Introduction
Despite advances in treatment, atherosclerotic cardiovascular disease remains the main cause of mortality worldwide [
1]. For the determination of the prognosis of future cardiovascular disease, risk categories based on the amount of coronary calcium are increasingly used [
2]. Coronary calcium can be quantified with computed tomography (CT) as coronary calcium scores (CCS), including Agatston scores, mass scores and volume scores [
3]. Correct and reproducible CCS are essential considering their use in the guidelines [
1,
4]. If risk-based treatment decision is uncertain after quantitative risk assessment, the 2013 guidelines recommend repeated CCS measurements. Also, CCS progression in repeated CCS measurements has recently been associated with heart failure [
5].
Several studies have demonstrated the dependence of CCS on motion [
6‐
9]. Motion artifacts can result in incorrect Agatston scores, which may thereby lead to incorrect risk classification. Depending on heart rate and anatomical location, coronary arteries move at velocities of 10–30 mm/s during the acquisition phase, whereas coronary arteries are often erroneously assumed to be stationary during CT acquisition [
10‐
12].
Previously published studies on the influence of heart rate on CCS have only focused on CT systems from a single manufacturer [
9,
13,
14]. Recently, substantial differences in CCS between new generation CT systems were demonstrated. Willemink et al. found differences in Agatston scores between CT systems of up to 43.9% for static calcifications [
15]. However, the effect of heart rate on CCS for different state-of-the-art CT systems remains unknown. The aim of the current study was therefore to determine the influence of heart rate on CCS for the high-end CT systems from four major manufacturers at routinely used clinical protocols with a dynamic phantom.
Discussion
This dynamic phantom study showed that for high-end CT systems of four major CT manufacturers coronary calcium scores were substantially, but not equally, influenced (up to 50%) by heart rate. Motion susceptibility depends on both calcification mass and CT system. At increased heart rates, the stability of the mass score is superior to Agatston score stability.
The most important findings in this study were as follows. First, we found that no modern CT system is capable of completely mitigating the influence of heart rate on Agatston score or calcification mass. Second, CMS Index analysis showed significant differences for both Agatston and mass scores between four high-end CT systems. Also, different heart rates were found to differentially affect calcification masses for two CT systems. Thus, for these CT systems the influence of heart rate on CCS depends on calcification density.
For follow-up imaging a valid and precise scoring method is essential, therefore subsequent scans should be made with the same CT system and heart rates should be comparable. For low mass calcifications, voxels above the 130 HU threshold at low heart rate might fall below this threshold as a result of motion blurring. As a result of this effect, CCS will decrease for these calcifications. For medium and high mass calcifications, the number of voxels above the 130 HU threshold will increase as a result of motion blurring, thereby increasing CCS. The registered calcification size is smaller for CT systems with a higher temporal resolution, which is demonstrated in Fig.
4 for CT3 in comparison with the other CT systems. However, many other parameters, including acquisition and reconstruction parameters, also influence the registered calcification size [
18]. In addition, technical advances, including dual-energy CT in combination with mono-energetic reconstructions, may influence the registered calcification quantification. With these mono-energetic reconstructions, it has been shown that blooming artefacts can be reduced [
19]. Although especially for high density calcifications improved size registration is expected, the influence of these reconstructions on motion artefacts is unknown at this time. Third, the stability of the mass score is clearly superior to the stability of the Agatston score. This can be explained from the different way of calculating the mass score in comparison with the Agatston score. While the Agatston score considers the highest HU value in the lesion, the mass score can be thought of as a weighted average, which is therefore more robust to the effects of motion blurring. We found that the use of modern, state-of-the-art CT systems still results in substantial deviation from the true physical mass.
Our results are in line with a study conducted by Groen et al., who found decreased CCS for low mass calcifications and increased CCS for high mass calcifications at increased heart rates [
7]. However, this study focused on only one CT system, while we used four high-end CT systems in the current study. Our findings are in agreement with the studies of Greuter et al. [
6] and Groen et al. [
7], who also showed the superior stability of mass scores in comparison to Agatston scores. These studies, however, were only performed on CT systems from one manufacturer. Tigges et al. used explanted coronary arteries mounted on a cardiac motion phantom to examine the influence of heart rate on CCS [
8]. The explanted coronary arteries were subdivided into categories based on their respective Agatston score. The study showed a decrease in CCS with increasing heart rates, independent of calcification mass. For low mass calcifications these results correspond well to our results, whereas the high mass calcification results differ from our results. This difference in results probably originates from a difference in calcification density, because a high Agatston score can be the result of either a large calcification or a high density. The density of the calcification was not specified in the study of Tigges et al.
The current study has limitations. First, the artificial coronary arteries used in this study were translated at constant linear motion, whereas complex 3D motions are observed in-vivo. This 3D motion is patient and artery specific. Nevertheless, the scan times were relatively short as a result of fast rotation times, whereby the constant linear motion of our phantom was deemed sufficient as a model of the complex in vivo motion of coronary arteries. Second, linear velocity was converted to heart rates based on only one available study [
11]. This limitation was accepted because of large variation in heart rates within the chosen heart rate groups (< 60, 60–75 and > 75 bpm). Third, slice thickness and increment were not the same for all CT systems. Previous studies demonstrated increased CCS for decreased slice thicknesses [
20,
21]. However, the current study focusses on CCS from clinically used CCS protocols to evaluate the influence of heart rate within each CT system. Also, the deviating slice thickness and increment for CT1 in comparison with the other systems is in line with protocol recommendations as described previously by McCollough et al. [
16]. With the clinically used CCS protocols, CCS were acquired with routine CT manufacturers’ software. However, it was demonstrated by Weininger et al. that different types of CCS scoring software resulted in similar CCS [
22].
In conclusion, the current dynamic phantom study showed that heart rate substantially influences CCS for the high-end CT systems of the four major CT manufacturers. Therefore, this study suggests that heart rate should be taken into account when interpreting the clinical impact of patients’ calcium scores. Moreover, it is essential to acquire CCS scans on similar CT systems between follow-up studies, preferably at comparable heart rates. Also, mass scores were found to be more stable then Agatston scores.