Dynamic CT Myocardial Perfusion Imaging

Abstract

Cardiac CT offers several approaches to establish the hemodynamic severity of coronary artery obstructions. Dynamic myocardial perfusion CT (MPI_CT) is based on serial CT imaging to measure the inflow of contrast medium into the myocardium and calculate absolute measures of myocardial perfusion. This review describes the MPI_CT acquisition protocol, post-image acquisition processing and calculation of quantitative parameters, the diagnostic performance of MPI_CT and the potential incremental value of this technique in comparison to alternative approaches. Further technical innovation using different scanner platforms and establishment of reproducible diagnostic thresholds to differentiate significant coronary artery disease will be crucial in the path to broader clinical implementation.

1.1. Principles of dynamic CT myocardial perfusion imaging

Dynamic stress myocardial perfusion CT (MPI_CT) is performed by acquiring a series of CT images after injection of a bolus of contrast medium during pharmacological hyperemia, similar to perfusion imaging technique by MRI or PET. Dynamic MPI_CT was first demonstrated in the late 1970s using electron-beam CT, but this technique never reached broad clinical application, partly due to the inability of scanners at the time to image the entire myocardium in a single acquisition¹. To measure blood flow throughout the left ventricle, today dynamic MPI_CT is performed on scanners with a detector-row width that covers the entire ventricle in 1 or 2 acquisitions. This requirement restricts dynamic MPI_CT to the latest generation wide-detector CT systems and 2^nd/3^rd generation dual-source CT systems. In order to limit the cumulative dose of a serial acquisition protocol, individual datasets are acquired using a low tube potential. Nevertheless, exposure is generally higher than what can be achieved with static MPI_CT protocols. By measuring regional myocardial attenuation over time quantitative measures of myocardial perfusion can be calculated throughout the myocardium of the left ventricle and displayed as volumetric perfusion maps. Due to the high iodine concentration in the ventricular cavity, beam hardening artifacts will inappropriately reduce attenuation values within the myocardium, particularly in the basal inferior wall positioned between the contra-enhanced ventricle and descending aorta. Correction of these beam hardening artifacts is essential for accurate myocardial perfusion imaging. Dynamic iterative correction algorithms that take into account different tissue types and changes in iodine content over time perform better than conventional beam-hardening correction algorithms^{2, 3}.

1.2. Scan protocol

Patients are required to abstain from caffeine 12-24h prior to the examination. If CT angiography is performed prior to the perfusion scan, a 10-15 minutes delay is recommended for washout of contrast medium. Adenosine or another vasodilators are administered through an intravenous cannula for 3-5 minutes. Regadenoson is administrated as a single bolus. The cardiac rhythm is continuously monitored and the blood pressure is measured at regular intervals. Just before the scan, a short, high-rate iodine contrast bolus is injected through a separate intravenous cannula followed by a saline bolus. Imaging starts just before the contrast reaches the right ventricle. Dual-source CT scanners have a longitudinal coverage of 3.8-5.8 cm, which is sufficient to cover the heart in two scans. The table moves back and forth between acquisitions resulting in a sampling rate of one complete dataset every 4 seconds, or every 6 seconds in cases of very fast heart rates. During a 30s breath hold, between 10 to 15 complete, low-dose datasets are acquired. Some MPI_CT protocols on wide-detector systems allow for superficial breathing by providing a post-acquisition displacement correction⁴. If a CT system with full cardiac coverage is used, a full dataset could be acquired every heart cycle. However, to limit roentgen exposure, the sampling rate can be altered during the exam with the objective of acquiring a maximum sampling rate during the crucial phase of myocardial contrast inflow and a lower sampling rate before and after⁴. In most studies, end-systolic datasets have been acquired to minimize beam-hardening artifacts from the contrast-filled LV cavity.

When MPI_CT is performed as part of a combined examination with coronary CTA, the so-called stress-rest protocol, this has logistic advantages because no delay is needed between the MPI_CT and CTA for washout of contrast medium from the myocardium. However, in clinical practice CTA effectively rules out coronary disease in the majority of patients with new suspected ischemic heart disease, in which case a perfusion scan is redundant and unnecessarily exposes patients to contrast medium and radiation. Particularly for patients without known coronary disease, an approach that starts with CTA, selectively followed by stress MPI_CT if angiographic lesions are detected on CTA, would be a more efficient approach in terms of effort and radiation exposure. The need for immediate CTA interpretation represents a logistic drawback to this approach, as well as the required contrast wash-out delay before MPI_CT can be performed⁵.

