. 298, no. 1, pp. 3–17, Nov 2021. “Photon-counting computed tomography of lanthanide contrast agents
4 [7] S. Leng, M. Bruesewitz, S. Tao, K. Rajendran, A. F. Halaweish, N. G. with a high-flux 330-µm-pitch cadmium zinc telluride detector in a
5 Campeau, J. G. Fletcher, and C. H. McCollough, “Photon-counting table-top system,” J. Med. Imaging, vol. 7, no. 03, p. 1, Jun 2020.
Detector CT: System Design and Clinical Applications of an Emerging [25] Z. Zhang, J. Hu, X. Zhang, Q. Xu, M. Li, C. Wei, L. Wei, and Z. Wang,
6 Technology,” RadioGraphics, vol. 39, no. 3, pp. 729–743, May 2019. “Experimental research of the energy bins for K-edge imaging using a
7 [8] M. J. Willemink, M. Persson, A. Pourmorteza, N. J. Pelc, and photon counting detector: a phantom and mice study,” Radiat. Detect.
8 D. Fleischmann, “Photon-counting CT: Technical principles and clinical Technol. Methods, no. 0123456789, 2020.
prospects,” Radiology, vol. 289, no. 2, pp. 293–312, Nov 2018. [26] S. A. Si-Mohamed, S. Boccalini, H. Lacombe, A. Diaw, M. Varasteh,
9 [9] T. Flohr, M. Petersilka, A. Henning, S. Ulzheimer, J. Ferda, and P. A. Rodesch, R. Dessouky, M. Villien, V. Tatard-Leitman, T. Bochaton,
10 B. Schmidt, “Photon-counting CT review,” Phys. Med., vol. 79, no. P. Coulon, Y. Yagil, E. Lahoud, K. Erhard, B. Riche, E. Bonnefoy,
11 October, pp. 126–136, Nov 2020. G. Rioufol, G. Finet, C. Bergerot, L. Boussel, J. Greffier, and P. C.
Douek, “Coronary CT Angiography with Photon-counting CT: First-In-
12 [10] A. S. Wang and N. J. Pelc, “Spectral Photon Counting CT: Imaging
Human Results,” Radiology, vol. 303, no. 2, pp. 303–313, May 2022.
Algorithms and Performance Assessment,” EEE Trans. Radiat. Plasma
13 Med. Sci., vol. 5, no. 4, pp. 453–464, Jul 2021. [27] D. Richtsmeier, J. O’Connell, P. A. Rodesch, K. Iniewski, and
14 [11] S. S. Hsieh, S. Leng, K. Rajendran, S. Tao, and C. H. McCollough, M. Bazalova-Carter, “Metal artifact correction in photon-counting de-
tector computed tomography: metal trace replacement using high-energy
15 “Photon Counting CT: Clinical Applications and Future Developments,”
data,” Med. Phys., Sep 2022.
EEE Trans. Radiat. Plasma Med. Sci., vol. 5, no. 4, pp. 441–452, Jul
16 2021. [28] K. Iniewski, C. Hansson, E. Guliyev, G. Prekas, and M. Ayukawa,
17 [12] K. Rajendran, M. Petersilka, A. Henning, E. R. Shanblatt, B. Schmidt, “Performance Characteristics of 250+ Mcps/mm2 CZT Detector Module
for Spectral Computed Tomography,” in 5th Workshop on Medical
18 T. G. Flohr, A. Ferrero, F. Baffour, F. E. Diehn, L. Yu, P. Rajiah,
Applications of Spectroscopic X-ray Detectors, CERN, Geneva., 2019a.
J. G. Fletcher, S. Leng, and C. H. McCollough, “First Clinical Photon-
19 counting Detector CT System: Technical Evaluation,” Radiology, vol. [29] K. Iniewski, M. C. Veale, and M. Bazalova-Carter, “High-flux CZT for
20 303, no. 1, pp. 130–138, Apr 2022. new Frontiers in computed tomography (CT), non-destructive testing
(NDT) and high-energy physics,” in IEEE Nuclear Sci. Symp. and Med.
21 [13] J. da Silva, F. Grönberg, B. Cederström, M. Persson, M. Sjölin,
Imaging Conf, 2019b.
Z. Alagic, R. Bujila, and M. Danielsson, “Resolution characterization
22 of a silicon-based, photon-counting computed tomography prototype [30] L. A. Feldkamp, L. C. Davis, and J. W. Kress, “Practical cone-beam
23 capable of patient scanning,” J. Med. Imaging, vol. 6, no. 04, p. 1, algorithm,” J. Opt. Soc. Am. A., vol. 1, no. 6, pp. 612–619, Feb 1984.
[31] A. Biguri, M. Dosanjh, S. Hancock, and M. Soleimani, “TIGRE: A
24 Oct 2019.
MATLAB-GPU toolbox for CBCT image reconstruction,” Biomed. Phys.
[14] X. Zhan, R. Zhang, X. Niu, I. Hein, B. Budden, S. Wu, N. Markov,
25 C. Clarke, Y. Qiang, H. Taguchi, K. Nomura, Y. Muramatsu, Z. Yu,
Eng. Express., vol. 2, no. 5, p. 055010, Sep 2016.
26 T. Kobayashi, R. Thompson, H. Miyazaki, and H. Nakai, “Compre-
[32] H. Yuen, J. Princen, J. Illingworth, and J. Kittler, “Comparative study
of Hough Transform methods for circle finding,” Image Vision Comput.,
27 hensive evaluations of a prototype full field-of-view photon counting
vol. 8, no. 1, pp. 71–77, 1990.
CT system through phantom studies,” Dec 2022. [Online]. Available:
28 https://arxiv.org/abs/2212.13337v1http://arxiv.org/abs/2212.13337
[33] T. Takenaga, S. Katsuragawa, M. Goto, M. Hatemura, Y. Uchiyama, and
29 [15] S. A. Si-Mohamed, S. Boccalini, M. Villien, Y. Yagil, K. Erhard,
J. Shiraishi, “Modulation transfer function measurement of CT images
by use of a circular edge method with a logistic curve-fitting technique,”
30 L. Boussel, and P. C. Douek, “First Experience With a Whole-Body
Radiol. Phys. Technol., vol. 8, no. 1, pp. 53–59, Jan 2015.
Spectral Photon-Counting CT Clinical Prototype,” Invest. Radiol., vol.
