Keywords
heart - CT-angiography - cardiac ct - dual-source ct - image quality - radiation dose
Introduction
Cardiac CT has become an established method for visualizing coronary vessels. The
diagnostic accuracy has been reported in meta-analyses with a sensitivity and specificity
of 95.2%/79.2% [1]
[2]. In the current “Nationale Versorgungsleitlinie Chronische KHK” (national care guidelines
for chronic coronary artery disease), cardiac CT is recommended as the method of choice
for patients with low to intermediate pretest probability [3]. It demonstrates identical accuracy compared to invasive coronary angiography but
with significantly fewer side effects and leads to reduced mortality in the long term
[4]
[5]. In Germany, the Bundesärztekammer (Federal Medical Association) regulates the minimum
requirements for device technology for various indications, including cardiac CT,
through guidelines for quality assurance in computed tomography [6]. Key parameters include coverage in the z-direction with a detector having at least
64 slices and a temporal resolution with a rotation time of ≤0.35s. This means that
only mid-range or higher-priced CT scanners qualify for cardiac CT applications.
For many outpatient practices, the question arises as to whether the additional cost
of a cardio-capable CT scanner is economically justifiable and what technical features
the CT scanner should have. The answer to the first question depends on local circumstances,
billing codes yet to be defined, quality assurance agreements, and respective regional
regulations. This publication aims to help answer the second question.
For more than ten years, cardiac CT scans have been performed at our institution using
a 128-slice single-source CT (SSCT) scanner with a rotation time of 0.3s. Annually,
over 300 cases have been examined, with most patients referred by specialized cardiologists.
In the context of a practice expansion, a new 128-slice dual-source CT (DSCT) scanner
with a rotation time of 0.33s was installed in the summer of this year. From the moment
of installation of the scanner, all cardiac CT patients have been exclusively examined
on the new DSCT scanner. Referring physicians, patient quality, staff, and contrast
media protocols have remained unchanged. This provided an opportunity to conduct a
retrospective analysis of image and metadata to make a robust assessment of the impact
of CT scanner technology on image quality and radiation dose in an outpatient setting.
These data can serve as guidelines for other providers of CT services when selecting
scanner technology.
Materials and Methods
CT Technology
Before the transition, a 128-slice single-source CT scanner (SOMATOM Definition AS+,
Siemens Healthineers, Forchheim, Germany) with a native temporal resolution of 150ms
(rotation time 0.3s) and a detector width of 38.4mm was used. After the transition,
a 128-slice dual-source CT scanner (SOMATOM Pro.Pulse, Siemens Healthineers) with
a native temporal resolution of 86ms (rotation time 0.33s) and a detector width of
38.4mm was employed. Further technical details of both CT scanners are listed in [Table 1].
Table 1 Comparison of the main image acquisition and reconstruction parameters for the SSCT
and DSCT scanners.
|
SSCT scanner
|
DSCT scanner
|
|
Number of detector rows
|
128
|
128
|
|
Number of X-ray tubes
|
1x Straton tube
|
2x Athlon tube
|
|
Rotation time (s)
|
0.3
|
0.33
|
|
Temporal resolution (ms)
|
150
|
86
|
|
Detector width z-axis (mm)
|
38.4
|
38.4
|
|
Recon slice thickness (mm)
|
0.75
|
0.8
|
|
Recon method
|
Iterative (SAFIRE)
|
Iterative (SAFIRE with ZeeFree)
|
|
Recon kernel
|
BV38
|
BV40
|
|
Contrast media protocol
|
50ml Imeron 400 @ 5ml/s
50ml Imeron 400/50ml Nacl @ 5ml/s
50ml NaCl @ 5ml/s
|
50ml Imeron 400 @ 5ml/s
50ml Imeron 400/50ml Nacl @ 5ml/s
50ml NaCl @ 5ml/s
|
|
Protocol selection (sequence/spiral)
|
Manual
|
Automatic
|
For both CT scanners, the manufacturer’s built-in dose modulation (Care-kV) was used
for CT angiography. Contrast agent timing was performed via bolus triggering in the
ascending aorta, with scan initiation occurring automatically upon reaching the threshold
of 140HU. On the SSCT scanner, CT angiography was acquired using prospectively triggered
sequences (up to a heart rate of 65 bpm). Above this threshold, retrospectively triggered
spiral acquisition was used. For the prospective sequence, a diastolic acquisition
window of 70–70% (low heart rate, stable) or 60–80% (low heart rate, instable) was
used. The spiral acquisition for higher heart rates used the 30–40% cardiac cycle
for the main dose. On the DSCT, program selection (sequence vs. spiral) was performed
automatically via built-in AI, which chose the optimal program based on heart rate
and rhythm. Typically, a sequence with an exposition window at 75–75% was used when
a low and rhythmic heart rate was detected. With faster or arrhythmic heart rates,
padding was added or a sequential systolic acquisition or if need be a pulsed spiral
with the main dose applied at 30–40% of the cardiac cycle was chosen by the scanner.