1.3. Calculation of quantitative parameters of myocardial perfusion

Despite best efforts by patients to hold still, often there will be some degree of displacement of the myocardium during the long scan period. Because it is important to compare attenuation values over time for the same tissue, correction of this gradual displacement is a crucial step in the data processing. Rigid and non-rigid transformation techniques can be applied to realign the 2D cross-sections, ideally compensating for in-plane as well as through-plane displacement of the ventricle. Rhythm irregularities that result in data acquisition during a different phase of contraction are more difficult to correct and may require exclusion from the quantitative analyses. The next step is to isolate the ventricle and discretize the myocardium into small volumetric elements. Within each myocardial element the measured attenuation values are plotted against time. The arterial input function is derived from a sample volume in the descending aorta. The myocardial time-attenuation curves are coupled with the arterial-input function using a hybrid deconvolution model, which uses a simplified impulse-residue function for modeling the interaction between the intravascular and extravascular compartments, after which the myocardial blood flow can be calculated by dividing the convoluted maximal slope of the myocardial time-attenuation curve by the maximum arterial input function⁶. Myocardial blood flow, as well as other parameters of myocardial perfusion like perfused capillary blood volume, and first-pass distribution volume, are reconstructed as color-coded volumetric maps. Regions of interest can be sampled manually to obtain the average myocardial blood flow. Alternatively, polar maps or bull-eye plots, can be created, which summarize myocardial perfusion parameters in a single image and report myocardial MBF per standardized myocardial segment using the 16/17-segment AHA classification. MBF measured by CT is generally lower than by other perfusion techniques, which is partly related to lower sampling rates of dual-source systems. MBF also varies between individuals, for various clinical and technical reasons^{7, 8}. Reported MBF cut-off values that signify hemodynamic significance vary substantially from 75 and 164 ml/min/100ml between studies [Table 1]. Due to these inconsistencies, normalization of regional MBF values relative to a measure of global MBF appears to improve the diagnostic accuracy of MPI_CT^{6, 9-11}.

Table 1: Diagnostic performance of dynamic MPI_CT.

Summary of studies comparing dynamic stress CT myocardial perfusion imaging (MPI) against invasive coronary angiography (ICA) with fractional flow reserve (FFR_cath) or other noninvasive functional tests. Sensitivity (sens), specificity (spec) and area under the curve (AUC) are listed on a per-vessel basis, and compared to CT angiography, if data available. Myocardial blood flow (MBF, ml/min/100cc myocardium), single/dual source computed tomography (SSCT/DSCT), coronary flow rate (CFR).

Author (year)	N	CT system (vendor)	Dose (mSv)	MBF Cut-off	Reference	CTA			MPICT
						Sens (%)	Spec (%)	AUC	Sens (%)	Spec (%)	AUC
Bamberg 2011	33	128-slice DSCT (Siemens)		75	ICA/FFRcath	100 (88-100)	51 (39-63)		93 (77-99)	87 (76-94)
Wang 2012	30	128-slice DSCT (Siemens)	9.5±1.3	NR	MPISPECT	90	51		100	75.7
Weiniger 2012	10	128-slice DSCT (Siemens)	12.8±2.4	Visual	MPIMRI				86	98
					MPISPECT				84	92
Huber 2013	32	256 SDCT (Philips)	9.5	164	ICA >75% FFRcath ≤ 0.75				76 (57-90)	100 (95-100)	0.86
Rossi 2014	99	128-slice DSCT (Siemens)		78	FFRcath ≤ 0.75	61 (77-94)	77 (84-97)		85 (69-94	89 (78-95)	0.95
Kono 2014	42	128-slice DSCT (Siemens)	9.4	0.85 (index)	FFRcath ≤ 0.80				98	70	0.85
Kikuchi 2014	32	320-slice SSCT (Canon)	12.8±2.9	2.97 (CFR)	MPIPET				86	92	0.90
Tanabe 2016	53	256 MDCT (Philips)	10.5-10.6	92	MPISPECT (N=25)				95 (52-100)	72 (53-91)	0.87
				98	MPIMRI (N=28)				78 (67-97)	80 (58-86)	0.89
Coenen 2017	74	128-slice DSCT (Siemens)	9.3-1.8	0.71 (index)	ICA/FFRcath ≤0.80	78 (65-90)	49 (38-61)	0.70	73 (61-86)	68 (56-80	0.75
Rossi 2017	115	128/192-DSCT (Siemens)	6.0-10.3	75 (index)	ICA/FFRcath ≤0.80	35 (21–53)	95 (81–99)	0.65	89 (76-96)	73 (59-83	0.85
Pontone 2019	85	256-slice SSCT (GE)	5.3±0.7	101	ICA/FFRcath	83 75-91)	66 (9-73)	0.83	73 (63-83)	86 (81-91)	0.88
Nishiyama 2019	38	256-slice SSCT (Philips)	10.2±1.2	126	ICA/FFRcath <0.75	96 (88–100)	57 (46–67)	0.84	83 (68–98)	93 (88–98)	0.96
Alessio 2019	34	256-slice SSCT (GE)	8.4±1.1	126	82-Rubidium PET				75	83	NR