31 Publish Ah, no. 00, Feb 2023.
32 [16] M. Mannil, T. Hickethier, J. Von Spiczak, M. Baer, A. Henning,
33 M. Hertel, B. Schmidt, T. Flohr, D. Maintz, and H. Alkadhi, “Photon-
Counting CT: High-Resolution Imaging of Coronary Stents,” Invest.
34 Radiol., vol. 53, no. 3, pp. 143–149, Mar 2018.
35 [17] J. Von Spiczak, M. Mannil, B. Peters, T. Hickethier, M. Baer, A. Hen-
36 ning, B. Schmidt, T. Flohr, R. Manka, D. Maintz, and H. Alkadhi, “Pho-
37 ton Counting Computed Tomography with Dedicated Sharp Convolution
Kernels: Tapping the Potential of a New Technology for Stent Imaging,”
38 Invest. Radiol., vol. 53, no. 8, pp. 486–494, Aug 2018.
39 [18] R. Symons, Y. De Bruecker, J. Roosen, L. Van Camp, T. E. Cork,
40 S. Kappler, S. Ulzheimer, V. Sandfort, D. A. Bluemke, and A. Pour-
morteza, “Quarter-millimeter spectral coronary stent imaging with
41 photon-counting CT: Initial experience,” J. Cardiovasc. Comput., vol. 12,
42 no. 6, pp. 509–515, Oct 2018.
43 [19] M. Sigovan, S. Si-Mohamed, D. Bar-Ness, J. Mitchell, J. B. Langlois,
P. Coulon, E. Roessl, I. Blevis, M. Rokni, G. Rioufol, P. Douek, and
44 L. Boussel, “Feasibility of improving vascular imaging in the presence of
45 metallic stents using spectral photon counting CT and K-edge imaging,”
46 Sci. Rep., vol. 9, no. 1, pp. 1–9, Dec 2019.
[20] G. Bratke, T. Hickethier, D. Bar-Ness, A. C. Bunck, D. Maintz, G. Pahn,
47 P. Coulon, S. Si-Mohamed, P. Douek, and M. Sigovan, “Spectral
48 Photon-Counting Computed Tomography for Coronary Stent Imaging:
49 Evaluation of the Potential Clinical Impact for the Delineation of In-
Stent Restenosis,” Invest. Radiol., vol. 55, no. 2, pp. 61–67, Feb 2020.
50 [21] J. R. Rajagopal, F. Farhadi, T. Richards, M. Nikpanah, and M. Y. Chen,
51 “Evaluation of Coronary Plaques and Stents with Conventional and
52 Photon-counting CT : Benefits of High-Resolution Photon-counting CT,”
Radiol. Cardiothorac. Imaging, vol. 3, no. 5, Oct 2021.
53 [22] B. Petritsch, N. Petri, A. M. Weng, M. Petersilka, T. Allmendinger, T. A.
54 Bley, and T. Gassenmaier, “Photon-Counting Computed Tomography for
55 Coronary Stent Imaging: In Vitro Evaluation of 28 Coronary Stents,”
Invest. Radiol., vol. 56, no. 10, pp. 653–660, Oct 2021.
56 [23] D. Richtsmeier, C. A. S. Dunning, K. Iniewski, and M. Bazalova-Carter,
57 “Multi-contrast K-edge imaging on a bench-top photon-counting CT
58 system: acquisition parameter study,” J. Instrum., vol. 15, no. 10, Oct
2020.
59
60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 8 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 1

1
2
3 The feasibility of accurate stent visualization with
4
5
6
photon-counting detector CT and K-edge imaging
7 Devon Richtsmeier, Pierre-Antoine Rodesch, Kris Iniewski, William Siu, and Magdalena Bazalova-Carter
8
9
10
11
12 Abstract—CT stent imaging suffers from blooming and metal be mitigated as much as possible. The artifacts obscure the
artifacts, reducing the diagnostic quality of images in the areas visualization of the stent lumen, which needs to be examined
13
around stents. Photon-counting detectors (PCDs) have been in order to assess whether or not there are any blockages
14 shown to reduce these artifacts. Two clinical scanners, a GE
15 Optima 580 and Discovery IQ, were compared to a bench-top which have developed within the stent. Stent imaging with
16 PCD-CT system for stent imaging with three stents: Medtronic CT is especially problematic with stents with lumen diameters
17 Protégé, Cordis Precise, and Cordis S.M.A.R.T. Control. The under 3 mm [5] , as the lumen covers a smaller area and
apparent strut thickness, lumen diameter, and lumen attenuation visualization is more readily impaired by artifacts. A number
18 of the stents were evaluated in reconstructed images. K-edge
19 of CT advances are currently approaching the point where they
images were also reconstructed to demonstrate more accurate
20 delineation of the tantalum radiopaque markers. PCD-CT offered could be used in the clinic to address some of the issues of
21 lower percent differences for strut thickness for all three stents stent imaging, including ultra-high-resolution (UHR) CT and
22 (p < 0.001) and for lumen diameter for the Protégé and Precise photon-counting detector (PCD) CT [6].
stents (p < 0.01). The lumen attenuation was more accurate
23 with PCD-CT as well (p < 0.01), excluding the comparisons
24 between the Optima 580 and PCD-CT for the Protégé and PCDs are direct-conversion x-ray detectors, which are able
25 Precise stents. The PCD-CT system was better able to delineate to distinguish the energy of incident x-ray and bin them within
26 stents, specifically strut thickness. The stents were more easily specific energy ranges. As PCDs have been developed which
27 distinguished in PCD-CT images and in 3D volume renderings
than the clinical systems. The tantalum radiopaque markers were are able to handle higher flux rates, research has been on-
28 clearly visible in K-edge images due to reduced metal artifacts. going to investigate replacing conventional energy-integrating
29 detectors (EIDs) in conventional CT scanners to investigate the
30 Index Terms—Photon-counting detector CT, photon-counting
detectors, CT angiography, stent, K-edge subtraction imaging benefits PCDs could offer [7]–[11]. To date this has resulted
31 in one PCD system being approved for clinical use [12], with
32 others currently under development [13]–[15]. The benefits
33 I. I NTRODUCTION offered by PCDs include reduced metal and blooming artifacts
34 CT angiography (CTA) is a common imaging assessment as well as increased spatial resolution [10], [11], among
35 used to diagnose vascular diseases and monitor potential others, which could offer gains for CTA for imaging stents. A
36 complications before and after treatment [1]. Vascular issues number of previous studies have investigated the comparison
37 such as stenoses can often be treated with the implantation between PCD-CT and EID-CT for stent imaging including
38 of a stent, which can prevent the blood vessel from nar- lumen visualization, lumen attenuation, lumen diameter, and
39 rowing. However, monitoring of the area in which the stent the apparent size of stent struts [16]–[22]. PCD-CT also
40 is implanted can be hindered due to the stent itself. Stents offers the potential for K-edge imaging of high-atomic number
41 are usually manufactured from various metals, which cause contrast agents [23]–[25], such as iodine and gadolinium,
42 metal artifacts [2], [3], especially blooming artifacts, due to as well stents containing proportions of high-atomic number
43 their high atomic number and density. These artifacts can elements [19]. Additionally, the first comparisons with in-
44 reduce the diagnostic value of CTA scans [4] and need to human data of the two modalities have been published [5],
45 [26].