Contrast agent was administered using a MedRad Stellant injector pump (Bayer, Leverkusen,
Germany). A total of 50ml of Imeron 400 (Bracco, Milan, Italy) was followed by an
Imeron/NaCl mixture (50ml/50ml) and 50ml NaCl at 5ml/s, delivered through a peripheral
18G venous cannula.
Patients
The data for all cardiac CT patients examined within 80 days before and after the
commissioning of the new CT scanner were retrospectively included. Image data were
queried through the PACS (Phönix-PACS, Freiburg, Germany), which allows anonymized
queries and data display. Exclusion criteria included patients undergoing coronary
calcium measurement without CT angiography and patients with bypasses. In total, data
from 245 patients were included.
Image Analysis
From the CT angiography (CTA) protocols stored in the PACS, the dose-length product
(DLP) for the CTA and the kV used for the CTA were extracted. The heart rate at the
time of acquisition and the acquisition mode were also extracted. The image quality
of the coronary vessels was independently evaluated by two experienced readers (Q2
and Q3 Cardiac CT of the Deutsche Röntgengesellschaft (German Radiological Society))
on a vessel basis for the RCA, LM, LAD, and LCX using an ordinal 3-point scale: 3
– excellent, 2 – artifacts but diagnostic, 1 – artifacts rendering the vessel not
fully evaluable.
Statistical Analysis
Using an alpha level of 0.05 and a power of 0.8, the given sample size would allow
robust detection of small differences between both groups (Cohen’s d = 0.4). Data
were anonymized and stored in Excel (Microsoft, Redmond, WA). Statistical analysis
was performed using JASP (University of Amsterdam, Netherlands) and ChatGPT 4o (OpenAI,
San Francisco, USA). Means, standard deviations, and two-sided Student’s t-tests were
used for continuous data, while medians and Wilcoxon signed-ranks tests were employed
for ordinal data. The agreement between readers was calculated using Cohen’s kappa
coefficient.
Results
A total of 113 patients (32 females/80 males, average age 66.1 years) were examined
using the SSCT scanner, while 132 patients (43 females/89 males, average age 64.7
years) were examined using the DSCT scanner. The average age of patients and the female-to-male
ratio were not significantly different between the two CT scanners. The mean heart
rate was also not significantly different between the two scanners, averaging approximately
61–62 beats per minute. The variability in heart rate, which was partly very high
on both systems due to extrasystoles during acquisition, was identical for both CT
scanners. A detailed overview of the demographic data is provided in [Table 2].
Table 2 Patients’ demographic data and acquisition-related data comparison.
|
SSCT scanner
|
DSCT scanner
|
|
|
bpm – beats per minute
|
|
Number of patients
|
113
|
132
|
|
|
Gender (F/M)
|
33/80
|
43/89
|
|
|
Age range (a)
|
43–90
|
40–90
|
|
|
Mean age (a)
|
66.0±10.5
|
64.7±10.5
|
p=0.3 (n.s.)
|
|
Heart rate range (bpm)
|
38–123
|
41–130
|
|
|
Mean heart rate (bpm)
|
61.1±10.8
|
61.6±14.1
|
p=0.7 (n.s.)
|
|
Sequence/spiral
|
95/18 (84%/16%)
|
120/12 (91%/9%)
|
|
|
Mean CTA kV
|
101.4±12.1
|
74.2±7.5
|
p<0.0001
|
|
DLP range (mGy∙cm)
|
27–923
|
17–558
|
|
|
Mean DLP CTA (mGy∙cm)
|
134.2±138.4
|
87.7±65.9
|
p=0.001
|
The mean DLP for the DSCT scanner was 87.7 mGy·cm, which was 52% lower than for the
SSCT scanner at 133.6 mGy·cm (p=0.013). The DLP varied on both CT scanners depending
on patient constitution and examination method, ranging from 17 mGy·cm to 923 mGy·cm.
For the DSCT scanner, the DLP showed less variation across all heart rates ([Fig. 1]). Using a conversion factor of 0.014, the mean effective radiation dose was calculated
as 1.23 mSv for the DSCT scanner and 1.87 mSv for the SSCT scanner. An example of
image quality is shown in [Fig. 2].
Fig. 1
a DLP across different heart rates – SSCT scanner. b DLP across different heart rates – DSCT scanner. The distribution of DLP across different
heart rates demonstrates significantly lower variability with the DSCT scanner. Additionally,
DLP increases less steeply at higher heart rates on the DSCT scanner. The graph has
been truncated for clarity, omitting high DLP values above 500 mGy · cm and heart
rates above 80 bpm for both devices.