2.1. Diagnostic performance

The diagnostic performance of MPI_CT has been validated in several animal studies^{6, 12, 13}. Myocardial blood flow measured by CT demonstrated good correlation with directly measured coronary flow, fractional flow reserve and MBF determined by microspheres. In human patients the performance of MPI_CT has been compared against catheter-based FFR as well as other myocardial perfusion imaging techniques^{4, 6, 9-11, 14-22}, summarized in table 1. The studied cohorts were typically referred for clinically indicated invasive angiography and therefore had a relatively high coronary disease burden. MBF or indexed MBF demonstrated superior and incremental discriminatory value over CTA in most studies. According to a meta-analysis by Lu et al, dynamic MPI_CT identifies hemodynamically significant coronary artery disease with a sensitivity and specificity of 85% and 81% on a per-vessel basis, compared to 82% and 61% by CTA²³. MPI_CT relies on consistent myocardial sampling and the diagnostic performance is negatively affected by gross cardiac motion and arrhythmia. Incidental rhythm irregularities can be corrected by excluding a specific phase from the analysis. Several studies reported improved diagnostic performance using indexed MBF values, which are relative MBF values normalized against remote myocardium or the 75^th percentile of the left ventricle MBF, to neutralize global differences in measured MBF values between individuals and examinations^{10, 11}.

To differentiate reversible ischemia from prior infarction some centers perform both a rest and a stress perfusion scan. To avoid an additional scan, the coronary CT angiogram can often serve as a static resting perfusion scan²⁴. In case of prior infarction, the dynamic MPI_CT will show very low MBF values²⁵. CT imaging of late iodine enhancement represents an alternative option to identify myocardial scar^26-28.

2.2. Prognostic value

There have been several registry reports from modest-size cohorts on the prognostic value of MPI_CT. Nakamura et al followed 332 patients with suspected coronary disease for 2.5 years and showed that the summed stress score based on normalized MBF by MPI_CT predicted a composite endpoint of cardiac death, nonfatal myocardial infarction, unstable angina, or hospitalization for congestive heart failure, with an incremental prognostic value on CTA based coronary stenosis: hazard ratio 5.7; 95% confidence interval: 1.9 to 16.9; p = 0.002²⁹. In a cohort of 81 patients, Assen et al reported that indexed MBF by MPI_CT predicted adverse events, including cardiac death, nonfatal myocardial infarction, unstable angina requiring hospitalization, or revascularization, over 18 months: hazard ratio 11.4 (95% confidence interval: 3.4 to 38.2; p<0.001), with superior and independent predictive value compared to CTA and CT-FFR⁸.

2.3. Clinical effectiveness

In one prospective, randomized controlled trial between a tiered cardiac CT protocol that included dynamic MPI_CT and conventional stress testing for patients with stable chest pain symptoms⁵, the cardiac CT approach was associated with a higher rate of coronary disease with a class I indication for revascularization by ESC standards (88% vs 50%), without increasing catheterization rates.