46 This work did not involve human subjects or animals in its research.
This work was partly funded by NSERC Alliance and Engage Plus grants,
47 NSERC CGS-D, an NSERC Discovery grant, the Canada Foundation for
48 In this study we evaluate the performance of a prototype
Innovation, the British Columbia Knowledge Development Fund, and the
49 Canada Research Chair program. bench-top PCD-CT system against two conventional clinical
50 CT systems by imaging several larger (> 5mm diameter)
Devon Richtsmeier is with the Department of Physics and Astronomy at
the University of Victoria, Victoria, BC, V8P 5C2, Canada. Corresponding
carotid artery stents. Though the stents are larger than those
51 author (email: [email protected]).
52 that are currently the most difficult to image, it provides a
Pierre-Antoine Rodesch is with the Department of Physics and Astronomy
53 baseline comparison between our bench-top PCD-CT system
at the University of Victoria, Victoria, BC, V8P 5C2, Canada (email: pier-
[email protected]). and every-day clinical systems. We evaluate the apparent
54 Kris Iniewski is with Redlen Technologies, Saanichton, BC, V8M 1X6,
55 Canada (email: [email protected]). lumen diameter, lumen attenuation and noise, and visualize
56 and quantify the size of radiopaque markers with K-edge
William Siu is with Fraser Health Authority, New Westminster, BC, V3L
3W7, Canada (email: [email protected]). subtraction imaging. We also demonstrate that our bench-top
57 Magdalena Bazalova-Carter is with the Department of Physics and Astron-
58 system provides the smallest measured apparent strut thickness
omy at the University of Victoria, Victoria, BC, V8P 5C2, Canada (email:
59 [email protected]). compared with previous studies.
60 https://mc.manuscriptcentral.com/trpms
Page 9 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 2

1
2 II. M ATERIALS AND M ETHODS
µ − µw
3 A. Imaging Systems HU = 1000 ∗ ( ) (1)
µw
4
Three different CT systems were used to image stents in in which µ is the attenuation values in the un-normalized
5
this study: a GE Optima 580 radiation therapy simulator, image and µw is mean signal within the water-containing
6
a GE Discovery IQ scanner (both GE Healthcare, Chicago, Eppendorf tubes within the un-normalized image. PCD-CT
7
IL), and a prototype bench-top PCD-CT scanner. All three and HR PCD-CT were reconstructed from the same projection
8
scanners can be seen in Fig. 1A–C. The bench-top PCD-CT data set. PCD-CT images were reconstructed in order to mimic
9
system consisted of a Comet MXR 160/22 x-ray tube (Comet the reconstructed clinical CT images as closely as possible in
10
Technologies, San Jose, CA), rotation and motion stages terms of pixel size and slice thickness, while HR PCD-CT
11
(Newport Corporation, Irving, CA), and a state-of-the-art flat offered the standard in-house reconstruction parameters.
12
panel PCD (Redlen Technologies, Saanichton, BC, Canada).
13
The source to isocenter distance was set to 322 mm with a
14 E. K-edge Image Reconstruction
source to detector distance of 578 mm. The PCD consisted of 2
15 K-edge subtraction images of tantalum were reconstructed
mm thick cadmium zinc telluride (CZT) crystal with a 330µm
16
pixel pitch and an active area of 8×190 mm2 , which gives a using the K-edge decomposition algorithm (KDA) described
17 by Zhang et al [25]. Briefly, Ta-specific sinograms were
106 mm field of view with 4.5 mm Z-coverage at isocenter
18 acquired according to the following equation:
[27]. The PCD is capable of the energy discrimination with
19
binning of up to six energy bins and can operate without
20 µ̂bg,U · TL − µ̂bg,L · TU
polarization at counts rates up to 650 Mcps/mm2 [28], [29]. DK (r) = (2)
21 µ̂K,L · µ̂bg,U − µ̂K,U · µ̂bg,L
22
where DK refers to the material-specific sinogram, µ̂bg is
23 B. Stents and Imaging Phantom
average the mass attenuation coefficient of the background,
24 Three nitinol stents were imaged in this study: a Medtronic µ̂K is the average mass attenuation coefficient of the K-edge
25 Protégé stent (Medtronic plc, Minneapolis, MN), a Cordis material, TL is the sinogram of the lower bin, and TU is the
26 Precise stent, and a Cordis S.M.A.R.T. Control stent (both sinogram of the upper bin. The subscripts U and L refer to
27 Cordis, Santa Clara, CA). Strut thicknesses were measured the upper and lower bin, respectively. So µ̂bg,L would be the
28 using a microscope with a 4X objective and found to be average mass attenuation coefficient of the background within
29 0.186 mm, 0.238 mm, and 0.177 mm, respectively. All other the energy range of the lower bin. To reconstruct the Ta-
30 physical dimensions were measured with calipers. The lumen specific sinogram the 67-81 keV and the 81-97 keV ranges
31 diameters of the stents were 5.82 mm, 5.71 mm, and 5.84 were used for the lower and upper bins, respectively. Once
32 mm, respectively. The two Cordis stents also had tantalum the Ta-specific sinogram was found, the K-edge images were
33 radiopaque markers. All stents were inserted into plastic straws reconstructed with the FDK algorithm and a Shepp-Logan
34 to mimic being compressed in vivo, inserted into a custom 100- filter.
35 mm diameter, 30-mm thick high-density polyethylene phantom
36 (Fig. 1D, E), and filled with water. In addition to the stents,
37 the phantom held 6.3 mm and 5.5 mm diameter Eppendorf F. Dose
38 tubes filled with water. CTDIvol (CT Dose Index) values were collected from the
39 clinical CT scanners and measured on the PCD-CT system
40 with the CTDI head phantom and 100-mm CTDI probe.
