There are currently two main CT techniques to perform MPI: static and dynamic MPI. Static MPI consists of a single electrocardiogram (ECG)-gated, iodinated contrast-enhanced CT acquisition of the heart. Static CT-MPI can be performed using either single- or dual-energy (SECT or DECT) settings. First described in 2008 [5], DECT is a relatively new technological development in the field of CT imaging with different vendor-specific solutions allowing the quantification of iodine concentrations. Static CT-MPI imaging indirectly evaluates myocardial perfusion by looking at the distribution of contrast (in the case of SECT) or iodine concentrations (in the case of DECT) throughout the myocardium at a single time point. In contrast with static CT-MPI, dynamic CT-MPI acquires multiple images over a specific time period, visualizing the entire in- and outflow of contrast medium. This dynamic nature enables direct quantification of myocardial perfusion. Figure 1 shows the differences between static and dynamic CTMPI.
PET-MPI is currently the reference standard for absolute quantification of myocardial perfusion and SPECT for visual analysis [6]. However, CT in addition offers anatomical CAD evaluation due to a higher spatial resolution at reduced cost [6]. CT is able to directly quantify the myocardial perfusion due to the linear relationship between iodine concentration and the change in signal intensity (HU values) over a wide range of iodine concentrations. This is in contrast to MRI-MPI in which the relationship between gadolinium and signal enhancement [7] becomes non-linear at the gadolinium concentrations that are commonly used for visual MPI analysis [7].
Technical information and requirements
The most important requirement for CT-MPI is a suitable CT system. For static CT-MPI, a 64-multidetector CT system is the minimum requirement, although wider volume CT systems are preferred. Smaller detector range systems require multiple slabs to image the heart completely which leads to a relatively low temporal resolution, thereby increasing overall scanning time. This increased scanning time can lead to changes in contrast enhancement of the myocardium during the acquisition resulting in both false-positive and false-negative results. The first DECT systems required a decrease in temporal resolution for cardiac acquisitions [8]. However, newer DECT systems addressed this limitation leading to temporal resolutions similar to SECT systems 83 ms or lower, depending on the dual-energy approach [8].
Dynamic CT-MPI is most commonly performed using dual-source CT (DSCT and wide multi-detector CT (MDCT) systems (256–320 rows). Second- and third-generation DSCT scanners, characterised by their relatively high temporal resolution (75 or 66 ms), have a coverage sof 7.3 and 10.2 cm. Using a shuttle mode, consisting of alternating back-and-forth table movements with sequential scanning, allows visualization of most of the left ventricle [8,9,10]. MDCT systems, used in only a limited number of dynamic CT-MPI studies [11, 12], allow for full-heart coverage within one gantry rotation using a single-tube wide detector with 256 or 320 detector rows covering 8 and 16 cm. The disadvantage of these MDCT systems is the decreased temporal resolution (140 ms), which can result in an increase in motion artifacts and an overall decrease in image quality [8]. However, wide-detector MDCT systems are able to acquire an image every heartbeat in contrast to requiring images every other heartbeat such as with DSCT systems. This is made possible by eliminating the time needed for the shuttle mode. The increase in acquired images will increase the information about the in- and outflow of contrast; however, every additional image comes at a cost of increased radiation dose. Figure 2 shows the difference between the DSCT shuttle mode and the MDCT wide detector approach.
Radiation dose
Static CT-MPI can be performed at the lowest radiation dose of all CT perfusion techniques, similar to CCTA imaging, with reported radiation doses between 2 and 9 mSv for combined rest-stress examinations [2, 4, 13]. The radiation dose of dynamic CT-MPI is relatively high due to the increased number of acquisitions that need to be taken, resulting in effective dose between 5 and 13 mSv [13, 14]. Nevertheless, recent studies that used the newest high-end CT systems generally show lower radiation doses (5–9 mSv) [11, 15]. Dynamic CT-MPI radiation doses could be reduced by applying new acquisition techniques that require only a limited number of optimally timed acquisitions. Additionally, these timed acquisitions could be used for conventional CCTA evaluation [16].
