Over the last few decades, the rapid development of cardiac computed tomography (CT) has been unequalled by any other imaging technique. The driving forces behind cardiac multidetector-row CT (MDCT) are the enormous socioeconomic relevance of coronary artery disease (CAD) and the need for comprehensive, reliable and reproducible diagnostic tests that overcome the morbidity and mortality of cardiac catheter angiography, which currently is considered the standard of reference in the assessment of CAD.
Various non-invasive imaging techniques were evaluated for assessing CAD. Some techniques such as single-photon-emission computed tomography (SPECT) focused on the assessment of myocardial perfusion, while other techniques concentrated on morphological imaging of the coronary arteries. Magnetic resonance imaging (MRI) was established as the gold standard for the non-invasive work-up of CAD. However, it failed to provide sufficient image quality for assessing the coronary arteries compared with CT.1 Moreover, its ability to evaluate myocardial perfusion, viability and ventricular function is challenged by MDCT, which also proved capable of functional cardiac imaging.2 Currently, a variety of factors are converging, potentially making coronary CT angiography a cornerstone in the management of CAD. Rapidly evolving MDCT technology is only one among these factors with a critical influence on the clinical implementation of cardiac CT.
History of Cardiac Computed Tomography
Cardiac CT was introduced at the same time that whole-body CT scanners became available. In 1977, only a few years after the introduction of CT in routine clinical practice, two methods of cardiac CT imaging were laid out: conventional CT using whole-body CT scanners3 and ultrafast CT, better known as electron-beam CT (EBCT).4 After an intense evaluation of both techniques, EBCT was considered the technique of choice for cardiac CT imaging. EBCT had a role in non-invasive cardiac imaging as a technique for coronary artery calcium scoring. With its high temporal resolution of 100ms, it was also well-suited for functional imaging. Although coronary imaging was also feasible with EBCT,5 it did not become a clinically accepted technique because of its limited spatial resolution. Despite the astonishing capabilities of cardiac EBCT imaging and the large numbers of patients who underwent coronary calcium scoring, cardiac EBCT imaging remained a niche technique.
Driven by the introduction of MDCT scanners, cardiac CT experienced a rapid rise from a niche application to a relevant mainstream technology. Starting in 1998, MDCT scanners became ubiquitously available within only a few years, providing a global infrastructure for large-volume cardiac CT imaging. The first generation of four-slice CT scanners enabled electrocardiography (ECG)-synchronised imaging of the heart with subsecond temporal resolution and a spatial resolution of about 1mm.6 Within the first year after its introduction, MDCT was shown to be capable of coronary calcium scoring and coronary CT angiography.
While four-slice CT provided a high sensitivity and specificity for the detection of coronary artery stenoses, there was a relevant proportion of up to 43% of coronary segments that could not be evaluated due to high heart rates or arrhythmia. With each subsequent scanner generation, including eight-, 16- and 40-slice scanners, the proportion of patients who could be successfully imaged with coronary CT angiography gradually increased.7 Technical improvements not only reduced the number of patients who were not evaluable because of motion artefacts, but also enabled shorter scan times, which helped to reduce contrast and radiation exposure and enabled ECG-synchronised large-volume scanning as used for chest pain imaging.
Currently, 64-row CT scanners are the most commonly used platform for cardiac CT. With a temporal resolution in the range of approximately 80–200ms, submillimetre spatial resolution and average scan times of about 10 seconds, these scanners were considered sufficiently robust for coronary CT angiography. Although the temporal and spatial resolution of MDCT is inferior compared with catheter coronary angiography, cardiac MDCT provides high diagnostic accuracies for the detection of significant CAD. In a current multicentre trial using 64-row CT scanners, a patient-based sensitivity and specificity of 85 and 90%, respectively, were reported.8 A recent meta-analysis showed even better results, with segment-based estimates of 90% sensitivity and 97% specificity and a negative predictive value of 99%.9 In fact, clinical acceptance of cardiac CT was paralleled by the clinical introduction of 64-row CT. However, with current 64-row CT there is still a relevant number of coronary artery segments that cannot be evaluated due to motion artefacts.
Technically, most authors used retrospective ECG gating for ECG synchronisation, resulting in an average radiation exposure of 12mSv with a very broad variation, depending on the scanner and the scan protocol used.10 The amount of iodinated contrast material used for the scan also varies widely in a range from about 70 to 120ml. With 64-row CT technology, the administration of beta-blockers is inevitable and arrhythmias such as atrial fibrillation are still a contraindication for coronary CT angiography. Considering these technical constraints, further improvements are needed.