C. Data Acquisition
41
42 Acquisition and reconstruction parameters for all systems
G. Conventional CT Image Analysis
43 can be found in Table I. PCD-CT and HR PCD-CT were
44 reconstructed from the same data set, with differing recon- CT images were analyzed in a number of ways. For both
45 struction parameters. For the PCD-CT acquisitions, the energyPCD-CT resolution image types, the 35-120 keV CT images
46 thresholds were set to 35, 52, 67, 81, 95, and 120 keV. This were analyzed in order to provide the closest comparison to the
47 created energy bins with ranges between each threshold pair, clinical images. First, 3-dimensional (3D) volume renderings
48 with the addition of a bin summing counts from all other of the stents were created in 3D Slicer and segmented using
49 energy ranges, 35-120 keV. a lower threshold of 1750 HU. Second, the lumen attenuation
50 was determined. A circular region-of-interest (ROI) between
51 3.5 and 3.7 mm in diameter was delineated within the stent
D. Image Reconstruction and the mean and variance within the ROI was calculated for
52
53 Clinical CT images were reconstructed using filtered back a minimum of seven slices. The lumen diameter and the strut
54 projection with the BONEPLUS kernel. PCD-CT images thickness were evaluated by measuring their apparent values in
55 were reconstructed bin-wise using the Feldkamp-David-Kress the CT images. The center of the stent was first found using
56 algorithm [30] with a Shepp-Logan filter implemented in the the Hough Gradient method [32], and fine-tuned manually.
57 TIGRE package for Python [31]. Once reconstructed, the Line profiles were then interpolated from the center through
58 PCD-CT images were normalized bin-wise to Hounsfield units individual struts and the full-width-half-maximum (FWHM)
59 (HU) utilizing Eq. 1: of the peak formed from the strut’s signal was found. The
60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 10 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 3

1
2 TABLE I
CT IMAGE ACQUISITION AND RECONSTRUCTION PARAMETERS .
3
4 Parameter GE Optima GE PCD-CT HR
5 580 Discovery PCD-CT
IQ
6 Collimation 16 × 16 × 24 × 24 ×
7 0.625mm 0.625mm 0.184mm 0.184mm
8 Tube Voltage 120 120 120 120
(kV)
9 Tube Current 180 180 1 1
10 (mA)
11 Rotation Time 1.0 1.0 180 180
12 (s)
Helical Pitch 0.5625 0.5626 0.5626 0.5626
13 Beam Filter HEAD HEAD 6 mm Al 6 mm Al
14 FILTER* FILTER*
15 Focal Spot Size 0.7 0.7 ∼0.4** ∼0.4**
(mm)
16 Source to De- 1063 949 578 578
17 tector Distance
18 Fig. 1. Setup images of the (A) GE Optima 580, (B) GE Discovery IQ, and (mm)
(C) the bench-top PCD-CT system. (D) Phantom layout. (E) Image the Cordis Source to Axis 606 541 322 322
19 Precise stent inside the straw. Distance (mm)
20 Reconstruction BONEPLUS BONEPLUS Shepp- Shepp-
21 Filter/Kernel *** *** Logan Logan
Reconstructed 15.3 12.7 10.5 10.5
22 lumen radius was calculated as the mean distance from the
FOV (cm)
23 center of the stent to the closer FWHM point. The strut Reconstructed 0.299 0.248 0.248 0.205
24 thickness was calculated as the FWHM of the peak itself. Pixel size (mm)
25 Finally, the modulation transfer function (MTF) was calculated Slice Thickness 0.625 0.625 0.625 0.208
(mm)
26 in all image sets based on the method detailed by Takenaga et MTF 10% 1.16 1.16 1.29 1.33
27 al [33]. Briefly, the center and radius of the 100 mm diameter (lp/mm)
28 phantom was found using the Hough method and fine-tuned *HEAD FILTER is the filtration used from head scans in GE CT systems.
**The PCD-CT x-ray tube focal spot was measured using the EN 12543
29 manually. A normalized image was then calculated based on
convention, the approximate IEC 60336 size is shown.
30 the equation: ***BONEPLUS is a GE proprietary reconstruction kernel.
31 PCD-CT = photon-counting detector CT
I(i, j) − IBG HR PCD-CT = high-resolution PCD-CT
32 IN (i, j) = (3) MTF = modulation transfer function
33 Iphantom − IBG FOV = field of view
34 in which IN is the normalized image, I is the original
35 image, i and j are the row and column numbers for individual
36 The size of each marker was measured as the FWHM of the
pixels, IBG is the mean signal outside of the phantom, and
37 profile, in the same manner as the width of the individual
Iphantom is the mean signal within the phantom. The signal
38 struts in the conventional CT images. The diameter of the Ta
of each pixel within 5 mm inside and outside of the edge of
39 markers were measured using calipers at ∼0.75 mm, though
the phantom was plotted against its distance from the center
40 the tantalum is pressed into the nitinol disks that make up the
of the circle, creating an over-sampled edge spread function
41 marker structure, so that actual size of the tantalum itself will
(ESF). The over-sampled ESF was fitted using least-squares
42 be somewhat smaller.
to the equation:
43
a
44 ESF (x) = +d (4) I. Statistical Analysis
45 1 + exp[−b(x − c)]
Normality of the quantitative variables were assessed via the
46 where x is the distance from the center of the circle. a, b, c, Shapiro-Wilk test. All data, excluding the Ta marker data, were
47 and d were initially set to 1, -1, rp hantom, and 0, respectively. found to be non-normal and thus displayed as the median and
48 rp hantom is the radius of the phantom. The ESF was then interquartile range (IQR). The Wilcoxon rank-sum test was
49 differentiated to find the line spread function (LSF), and the used to compare the variables. Given the Ta marker data had
50 Fast Fourier Transform applied to the LSF to find the MTF. limited data points, it was reported as the mean and minimum
51 and maximum values.