Comparing SECT with DECT CT-MPI shows that used radiation doses are similar [17,18,19]. Studies about dual-source DECT imaging show strong evidence that DECT imaging is not associated with increased radiation dose levels compared to SECT acquisitions acquired at 120 kVp [20]. Detector-based DECT systems are expected to produce similar radiation doses to SECT. However, these detector-based DECT scans are standardly acquired at a high tube voltage, and therefore, the tube current should be lowered to maintain similar radiation dose levels [19].
Traditionally, radiation exposure can be reduced by using lower kVp levels (100 kVp instead of 120 kVp) while maintaining optimal image quality in high-end systems. Nonetheless, DECT scanning at lower kVp will only marginally reduce the radiation dose compared to SECT, since both low and high kVp spectra are fundamentally required for DECT. In addition, dual-source DECT systems offer the use of a tin filter (Sn). Sn filters are used to block low-energy photons of the high-energy beam (100/150 kVp) improving spectral separation and reducing total radiation dose [21].
Image protocol
Exam preparations
Patients should be instructed to not consume any caffeine-containing substances (coffee, tea, chocolate, etc.) for 24 h before the examination, because of the interfering effects of caffeine on the effectiveness of the stressor agent. Especially adenosine is sensitive to the competing effect of caffeine, while early-stage research indicates that regadenoson is less affected [22, 23].
Heart rate-limiting medication such as beta-blockers and coronary dilating medication such as sublingual nitrates are known to influence myocardial perfusion; therefore, it is advised to only use them when necessary to achieve diagnostic image quality for CCTA imaging [24, 25].
CCTA, rest, and stress phase acquisitions
Traditionally, MPI can be performed by using a rest and stress phase acquisition to discriminate myocardial ischemia from infarction. Ischemic defects are present on stress images only while infarcted defects are present on both rest and stress images. Using CT, these functional acquisitions can be performed in conjunction with a CCTA acquisition for anatomical coronary evaluation.
For static CT-MPI, the rest phase acquisition is similar to the CCTA and can be performed with the standard CCTA protocol [3]. For dynamic CT-MPI, the CCTA need to be acquired separately due to the dynamic nature of the rest acquisition. For dynamic CT-MPI, the use of CCTA, rest, and stress acquisitions will result in high radiation doses. Therefore, a stress-only approach is used in several dynamic CT-MPI studies [9, 26]. By using the quantification abilities of dynamic CT-MPI, discrimination between ischemic and infarcted myocardium is still possible.
Point of debate is the order of acquisitions, which will ultimately depend on which evaluation takes precedence. In patients with a high probability or known CAD or in patients with stents or bypasses, myocardial ischemia is likely, and functional evaluation is more important than anatomical evaluation. In these cases, stress CT-MPI imaging should be performed first, eliminating the effect of contrast contamination and the suppressing effect of beta-blockers and nitroglycerin on myocardial ischemia [24, 25]. In patients with a low to intermediate probability of CAD, it would be preferable to start with CCTA (rest phase for static CT-MPI), in order to avoid unneeded testing in case of no stenosis. An important factor for successful multi-acquisition imaging is to allow sufficient time between each examination to allow complete outflow of contrast agent to avoid contrast contamination and reduce the effect the stressor agent for the subsequent acquisition. Figure 3 shows a schematic representation of a complete cardiac work-up protocol, including CCTA, perfusion, and delayed enhancement imaging.
Stressor agents
For stress CT-MPI, a pharmacological stressor agent is used to achieve maximal hyperemia, with adenosine the most frequently used. However, the use of adenosine causes unwanted short-term side effects such as bronchial constriction. This is especially apparent in patients with reactive airway disease, such as COPD, a frequent comorbidity in patients with CAD. Although the short half-life of adenosine allows for the abrupt discontinuation of administration and rapid disappearance of potential harmful side effects, it requires continuous intravenous administration during acquisition and weight-based dosing [23, 27, 28]. Another stressor agent that is increasingly used is regadenoson. It is a potent and selective coronary vasodilator with a rapid onset of action and a longer half-life compared to adenosine. Therefore, regadenoson can be administered as a fixed-dose bolus without weight adjustments. Furthermore, given its selectivity, there are less serious side effects making it also suitable and safe to use in COPD patients [22, 29]. Nevertheless, some studies also report negative effects of the use regadenoson, like higher rates of arrhythmias and increased use of rescue agent aminophylline [30]. Concluding, stress acquisitions require the administration of adenosine for 2–5 min with a rate of 140 μg/min/kg or a single injection of 0.4 mg of regadenoson.