State-of-the-art Technology
Currently, different technical developments can be observed, aiming at:
- improvement of temporal resolution in order to further reduce the proportion of patients ineligible for coronary CT angiography and to overcome the need for pharmacological heart rate control;
- reduction of radiation exposure in order to bring down radiation exposure below the level needed for catheter coronary angiography; and
- extension of cardiac CT imaging beyond coronary CT angiography in order to evaluate the physiological relevance of coronary artery stenoses.
Improvement of Temporal Resolution
As known from conventional coronary angiography, temporal resolution for motion-free imaging of the heart should be as low as 30ms.11 Temporal resolution of 64-row CT scanners is roughly five to 10 times worse. Therefore, all vendors are looking into software- and hardware-based solutions to further improve temporal resolution. Reducing gantry rotation time is the most straightforward approach to achieve this. Usually, projection data from half a gantry rotation are needed for image reconstruction. Bringing down gantry rotation times from 500ms in four-row CT-scanners to about 330ms in 64-slice machines improved temporal resolution from 250 to 165ms. Recent CT scanners have gantry rotation times down to 280ms. However, this approach is physically limited. Even if air bearings are used, the shortest gantry rotation time that is thought to be achievable at reasonable effort will be in the range of 200ms.
In multisegment image reconstruction projections, data from several cardiac cycles are used to compute an image. This software approach eventually brings down temporal resolution to 43ms. The implementation and efficacy of this technique vary between the different vendors. In general, data from two to four cardiac cycles are used. Unfortunately, the efficacy of this technique strongly depends on the individual patient’s heart rate and may even vary during the same scan. As data from multiple RR intervals are used, this technique is prone to motion artefacts in patients with irregular heart rates and clinical results are not unequivocally in favour of this approach.12,13 Consequently, pharmacological heart-rate control in patients with heart rates >60–70 beats per minute (bpm) remains a necessity.
Another hardware-based approach is dual-source CT (DSCT), which was introduced in 2006. Its design reflects the concept of a 30-year-old experimental scanner known as a ‘dynamic spatial reconstructor’. While the latter consisted of 28 X-ray sources,14 modern DSCT scanners consist of two tube-detector systems mounted perpendicularly in the same gantry.15 This design permits sampling of sufficient projection data from a single-gantry rotation. Compared with conventional single-source scanners, the temporal resolution is twice as high if the same gantry rotation time is applied, i.e. temporal resolution equals one-quarter of the gantry rotation time. With a temporal resolution of 75–83ms, these scanners were reported to further reduce the proportion of unevaluable coronary segments down to almost zero, even at irregular or elevated heart rates.16,17 Compared with multisegment image reconstruction algorithms, this technique has the decisive advantage of being independent of the individual patient’s heart rate. In theory it can be combined with multisegment image reconstruction techniques to even further improve temporal resolution.
Dose Reduction
Electrocardiographic Synchronisation Strategy
Basically, there are two ECG synchronisation strategies: prospective or retrospective gating. Of these, prospective gating is the most dose-efficient method, as only data that are used for image reconstruction are acquired. By contrast, retrospective ECG gating requires continuous data acquisition, enabling image reconstruction from different cardiac phases. Projection data that are not used for image reconstruction are discarded.18
Although techniques such as ECG pulsing were developed to reduce radiation exposure, retrospective ECG gating remains less dose-efficient than prospective ECG triggering.19 While retrospective ECG gating was a prerequisite in scanners with low temporal resolution, the introduction of DSCT enabled the routine use of prospective ECG triggering for coronary CT angiography. In patients with low heart rates (<60bpm), acceptable negative predictive values for exclusion of obstructive CAD can be achieved (see Figure 1).20 Thereby, radiation exposure is brought down to 2–5mSV. However, this technique depends on a stable sinus rhythm at low heart rates and the availability of a scanner with high temporal resolution. As there is a risk of missing the ideal phase for coronary artery visualisation, hybrid techniques were developed combining elements from retrospective gating and prospective triggering. With these techniques, data of pre-defined duration are acquired using a prospective trigger technique. As the length of data acquisition can vary, this approach allows the reconstruction of several image data sets at different points of the cardiac cycle, depending on the length of the trigger window. Although this technique is not as dose-efficient as pure prospective triggering, it appears to be a virtually ideal mixture to reduce radiation exposure while ensuring sufficient safety for identifying the ideal phase for image reconstruction.