52 H. K-edge Image Analysis
53 III. R ESULTS
54 The Ta marker size for the Cordis Precise and S.M.A.R.T.
55 Control stents was compared between clinical system images, A. Qualitative Evaluation
56 the two PCD-CT method images, and the two resulting K- Fig. 2 demonstrates representative CT slices from the GE
57 edge images from both PCD-CT methods. The size of the Optima 580 (Fig. 2A, E, I), GE Discovery IQ (Fig. 2B, F, J),
58 markers was measured by drawing a single profile through PCD-CT (Fig. 2C, G, K), and HR PCD-CT (Fig. 2D, H, L) for
59 each Ta marker in the slice in which they appeared largest. the Medtronic Protégé, Cordis Precise, and Cordis S.M.A.R.T
60 https://mc.manuscriptcentral.com/trpms
Page 11 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 4

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 Fig. 3. 3D volume renderings of all three stents. Volume renderings created
16 from scans of (A–D) the Medtronic Protégé stent, (E–H) the Cordis Precise
17 stent, and (I–L) the Cordis S.M.A.R.T. Control stent with each of the four
imaging methods.
18
19
20 Fig. 2. CT images of the plastic phantom with close-ups of the stents. Scans
of (A–D) the Medtronic Protégé stent, (E–H) the Cordis Precise stent, and
21 (I–L) the Cordis S.M.A.R.T. Control stent with each of the four imaging
22 methods. W/L = [2100/450].
23
24
Control stents, respectively. The CT images for both PCD-
25
CT options were reconstructed using the 35-120 keV bin. The
26
lumen of the stents appears larger in PCD-CT and HR PCD-
27
CT images compared to both GE scanners, and the struts are
28
easily delineated in PCD-CT, whereas they are not as easily
29
separated in the GE CT images.
30 The 3D stent volume renderings created for all four image
31 sets are shown in Fig. 3. The most accurate representation
32 of the stents was produced by HR PCD-CT (Fig. 3D, H, L).
33 Individual struts and the greater pattern of the stents were
34 visible and easily recognizable. PCD-CT (Fig. 3C, G, K)
35 offered less accurate representations, although individual struts
36 could be distinguished and the pattern could still be recognized
37 in some areas of the stents when comparing them to the HR
38 PCD-CT renderings. Neither of the clinical scanners offered
39 the same strut delineation or greater pattern recognizability
40 (Fig. 3A–B, E–F, I–J) and appeared largely as solid cylinders
41 with the occasional window into the interior lumen. Fig. 4. K-edge subtraction of radiopaque markers of the Cordis S.M.A.R.T.
42 In addition to better visualization of the stent structure in Control stent. CT image acquired with the (A) GE Discovery IQ and (B)
43 PCD-CT compared to clinical CT, radiopaque markers can also PCD-CT. (C) Ta K-edge decomposition image. (D) PCD-CT image with the
44 be better defined through the use of the energy information that
Ta K-edge image overlaid on top of it. W/L = [2100/450].
45 is inherent to PCD-CT. Fig. 4 shows the radiopaque markers
46 on the Cordis S.M.A.R.T Control stent in images from both
47 and HR PCD-CT demonstrated strut thicknesses closer to
the GE Discovery IQ (Fig. 4A) and PCD-CT (Fig. 4B). Metal the true thicknesses than either clinical scanner (p < 0.001)
48 and blooming artifacts can be seen in both, although they are(Fig. 5A–C). PCD-CT and HR PCD-CT also demonstrated
49 more prevalent in the PCD-CT image than in the HR PCD- values closer to the true value for the lumen diameter than
50 CT image. Using K-edge subtraction imaging, a Ta-specific either clinical scanner for the Medtronic Protégé and Cordis
51 image could be created (Fig. 4C), which mitigated many of Precise stents (p < 0.01). The lumen diameter results were not
52 the metal and blooming artifacts. The Ta-specific image couldsignificantly different for the Cordis S.M.A.R.T. Control stent
53 be viewed alone or as an overlay to the PCD-CT image (Fig. when comparing the clinical scanners to the PCD-CT system
54 4D) in order to better register the markers within the phantom.
(p > 0.01).
55
56 Table II shows the absolute difference of lumen attenuation
B. Quantitative Evaluation from water (0 HU). The median HU value fell closer to zero
57
58 Fig. 5 shows boxplots of the median and IQR for strut with PCD-CT and HR PCD-CT than the clinical scanners (p
59 thickness and lumen diameter. For all three stents, PCD-CT < 0.01). The only comparisons between HU data that were
60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 12 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 5

1
2
3
4
5
6
7
8
9
10
11
12 Fig. 5. Measured stent dimensions. The shaded red bands indicate the true
dimensions. Orange lines indicate median values, and the boxes indicate
13 interquartile range. (A–C) The measured strut thickness for the (A) Medtronic
14 Protégé, (B) Cordis Precise, and (C) Cordis S.M.A.R.T Control stents with
15 the four image sets
. (D–F) The measured lumen diameter for the (D) Medtronic
16 Protégé, (E) Cordis Precise, and (F) Cordis S.M.A.R.T Control
17 stents with the four image sets.
18
19
TABLE II
20 A BSOLUTE DIFFERENCE OF LUMEN ATTENUATION .
21
Medtronic Cordis Precise Cordis
22 Protégé S.M.A.R.T.
23 Control
24 (median; (median; (median;
IQR) IQR) IQR) Fig. 6. Measured tantalum diameter for (A) the Cordis Precise stent and (B)
25 Optima 580 -14 HU; -10 HU; -5 HU; the Cordis S.M.A.R.T. Control stent.
26 -61–41HU -66–44 HU -66–43 HU
27 Discovery -15 HU; -23 HU; -22 HU;
28 IQ -56–25 HU -71–21 HU -64–21 HU PCD-CT and HR PCD-CT for all stents. To our knowledge,
PCD-CT -4 HU; -5 HU; 4 HU; we also demonstrated the smallest measured apparent strut
29 -63–64 HU -86–88 HU -59–63 HU
30 HR 2 HU; 5 HU; 12 HU; thickness. The significant improvement in the measurement of
31 PCD-CT -52–65 HU -82–104 HU -49–79 HU the strut thickness with PCD-CT can be largely attributed to
Lumen attenuation is presented as the median absolute difference
the higher inherent spatial resolution of PCDs compared with
32 and IQR from 0 HU, the expected mean attenuation of water.