Acquisition timing
For static CT-MPI, scan timing is an important factor. The image should be acquired during the early arterial phase of first-pass contrast enhancement at the peak of the contrast enhancement curve. The timing of the peak can be estimated using a test-bolus or bolus-tracking technique. Optimal time delays vary between 2 and 4 s depending on measurement location (ascending or descending aorta) and on the HU threshold (150 or 250 HU) [31].
For dynamic CT-MPI, temporal sampling rates become a significant factor. Several studies have shown that the limited temporal sampling rates that are currently being used lead to an underestimation of the perfusion compared to the true perfusion values and PET determined values [32, 33].
Another important timing factor for dynamic CT-MPI is the timing of the images with respect to the cardiac phase. Myocardial perfusion is not constant but varies between the systolic and diastolic phase [34, 35]. As a result of the myocardial contraction, myocardial perfusion takes place during the diastolic phase and is maximal end-diastolic. However, systolic phase imaging offers some important advantages. Firstly, during systole, the heart is contracted, resulting in a smaller heart volume and in particular a shorter basal-apical length. This is beneficial in scanner systems with limited heart coverage. Wide-detector MDCT studies are performed mostly in diastolic phase, while still being able to capture the entire heart at its maximal volume [11, 36]. Secondly, the systolic phase is constant in duration independent of heart rate and is less sensitive to arrhythmia which is beneficial during stress acquisitions. Finally, because of the reduced volume and reduced contrast dose in the left ventricle, beam-hardening artifacts are reduced.
Image analysis
Qualitative analysis
CT-MPI can be qualitatively analyzed by visual inspection of the contrast attenuation on short-axis view images of the left ventricle. For optimal assessment of perfusion defects, thick sections, minimum intensity projection, and a narrow window width can be used [37]. A standard AHA-17-segment model is used to describe perfusion defects.
Additionally, for DECT static CT-MPI, virtual mono-energetic images (VMI) can be created. VMI images offer the advantages of being able to be reconstructed at multiple energy levels by using different percentages of the low and high energy data. Low energy reconstruction (e.g., below 80 kV), will lead to an increase in contrast attenuation exceeding the increase in noise level. These images can be used to assess patients who cannot receive the normal dose of contrast agent or when the contrast-bolus given was sub-optimal [38, 39]. High energy reconstructions (e.g., 110 kV and higher) can be used to reduce blooming artifacts or metal artifacts, enhancing the ability to analyze severe calcified plaques or stents [40, 41].
Dynamic CT-MPIMPI offers the possibility of visually analyzing the differences in HU values over time. Regularly, this is represented as a color-coded perfusion maps based on quantitative MBF values, representing MBF values per pixel over time. This approach allows for the visual detection of three-vessel disease. Figure 4 shows an overview of all the options for qualitative analysis
Semi-quantitative analysis
Although semi-quantitative analysis of static CT-MPI acquired at single-energy is attempted using parameters such as the transmural perfusion ratio, visual analysis is still superior [42,43,44].
Currently, only DECT acquisitions allow for semi-quantitative analysis of static CT-MPI using an iodine concentration map, see Fig. 5. The accuracy of the third-generation DSCT and first-generation dual-layer CT (DLCT) systems for the quantification of iodine is shown to be highly accurate [45]. The DSCT systems showed slightly more accurate measurements, especially using the 150Sn/70 or 150Sn/80 kVp combination, with a lower measurement error compared to DLCT systems. Fast kV-switching systems show similarly high accuracy to the DSCT systems [46, 47]. A limited number of studies on iodine quantification have shown that thresholds between 2.1 and 2.5 mg/mL of iodine are optimal to discriminate diseased from normal myocardium using a stress acquisition, while at rest, a threshold of 1.0 mg/mL of iodine can be used to discriminate between infarcted and ischemic myocardium [48, 49].