3.1. Comparison of dynamic MPI_CT to other noninvasive functional tests

Dynamic MPI_CT is technically somewhat more demanding than static MPI_CT, but has the conceptual advantage of absolute blood flow measurements. Quantitative measures are potentially valuable for differentiating degrees of ischemia and changes over time, and may allow for quantification of microvascular disease. Anecdotally, dynamic perfusion MPI_CT is less susceptible to beam-hardening artifacts. No large-scale direct comparative studies have been performed between dynamic and static MPI_CT in humans. However, a meta-analysis by Danad et al, suggests that dynamic MPI_CT may be more accurate, albeit at the expense of a higher radiation dose³⁰. At the time of their analysis, the radiation dose of MPI_CT was around 10mSv, based largely on studies with 2^nd generation dual-source CT systems and higher kVp settings. More recent studies using contemporary CT technology show that dynamic MPI_CT is possible at doses around 5mSv or less^{4, 14, 31}. Meta-analyses indicate that the diagnostic accuracy of static/dynamic MPI_CT is comparable to other stress imaging modalities using invasive FFR as reference³² . On a per-vessel level, Takx et al reported that the diagnostic accuracy of MPI_CT (AUC 0.93) was in the same range as perfusion imaging by MRI (AUC 0.94) and PET (AUC 0.93), and these perfusion imaging modalities outperformed stress echocardiography (AUC 0.82) and SPECT perfusion imaging (AUC 0.83)³². Although meta-analyses cannot be considered conclusive, there is currently no indication that MPI_CT performs inferior to other stress imaging techniques. MPI_CT also has the practical advantage that it can be performed in conjunction with coronary CTA, and images may be merged for a comprehensive interpretation of coronary artery disease. Additional radiation and contrast medium exposure represent a drawback of MPI_CT compared to MRI, echocardiography and stress testing without imaging.

CT-derived FFR represents an alternative approach to assess the hemodynamic significance of coronary disease on CTA. Dynamic MPI_CT has been compared to CT-FFR in several studies, and showed comparable performance and complementary value^{4, 9}. Coenen, et al. concluded that for patients with an on-site performed CT-FFR result within the diagnostic grey zone, MPI_CT provided an efficient approach to improve diagnostic accuracy in the detection of hemodynamically significant coronary lesions⁹. Pontone, et al. confirmed these observations using a commercially available CT-FFR solution and showed that a similar stepwise approach increased diagnostic performance (AUC increased from 0.88 to 0.92, P<0.05)⁴.

4.1. Future developments

The current evidence for dynamic MPI_CT is based on relatively small, often single-center cohorts. More than half of these studies were performed on dual-source CT systems from one specific manufacturer. For dynamic MPI_CT to mature into a clinically used diagnostic tool it will be important to validate the technique in larger multicenter cohorts using different CT systems³³. Inter-individual variation in measured global myocardial blood flow challenges interpretation of absolute perfusion values. This necessitates the use of normalized MBF indexes to identify inducible myocardial ischemia. Further research is needed to identify perfusion parameters that best identify significant myocardial ischemia. Comprehensive interpretation and revascularization decision making would benefit from a robust infarct imaging technique for CT. Finally, it will be important to develop more efficient scan protocols to achieve the highest diagnostic value with the lowest effective dose. Prospective studies in larger cohorts are needed to determine the most efficient deployment of MPI_CT and assess its performance against other diagnostic tests for the diagnostic evaluation of patients with suspected or known coronary disease.

Figure 1: Acquisition protocol.

For a “rest-stress approach”, CTA is performed first, followed by dynamic stress CT myocardial perfusion imaging in case of abnormalities. A delay between both scans is required for contrast medium washout. A low-dose native scan is performed during systole to plan the perfusion scan, which is also performed during systole. Attenuation values throughout the myocardium and the aorta are plotted against time, from which myocardial perfusion maps can be reconstructed. Nitroglycerin (NTG), beta-blocker (BB), contrast medium (CM), myocardial perfusion imaging (MPI), myocardial blood flow (MBF).

Figure 2: MPI_CT case example.

CT angiography demonstrating a severe lesion in a large marginal branch off the left circumflex coronary artery (A). The myocardial blood flow map obtained by dynamic MPI_CT shows a distinct defect in the mid antero-lateral wall indicating hemodynamic significance of the coronary lesion (B). Confirmation of a high-grade stenotic lesion in the marginal branch by invasive angiography.

Abbreviations

AUC

Area under the curve

CT

Computed tomography

CTA

Computed tomography angiography

ESC

European Society of Cardiology

FFR

Fractional Flow Reserve

MBF

Myocardial blood flow

MPI

Myocardial perfusion imaging

MRI

Magnetic resonance imaging

PET

Positron emission tomography

SPECT

Single-photon emission computed tomography