33 the EIDs on the clinical scanners. The PCD has a detector
34 pixel pitch of 0.33 mm, which is likely smaller than the
35 not significantly different were between the GE Optima 580 EID detectors, allowing for less partial volume effects in the
36 and PCD-CT for the Protégé and Precise stents (p = 0.065 and projections. This allowed for reduced blooming artifacts in
37 p = 0.042, respectively). However, the IQRs for all the stents the reconstructions, even though the PCD-CT and clinical CT
38 were smaller with the clinical scanners than either PCD-CT scans were reconstructed with nearly identical voxel sizes,
39 method. which should ensure fewer differences in terms of partial
40 Fig. 6 shows the Ta marker size tangential to the edge of the volume effects in the reconstructed images. With HR PCD-CT,
41 circle made by the stent; perpendicular to the most prevalent which had a smaller voxel size compared to the other three
42 metal artifacts in the PCD-CT image (Fig. 4B). Both the image sets, there were even better results in terms of apparent
43 PCD-CT and HR PCD-CT demonstrate smaller diameters than strut thickness, demonstrating that partial volume effects do
44 the measured diameter of the nitinol and tantalum together, have some effect on the blooming artifacts. With smaller stents
45 while the clinical systems show larger values. The K-edge it would be crucial to reconstruct at a higher resolution. The
46 measurements are similar for the Precise stent, but smaller for two PCD-CT methods also benefited from a smaller focal spot
47 the S.M.A.R.T. Control. size, which would cause less focal spot blurring, as well as
48 The 10% value of the MTF for each system can be seen smaller source to axis and source to detector distances, leading
49 in Table I. The reported CTDIvol values for the Optima 580 to less object magnification over multiple pixels.
50 and the Discovery IQ were 28.60 mGy and 62.16 mGy, For lumen diameter (Fig. 5D-F), there was little variation in
51 respectively. The measured CTDIvol for the PCD-CT setup the measurements, due in part to the large lumen diameter. The
52 was 93.30 mGy. lumen edges were more visible in the PCD-CT and HR PCD-
53 CT images, though this was likely due to the BONEPLUS
54 kernel creating dark undershoots around the individual struts.
IV. D ISCUSSION Median lumen attenuation (Table II) was more accurate
55
56 We evaluated the performance of our bench-top PCD-CT for PCD-CT and HR PCD-CT compared to the two clinical
57 scanner with respect to two clinical EID-CT scanners for scanners, though again the differences were small. A more
58 stent imaging. Quantitatively, we found that the apparent accurate representation of what would be seen in vivo would
59 strut thickness was closest to the physical measurement using be if there was contrast within the stent lumen. This would
60 https://mc.manuscriptcentral.com/trpms
Page 13 of 14 IEEE Transactions on Radiation and Plasma Medical Sciences
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 6

1
2 make the stent structures less visible as the attenuation would spots which were larger than the PCD-CT system’s focal
3 be higher within the lumen and could have some effect on the spot and had larger source to detector and source to axis
4 noise within the lumen. However, in this simpler case, we were distances. However, the PCD-CT system is not equipped with
5 also concerned with the accuracy of the signal to the expected an anti-scatter grid, which would serve to limit the noise in
6 value of 0 HU as well as the noise that was inherent to the reconstructed images and has a prototype detector, which is
7 stent. In respect to the noise, the IQR of the lumen attenuation still being developed. With that said, the bench-top PCD-
8 is larger for both PCD-CT methods, indicating higher image CT system outperformed the current, clinical EID scanners
9 noise within the lumen diameter. The higher noise can be in terms measuring apparent strut thickness.
10 attributed mostly to the high degree of non-conforming pixels In regard to the future clinical possibilities with this PCD
11 whose signal deviations are not fully mitigated by flat-field technology, the prospects are promising. The superior colli-
12 corrections. We also notice that these pixels, especially those mation and resolution of the PCD-CT methods is due almost
13 that create ring artifacts, can lead to additional metal artifact exclusively to the detector itself. Unlike with an EID, PCDs
14 streaks between the metal and artificially high signal areas require no septa between pixels in the detector; there is
15 within the ring artifacts [27]. And we see residual metal artifact relatively little spread of the electrons produced in the PCD
16 streaks within the PCD-CT images. For HR PCD-CT, the crystal when compared with the potential light spread in EIDs.
17 increase in noise over PCD-CT was also due to the smaller The minimal spread of electrons is due to the high bias
18 slice thickness, which resulted in reduced counts per voxel, and voltage applied and the fact that the anodes are pixelated. The
19 thus higher noise. The dose we saw with PCD-CT was also collimation in the PCD-CT methods is due to the size of the
20 higher than either clinical scanner, in fact it was three times as individual rows of the detector and the geometry of the setup.
21 high as the dose recorded by the Optima 580. This is due in Given that the detector specifications should remain the same
22 large part to the source to axis distance differences between when implemented in a clinical system, the collimation would
23 the scanners. If the dose was scaled accordingly, we would be approximately 24×0.188 mm with a geometry comparable
24 record CTDIvol values of 90.67 mGy and 175.47 mGy for the to the clinical systems studied here. Clinical results could even
25 Optima 580 and Discovery IQ, respectively, which compare be further improved with the use of proprietary kernels and
26 well to the 93.30 mGy for the PCD-CT system. other, more sophisticated reconstruction methods.
27 Qualitatively, the best results were seen with HR PCD-CT,
28 which offered the best 3D volume rendering of the stents and
V. C ONCLUSION
29 subtly reduced blooming artifacts when compared to PCD-
30 CT. However, both PCD-CT methods outperformed the EID- PCD-CT offered better stent visualization than current clin-
31 CT scanners, offering much better delineation of individual ical CT scanners including more accurate quantification of
32 struts in CT images as well as better 3D volume rendering of apparent strut thickness due to the increased spatial resolution
33 the stents. In addition, PCD-CT offers the ability to visualize offered by photon-counting detectors. Additionally, the energy
34 the radiopaque markers with K-edge subtraction imaging, information of PCDs could be used to visualize radiopaque
35 allowing the marker location and shape to be more accurately markers without metal artifacts in tantalum-specific K-edge
36 determined (Fig. 4C, D). The measured Ta sizes showed images.
37 the best results with the K-edge images, as they mitigated
38 the majority of the metal artifacts seen in the conventional ACKNOWLEDGMENT
39 images (Fig. 4A, B). A large difference was seen between
the sizes measured in the K-edge images compared to the The authors would like to thank Wilson Lo and Clay
40
41 clinical systems, while less of a difference was seen between Lindsay for their assistance in collecting the clinical data,
42 the the K-edge and conventional PCD-CT images. The main along with Mina Bechai and the staff at Initio Medical, as
43 reason for this being that the thickness of the Ta markers was well as Kyle Bromma for acquiring microscope images of the
44 measured along the short axis of the markers (Fig. 4C) in order stents. All authors declare that they have no known conflicts of
45 to avoid the excessive metal artifacts which occurred along the interest in terms of competing financial interests or personal
46 longer axis in the conventional PCD-CT images (Fig. 4B). One relationships that could have an influence or are relevant to
47 benefit of Ta-specific images for clinical imaging would be the the work reported in this paper.