Fig. 2 Exemplary depiction of the right coronary artery and its branches. This is the patient
with the lowest dose of 0.24 mSv. The minimal differences in contrast, particularly
in the right ventricle, are also clearly visible due to the sequential acquisition,
without any visible steps in the RCA. The image quality was rated as very good (score
3) by both reviewers.
The automatic kV selection allowed for a significantly lower average tube voltage
on the DSCT scanner (74.2 kV) compared to the SSCT scanner (101.4 kV) (p<0.001). The
distribution of kV across individual examinations is shown in [Fig. 3]. The complete technical results are also provided in [Table 2].
Fig. 3 Representation of the kV distribution across different patients. In the case of the
SSCT scanner, most patients were examined with 100 kV, with a few cases using 120
kV or 80 kV. In the case of the DSCT scanner, the stronger X-ray tube allowed for
examinations predominantly at 70 kV and 80 kV.
In the subjective evaluation of vessel assessability by the two readers, diagnostic
image quality of the right coronary artery (RCA) was achieved for the SSCT scanner
in 37 of 113 cases and 42 of 113 cases (readers 1 and 2, respectively), as shown in
[Fig. 4]. Calculated as percentages, 33% and 37% of RCA examinations on the SSCT scanner
were non-diagnostic (readers 1 and 2, respectively). For the DSCT scanner, the number
of non-diagnostic RCAs was 8 and 5 of 132 cases (readers 1 and 2, respectively). In
percentages, this corresponds to 6% and 4% non-diagnostic RCA examinations with the
DSCT scanner.
Fig. 4 Comparable slices from an SSCT examination (a) and a DSCT examination (b). The LAD and distal RCA are clearly visible, but the mid-RCA in the sulcus is blurred
due to pulsation in the SSCT examination (arrow) and cannot be assessed, whereas it
is well-assessable in the DSCT examination (arrowhead). To better visualize the artifacts
in the sulcus, a thin MIP was used for the SSCT examination in this representation.
For the circumflex artery (LCX), the percentage of non-diagnostic examinations with
the SSCT scanner was 6% and 7% (readers 1 and 2, respectively), while for the DSCT,
it was 2% and 0% (readers 1 and 2, respectively). The percentage of non-diagnostic
left anterior descending artery (LAD) examinations was 4% with the SSCT scanner (readers
1 and 2, respectively) and 0% with the DSCT scanner (readers 1 and 2, respectively).
These differences were statistically significant: RCA p<0.0001, LCX p<0.0001, LAD
p=0.0004. No relevant differences were found for the left main artery (LM).
The agreement between readers was good, with a Cohen’s kappa of 0.71 (0.61–0.8 – good
agreement). A summary of the results of the subjective evaluation is presented in
[Table 3].
Table 3 Display of the results of the vessel assessment for both readers where 1 indicates
a non-diagnostic vessel/vessel segment, 2 indicates impaired but diagnostic image
quality, and 3 indicates excellent image quality.
|
|
Reader 1
|
Reader 2
|
|
Score
|
RCA
|
LM
|
LAD
|
LCX
|
RCA
|
LM
|
LAD
|
LCX
|
|
SSCT
|
1
|
37
|
2
|
5
|
7
|
42
|
2
|
4
|
8
|
|
2
|
31
|
8
|
39
|
24
|
31
|
8
|
33
|
22
|
|
3
|
45
|
103
|
69
|
82
|
40
|
103
|
76
|
82
|
|
DSCT
|
1
|
8
|
0
|
0
|
3
|
5
|
0
|
0
|
0
|
|
2
|
33
|
1
|
24
|
13
|
32
|
1
|
11
|
16
|
|
3
|
91
|
131
|
108
|
116
|
95
|
131
|
121
|
115
|
Discussion
This study provides two key insights: first, the improved technology of newer CT devices,
such as the dual-source CT (DSCT) scanner used in this study, leads to significantly
more frequent utilization of “low-kV” imaging, which is particularly dose-efficient.
As this method utilizes radiation closer to the k-edge of iodine, it achieves a high
arterial vessel contrast on average [7]. However, measuring contrast was outside the focus of this study.
Second, the data clearly show that increased temporal resolution, as achieved with
a DSCT scanner, results in a much more robust execution of cardiac CT in routine clinical
practice. Even with the new DSCT scanner, non-diagnostic cases were observed, but
the number of such cases was significantly lower compared to the single-source CT
(SSCT) scanner. In particular, the relatively frequent pulsation artifacts of the
RCA in the atrioventricular sulcus with the SSCT scanner were, while not completely
eliminated with the DSCT scanner, significantly less frequent and pronounced. This
resulted in a statistically significant improvement in the overall evaluability of
the examinations, making routine practice examinations considerably more robust.