Scanner Technology – Dual-source Computed Tomography versus 320-detector Rows
Starting with four-slice CT scanners in 1998, a ‘slice race’ was started with constantly increasing numbers of detector rows. This slice race did not stop with the introduction of 64-row CT scanners. Recently, 128-, 256- and 320-row single-source systems, as well as 128-slice DSCT scanners, have been introduced into clinical routine practice. While there are great differences in detector widths, ranging from 3.8 to 16cm, all of these scanners have broader detectors compared with the preceding scanner generation. Although this development appears to be quite predictable, there are different philosophies at work. The basic idea is that complete volume coverage of the heart within a single heartbeat may reduce susceptibility to arrhythmia and bring down radiation exposure to the patient. Both approaches use prospective ECG triggering utilising data from a single RR interval. However, with a 16cm detector, no table movement is needed; by contrast, with comparably small detectors such as those mounted in current DSCT scanners, rapid table movement is needed to acquire all projection data within a single heartbeat. Both techniques have specific advantages.
While the first approach provides truly isophasic data, it suffers from worse temporal resolution unless data from several rotations and therefore different heartbeats are acquired. It thereby loses the advantage of isophasic data acquisition and limits dose efficiency. The second approach is limited, as data are obtained from different points within the same RR interval. With an average cardiac scan taking about 270ms, the data come from different phases of the diastole. However, high temporal resolution and very low radiation exposure are achieved, and this technique appears to be well-suited for patients with heart rates <60bpm. While sensitivity and specificity for detection of obstructive CAD are identical with both techniques, the first approach is reported to result in an average radiation exposure of 4mSv,21 while the second technique brings exposure down in the range of approximately 1mSv.22 Therefore, both techniques are comparable or even superior to catheter coronary angiography in terms of radiation exposure.
Multidetector Computed Tomography in Coronary Artery Disease – Imaging Beyond the Coronary Arteries
The most important limitation of CT coronary angiography in the work-up of suspected or true CAD is the fact that it does not provide information on the haemodynamic relevance of a coronary artery lesion. Previous studies comparing MDCT and myocardial perfusion imaging revealed substantial differences between the presence of stenotic lesions on cardiac CT and signs of ischaemia on myocardial perfusion imaging.23 Herein lies the greatest challenge for cardiac CT imaging in CAD. The introduction of recent CT scanner technology stirred up this field by introducing a number of revolutionary changes. Accordingly, the potential of cardiac CT to evaluate coronary morphology in combination with different approaches towards myocardial perfusion imaging in a comprehensive examination are under investigation.
Perfusion-weighted Imaging
Assessment of myocardial perfusion using CT is based on changes in myocardial tissue attenuation after contrast medium administration. Perfusion defects may be identified as areas of reduced contrast enhancement during rest and/or stress imaging. The first applications of this technique were described as early as 1976.24 Since then, the technique has markedly improved, but was limited to rest imaging. The introduction of isotropic voxels enabled computation of high-quality multiplanar reformats and thereby presentation of CT images along the short and long axes of the heart, as performed in MRI or echocardiography. The development of perfusion-weighted colour maps improved the sensitivity for detecting perfusion abnormalities.25
However, the value of arterial-phase imaging at rest is limited as it provides neither quantitative data on myocardial perfusion nor direct information on myocyte injury. This is because hypoattenuating myocardium represents not only perfusion deficits but also myocardial infarction. Therefore, other techniques for assessing the effects of CAD on myocardial perfusion were sought.
To obtain information on myocardial perfusion reserve, myocardial perfusion imaging needs to be performed under stress (and rest) conditions. An initial study combining arterial-phase rest and stress MDCT including coronary CT angiography was performed in 2005 using 16-slice CT.26 This technique was refined with 64-slice CT27 and an initial patient study showed promising results with a sensitivity of 93% for detecting vessels with >50% stenosis and a corresponding perfusion abnormality on SPECT.28 As the combination of stress and rest imaging overcomes most of the inherent limitations of arterial phase imaging at rest, this approach is likely to be investigated further. With radiation exposure of ≤4mSv per acquisition, this technique will be strong competitor for SPECT imaging, as it provides myocardial perfusion imaging at the same radiation dose as SPECT with the advantage of also supplying information on coronary artery stenoses.
Dual-energy Computed Tomography
An exciting technology for the assessment of myocardial blood supply is dual-energy CT imaging. The recent availability of DSCT with its two tube-detector systems enables the simultaneous ECG-synchronised acquisition of high and low X-ray energy spectra with a single CT scan. This type of data acquisition permits separation of materials utilising the information from the different X-ray spectra. This is particularly effective in materials with high atomic numbers, such as iodine.29 As iodinated contrast material mixes with blood it can be used as a ‘tracer’ for dual-energy imaging of the myocardial blood supply. Compared with perfusion-weighted imaging, as described above, this approach provides a higher sensitivity for areas of diminished iodine uptake in the left ventricular myocardium.