48 reduction of metal artifacts around the markers, leading to less
49 obfuscation of the stent lumen. R EFERENCES
50 There are a number of limitations to our study. First, neither
clinical scanner would be utilized for coronary CT angiog- [1] K. K. Kumamaru, B. E. Hoppel, R. T. Mather, and F. J. Rybicki, “CT
51 Angiography: Current Technology and Clinical Use,” Radiol. Clin. N.
52 raphy as they would not supply the necessary resolution to Am., vol. 48, no. 2, pp. 213–235, Mar 2010.
53 image the small stents utilized in coronary arteries. However, [2] W. A. Kalender, R. Hebel, and J. Ebersberger, “Reduction of CT artifacts
the stents used in this study are carotid artery stents and were caused by metallic implants,” Radiology, vol. 164, no. 2, pp. 576–577,
54 Aug 1987.
55 used mainly as a comparison to evaluate the performance [3] F. E. Boas and D. Fleischmann, “CT artifacts: Causes and reduction
56 of our prototype bench-top PCD-CT system. The clinical techniques,” Imaging Med., vol. 4, no. 2, pp. 229–240, 2012.
scanners also used the BONEPLUS kernel which had some [4] O. Ghekiere, R. Salgado, N. Buls, T. Leiner, I. Mancini, P. Vanhoenacker,
57 P. Dendale, and A. Nchimi, “Image quality in coronary CT angiography:
58 dark undershoots are the stent struts, which affected the lumen Challenges and technical solutions,” Brit. J. Radiol., vol. 90, no. 1072,
59 visibility in images. Both clinical scanners had x-ray focal 2017.

60 https://mc.manuscriptcentral.com/trpms
 IEEE Transactions on Radiation and Plasma Medical Sciences Page 14 of 14
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, JANUARY 2023 7

1
2 [5] S. Boccalini, S. A. Si-Mohamed, H. Lacombe, A. Diaw, M. Varasteh, Coronary Stent Imaging: In Vitro Evaluation of 28 Coronary Stents,”
P. A. Rodesch, M. Villien, M. Sigovan, R. Dessouky, P. Coulon, Invest. Radiol., vol. 56, no. 10, pp. 653–660, Oct 2021.
3 Y. Yagil, E. Lahoud, K. Erhard, G. Rioufol, G. Finet, E. Bonnefoy- [23] D. Richtsmeier, C. A. S. Dunning, K. Iniewski, and M. Bazalova-Carter,
4 Cudraz, C. Bergerot, L. Boussel, and P. C. Douek, “First In-Human “Multi-contrast K-edge imaging on a bench-top photon-counting CT
5 Results of Computed Tomography Angiography for Coronary Stent system: acquisition parameter study,” J. Instrum., vol. 15, no. 10, Oct
Assessment with a Spectral Photon Counting Computed Tomography,” 2020.
6 Invest. Radiol., vol. 57, no. 4, pp. 212–221, Aug 2022. [24] C. A. S. Dunning, J. O’Connell, S. M. Robinson, K. J. Murphy, A. L.
7 [6] A. C. Kwan, A. Pourmorteza, D. Stutman, D. A. Bluemke, and J. A. Frencken, F. C. J. M. van Veggel, K. Iniewski, and M. Bazalova-Carter,
8 Lima, “Next-generation hardware advances in CT: Cardiac applications,” “Photon-counting computed tomography of lanthanide contrast agents
Radiology, vol. 298, no. 1, pp. 3–17, Nov 2021. with a high-flux 330-µm-pitch cadmium zinc telluride detector in a
9 [7] S. Leng, M. Bruesewitz, S. Tao, K. Rajendran, A. F. Halaweish, N. G. table-top system,” J. Med. Imaging, vol. 7, no. 03, p. 1, Jun 2020.
10 Campeau, J. G. Fletcher, and C. H. McCollough, “Photon-counting [25] Z. Zhang, J. Hu, X. Zhang, Q. Xu, M. Li, C. Wei, L. Wei, and Z. Wang,
11 Detector CT: System Design and Clinical Applications of an Emerging “Experimental research of the energy bins for K-edge imaging using a
Technology,” RadioGraphics, vol. 39, no. 3, pp. 729–743, May 2019. photon counting detector: a phantom and mice study,” Radiat. Detect.
12 [8] M. J. Willemink, M. Persson, A. Pourmorteza, N. J. Pelc, and Technol. Methods, no. 0123456789, 2020.
13 D. Fleischmann, “Photon-counting CT: Technical principles and clinical [26] S. A. Si-Mohamed, S. Boccalini, H. Lacombe, A. Diaw, M. Varasteh,
14 prospects,” Radiology, vol. 289, no. 2, pp. 293–312, Nov 2018. P. A. Rodesch, R. Dessouky, M. Villien, V. Tatard-Leitman, T. Bochaton,
[9] T. Flohr, M. Petersilka, A. Henning, S. Ulzheimer, J. Ferda, and P. Coulon, Y. Yagil, E. Lahoud, K. Erhard, B. Riche, E. Bonnefoy,
15 B. Schmidt, “Photon-counting CT review,” Phys. Med., vol. 79, no. G. Rioufol, G. Finet, C. Bergerot, L. Boussel, J. Greffier, and P. C.
16 October, pp. 126–136, Nov 2020. Douek, “Coronary CT Angiography with Photon-counting CT: First-In-
17 [10] A. S. Wang and N. J. Pelc, “Spectral Photon Counting CT: Imaging Human Results,” Radiology, vol. 303, no. 2, pp. 303–313, May 2022.
Algorithms and Performance Assessment,” EEE Trans. Radiat. Plasma [27] D. Richtsmeier, J. O’Connell, P. A. Rodesch, K. Iniewski, and
18 Med. Sci., vol. 5, no. 4, pp. 453–464, Jul 2021. M. Bazalova-Carter, “Metal artifact correction in photon-counting de-
19 [11] S. S. Hsieh, S. Leng, K. Rajendran, S. Tao, and C. H. McCollough, tector computed tomography: metal trace replacement using high-energy
20 “Photon Counting CT: Clinical Applications and Future Developments,” data,” Med. Phys., Sep 2022.