Translated into a radiological report, this means that out of the 113 cardiac CT examinations
conducted on an SSCT scanner, depending on the reader, 37–42% would have to be rated
as CAD-RADS N because not all segments were reliably evaluable. With the DSCT scanner,
this percentage decreases to 4–6%. In practice, radiologists often resort to providing
a summary evaluation of visible segments. However, a secure and comprehensive assessment
of coronary vessels is desirable.
With the imminent rollout of cardiac CT as a service covered by statutory health insurance
in Germany, an increasing number of examinations can be expected. To ensure that the
method, which has so far been predominantly used in academic and university settings,
gains acceptance in the private cardiology sector, consistently high radiological
process quality and thus consistently good evaluability of coronary vessels are a
fundamental requirement.
The minimum technical standards for cardiac CT set forth by the Federal Joint Committee
(Gemeinsamer Bundesausschuss, GBA) and outlined in the Federal Medical Association’s
guidelines [6] would be fully met with the SSCT scanner used in this study. The SSCT scanner that
was used even had a slightly better rotation time (0.3s) than the legally required
minimum standard (≤0.35s). However, given the results presented in this study, the
technical minimum requirements mandated by law should be critically re-evaluated.
Of course, the successful performance of cardiac CT also requires proper patient premedication,
including beta blockers if necessary, unless contraindicated. This was and is routinely
done at our institution. However, not all patients can achieve the desired heart rate
of below 60 bpm, and some may exhibit higher heart rates or even extrasystoles during
acquisition due to the stress of the cardiac CT examination.
Patients with extrasystoles during acquisition or arrhythmias were also those who,
along with individuals with higher body weight and compact body structures, showed
significant outliers in radiation dose on both systems. The difference between the
lowest and highest doses on the DSCT scanner was nearly a factor of 33: DLP ranged
from 17 mGy·cm to 558 mGy·cm (0.24 mSv to 7.8 mSv). Similar differences were also
observed with the SSCT scanner.
On average, the effective radiation dose on both systems was significantly lower than
the previously published data, at 1.23 mSv (DSCT) and 1.87 mSv (SSCT). A recent review
article published in the German Medical Journal reported the dose for prospectively
triggered coronary CTA as 5.7 mSv [8]. The doses presented in this study are thus approximately 70–80% lower than published
values.
This is particularly important given the high visibility and health policy impact
of the German Medical Journal in Germany. However, it should be noted that the dose
values reported here do not include the topogram and the dose for bolus tracking.
Including these would increase the dose for both devices by approximately 0.2–0.5
mSv. For conventional coronary angiography, the reference value for a purely diagnostic
coronary angiography is listed at 2800 cGy·cm², or approximately 4.48 mSv [9]—thus about 2–3 times higher than the effective doses for cardiac CT presented here.
These positive data must be proactively communicated by radiologists to referring
colleagues to increase the acceptance of cardiac CT and highlight its added value
compared to diagnostic coronary angiography.
It is also notable that the use of automated software for program selection on the
CT scanner, as with the DSCT scanner, reduced the use of spiral acquisition from 16%
of cases to 9%. This automation can also lead to higher process quality and reduce
the workload for radiology technologists. It should also be mentioned that spiral
acquisition does not automatically imply a high radiation dose: spiral acquisitions
with DLPs as low as 73 mGy·cm (1.0 mSv) were found in this study for heart rates of
82 bpm. A systematic analysis of these cases, however, falls outside the scope of
this study.
In summary, this practical, retrospective analysis of 245 cardiac CT cases shows that
using the minimum technical requirements for cardiac CT examination (64-slice SSCT
scanner, 0.35s rotation time) leads to a significantly higher rate of non-diagnostic
findings compared to newer systems with higher temporal resolution. The radiation
dose required for cardiac CT examination is significantly lower with both systems
than in previously published data, with the DSCT scanner demonstrating nearly 34%
lower doses than the SSCT scanner.
Clinical Relevance
-
Switching from an SSCT scanner to a DSCT scanner in an outpatient setting results
in a significant dose reduction of approximately 34%.
-
The number of artifact-related non-diagnostic examinations is significantly higher
with an SSCT scanner (0.3s rotation time) compared to a DSCT scanner (0.33s rotation
time).
-
Meeting only the minimum standards for cardiac CT (64-slice SSCT scanner, 0.35s rotation
time) is likely to result in a similarly high rate of non-diagnostic examinations
as found in this study for the 128-slice SSCT scanner with a 0.3s rotation time.
-
Outpatient cardiac CTA can be performed with an average radiation dose well below
2 mSv using both SSCT and DSCT systems.