While this technique is promising, it suffers two relevant limitations: it is limited in the same way as perfusion-weighted imaging at rest, because areas of reduced iodine uptake may correspond to perfusion deficits and myocardial infarction; also, the dual-energy technique results in decreased temporal resolution, because when both tubes are operated at different tube voltages, a half-scan is required to obtain enough projection data for image reconstruction. By contrast, a quarter scan suffices when both tubes are operated at the same voltage.
Therefore, temporal resolution of dual-energy imaging with current DSCT scanners is limited to 150–165ms, while single energy scanning provides a temporal resolution of 75–83ms. Nevertheless, this technique has successfully been employed in early clinical studies with small patient numbers.30,31 With its high sensitivity it will soon be employed in stress imaging as well. In addition, other strategies for the acquisition of dual-energy image data are sought that will provide integrative information on the coronary arteries as well as the state of myocardial perfusion. These approaches include rapid switching of kV levels during data acquisition and multilayer detectors that filter specific photon energies from the X-ray spectrum.
Myocardial Perfusion Imaging
Dynamic myocardial perfusion CT imaging is a functional imaging technique with the potential to detect impaired microvascular function before clinical symptoms, ECG changes or wall-motion abnormalities occur. Therefore, it goes beyond coronary CT angiography and has been established as the reference standard for prognosis and clinical decision-making of patients with CAD. Today, imaging of myocardial perfusion in clinical routine practice is dominated by SPECT and MRI.
The potential of CT to quantify myocardial perfusion was recognised as early as the 1980s, and early EBCT studies in animal models as well as in healthy volunteers demonstrated the ability of CT to determine myocardial blood flow.32,33 Exploratory studies with previous generations of MDCT scanners have described dynamic imaging approaches for quantification of myocardial perfusion.34,35 A major drawback was the limited volume coverage, limiting perfusion imaging to an area of approximately 2–3cm.
The availability of detector arrays that are wide enough to cover the entire cardiac anatomy21 or scanners that are fast enough to cover the entire left ventricle in a single heartbeat22 permit whole-heart acquisition of dynamic, time-resolved data on myocardial perfusion. There are two techniques: with 256- or 320-slice scanners the entire heart is covered during a single rotation and repeated scans allow for quantitative perfusion analysis; alternatively, repeated rapid ECG-gated dynamic spiral scanning using recent DSCT scanners also permits dynamic whole-heart myocardial perfusion imaging. These techniques are under investigation for myocardial perfusion imaging at stress and rest, with very promising early results from animal studies (see Figure 2).36 However, dynamic scanning results in repeated data acquisition and therefore increased radiation exposure compared with static perfusion-weighted imaging. New techniques such as iterative image reconstruction are under investigation to bring down radiation exposure into ranges below that of myocardial SPECT perfusion imaging.
This approach holds the potential to extend myocardial perfusion imaging into a truly quantitative 3D imaging technique simultaneously providing information on regional myocardial perfusion and the coronary arteries. Considering CT’s inherent advantages such as the linear relation between contrast enhancement and iodine concentration or its ease of use, CT will become a serious competitor for SPECT and MRI. Finally, not only is cardiac CT useful in the diagnostic work-up of patients with suspected CAD, it also permits a detailed assessment of the sequelae of CAD, allowing for the assessment of myocardial viability using delayed contrast enhancement.2 Only recently, it was shown to be feasible to assess myocardial oedema in acute MI.37 Its potential for assessing global and regional left ventricular function has repeatedly been reported.38 With its ability to dynamically visualise cardiac structure at high spatial resolution, it permits a detailed assessment of the cardiac anatomy, including the heart valves. It can be used to plan cardiac surgery or resynchronisation therapy. Today, it is an appropriate technique for the non-invasive assessment of the coronary arteries in uninterpretable ECG and is the reference technique for assessing coronary anomalies.39
With the broad introduction of state-of-the-art CT scanners, many indications that nowadays are considered inappropriate or uncertain for cardiac CT will become appropriate for routine cardiac CT. While the diversity and sophistication of cardiac imaging techniques commonly results in the question of ‘which test for which patient?’, the answer will more often be ‘cardiac CT’ as a single test for a broad range of clinical questions.