EEE Trans. Radiat. Plasma Med. Sci., vol. 5, no. 4, pp. 441–452, Jul [28] K. Iniewski, C. Hansson, E. Guliyev, G. Prekas, and M. Ayukawa,
21 2021. “Performance Characteristics of 250+ Mcps/mm2 CZT Detector Module
22 [12] K. Rajendran, M. Petersilka, A. Henning, E. R. Shanblatt, B. Schmidt, for Spectral Computed Tomography,” in 5th Workshop on Medical
23 T. G. Flohr, A. Ferrero, F. Baffour, F. E. Diehn, L. Yu, P. Rajiah, Applications of Spectroscopic X-ray Detectors, CERN, Geneva., 2019a.
J. G. Fletcher, S. Leng, and C. H. McCollough, “First Clinical Photon- [29] K. Iniewski, M. C. Veale, and M. Bazalova-Carter, “High-flux CZT for
24 counting Detector CT System: Technical Evaluation,” Radiology, vol. new Frontiers in computed tomography (CT), non-destructive testing
25 303, no. 1, pp. 130–138, Apr 2022. (NDT) and high-energy physics,” in IEEE Nuclear Sci. Symp. and Med.
26 [13] J. da Silva, F. Grönberg, B. Cederström, M. Persson, M. Sjölin, Imaging Conf, 2019b.
Z. Alagic, R. Bujila, and M. Danielsson, “Resolution characterization [30] L. A. Feldkamp, L. C. Davis, and J. W. Kress, “Practical cone-beam
27 of a silicon-based, photon-counting computed tomography prototype algorithm,” J. Opt. Soc. Am. A., vol. 1, no. 6, pp. 612–619, Feb 1984.
28 capable of patient scanning,” J. Med. Imaging, vol. 6, no. 04, p. 1, [31] A. Biguri, M. Dosanjh, S. Hancock, and M. Soleimani, “TIGRE: A
29 Oct 2019. MATLAB-GPU toolbox for CBCT image reconstruction,” Biomed. Phys.
[14] X. Zhan, R. Zhang, X. Niu, I. Hein, B. Budden, S. Wu, N. Markov, Eng. Express., vol. 2, no. 5, p. 055010, Sep 2016.
30 C. Clarke, Y. Qiang, H. Taguchi, K. Nomura, Y. Muramatsu, Z. Yu, [32] H. Yuen, J. Princen, J. Illingworth, and J. Kittler, “Comparative study
31 T. Kobayashi, R. Thompson, H. Miyazaki, and H. Nakai, “Compre- of Hough Transform methods for circle finding,” Image Vision Comput.,
32 hensive evaluations of a prototype full field-of-view photon counting vol. 8, no. 1, pp. 71–77, 1990.
CT system through phantom studies,” Dec 2022. [Online]. Available: [33] T. Takenaga, S. Katsuragawa, M. Goto, M. Hatemura, Y. Uchiyama, and
33 https://arxiv.org/abs/2212.13337v1http://arxiv.org/abs/2212.13337 J. Shiraishi, “Modulation transfer function measurement of CT images
34 [15] S. A. Si-Mohamed, S. Boccalini, M. Villien, Y. Yagil, K. Erhard, by use of a circular edge method with a logistic curve-fitting technique,”
35 L. Boussel, and P. C. Douek, “First Experience With a Whole-Body Radiol. Phys. Technol., vol. 8, no. 1, pp. 53–59, Jan 2015.
Spectral Photon-Counting CT Clinical Prototype,” Invest. Radiol., vol.
36 Publish Ah, no. 00, Feb 2023.
37 [16] M. Mannil, T. Hickethier, J. Von Spiczak, M. Baer, A. Henning,
38 M. Hertel, B. Schmidt, T. Flohr, D. Maintz, and H. Alkadhi, “Photon-
Counting CT: High-Resolution Imaging of Coronary Stents,” Invest.
39 Radiol., vol. 53, no. 3, pp. 143–149, Mar 2018.
40 [17] J. Von Spiczak, M. Mannil, B. Peters, T. Hickethier, M. Baer, A. Hen-
41 ning, B. Schmidt, T. Flohr, R. Manka, D. Maintz, and H. Alkadhi, “Pho-
ton Counting Computed Tomography with Dedicated Sharp Convolution
42 Kernels: Tapping the Potential of a New Technology for Stent Imaging,”
43 Invest. Radiol., vol. 53, no. 8, pp. 486–494, Aug 2018.
[18] R. Symons, Y. De Bruecker, J. Roosen, L. Van Camp, T. E. Cork,
44 S. Kappler, S. Ulzheimer, V. Sandfort, D. A. Bluemke, and A. Pour-
45 morteza, “Quarter-millimeter spectral coronary stent imaging with
46 photon-counting CT: Initial experience,” J. Cardiovasc. Comput., vol. 12,
no. 6, pp. 509–515, Oct 2018.
47 [19] M. Sigovan, S. Si-Mohamed, D. Bar-Ness, J. Mitchell, J. B. Langlois,
48 P. Coulon, E. Roessl, I. Blevis, M. Rokni, G. Rioufol, P. Douek, and
49 L. Boussel, “Feasibility of improving vascular imaging in the presence of
metallic stents using spectral photon counting CT and K-edge imaging,”
50 Sci. Rep., vol. 9, no. 1, pp. 1–9, Dec 2019.
51 [20] G. Bratke, T. Hickethier, D. Bar-Ness, A. C. Bunck, D. Maintz, G. Pahn,
52 P. Coulon, S. Si-Mohamed, P. Douek, and M. Sigovan, “Spectral
Photon-Counting Computed Tomography for Coronary Stent Imaging:
53 Evaluation of the Potential Clinical Impact for the Delineation of In-
54 Stent Restenosis,” Invest. Radiol., vol. 55, no. 2, pp. 61–67, Feb 2020.
55 [21] J. R. Rajagopal, F. Farhadi, T. Richards, M. Nikpanah, and M. Y. Chen,
“Evaluation of Coronary Plaques and Stents with Conventional and
56 Photon-counting CT : Benefits of High-Resolution Photon-counting CT,”
57 Radiol. Cardiothorac. Imaging, vol. 3, no. 5, Oct 2021.
58 [22] B. Petritsch, N. Petri, A. M. Weng, M. Petersilka, T. Allmendinger, T. A.
Bley, and T. Gassenmaier, “Photon-Counting Computed Tomography for
59
60 https://mc.manuscriptcentral.com/trpms