Quantitative analysis
Quantitative analysis using dynamic CT-MPI can be performed by direct or indirect methods. Indirect perfusion parameters are derived from the tissue attenuation curves (TAC) and the arterial input function curve (AIF) using the so called upslope method. The upslope method takes the ratio between the maximum upslope of the TAC curve and the maximal AIF value as a measure of perfusion. The main advantage of this method is that it is fairly robust and computationally easy while only the upslope is needed thereby eliminating all acquisitions made after the TAC peak, possibly reducing radiation dose. Figure 6 shows the basic principle of the upslope method based on the attenuation curves.
Direct quantitative analysis of dynamic CT perfusion data is done using a model-dependent deconvolution approach, similar to MRI perfusion [7]. For CT-MPI, there are several models available. From these models, the myocardial blood flow (MBF) can be determined [7, 50, 51]. The most used approach for CT perfusion quantification is a hybrid approach where the deconvolution technique is used to model the curves, and subsequently, the upslope method is used to calculate MBF [52]. Studies have shown that there is no difference in the diagnostic accuracy between different models; however, the absolute values can be significantly different and this should be taken into account when threshold is chosen to detect abnormalities [53]. Figure 6 shows an example of a perfusion map based on the hybrid approach compared to SPECT perfusion imaging results.
Absolute versus relative perfusion values
A wide variety of absolute and MBF threshold values for the detection of perfusion defects are reported in the literature [13, 14, 54]. Variation in these values can be caused by differences between patient populations, CT systems, image protocols, and analysis procedures. A solution is the use of a relative measure that compares healthy myocardium with ischemic myocardium within a patient. However, using relative values eliminate the benefit of evaluating global perfusion defects.
Several studies reported that relative MBF values yielded a higher diagnostic accuracy compared to absolute MBF values [55,56,57]. However, one study reported that myocardial flow reserve estimations were not as accurate as individual MBF estimates using PET perfusion as a reference [11].
MBF versus MBV
Analysis of the myocardial blood volume (MBV) could assist in the discrimination of ischemic and infarcted myocardium in combination with MBF. MBV can be derived by using the peak enhancement of the TAC [58]. In ischemic myocardium, the arterioles dilate to compensate for the decreased flow caused by a coronary stenosis keeping the MBV in the myocardium constant. In infarcted myocardium, this compensatory mechanism is not available leading to a decrease in MBV [58,59,60,61,62].
Diagnostic accuracy
The largest study on static CT-MPI (381 patients) is the CORE320 study. This study showed an accuracy of 0.93, with a sensitivity and specificity of 80 and 74%, compared to combined SPECT and invasive coronary angiography [63,64,65,66]. Another multi-center study (110 patients), performed by Cury et al., investigated the accuracy of static CT-MPI compared to SPECT-MPI and showed slightly higher sensitivity and specificity of 90 and 84% with an overall accuracy of 87% [67]. Meta-analyses on static CT-MPI showed that the average sensitivity and specificity ranged between 75–84% and 78–95% for a variety of image protocols and reference standards used [2, 68].
Overall DECT-MPI studies showed good accuracy compared to various reference modalities [4, 48, 49, 69,70,71,72,73]. Sensitivity and specificity are reported around 89 and 78%. A limited number of studies have been performed on the quantitative analysis of DECT-MPI acquisitions. They showed that a rest acquisition is needed to discriminate between ischemic and infarcted myocardium [48, 49].
For dynamic CT-MPI, similar accuracies have been reported as with static CI MPI with an average sensitivity and specificity ranging between 76–100% and 74–100%, respectively [2, 13, 14, 74, 75]. Quantitative analysis studies have focused on determining an MBF threshold to discriminate ischemic from normal myocardium resulting in a wide range of threshold values between 75 and 136 mL/100 mL/min [13, 14, 54].