Stress Cardiovascular Magnetic Resonance: Clinical Myocardial Perfusion


Considerable progress in myocardial perfusion cardiovascular magnetic resonance (CMR) has been achieved recently and its diagnostic performance has been documented in single-center and multicenter trials. From single-center trials, and particularly from multicenter trials, the knowledge of how to perform and interpret perfusion CMR studies has improved considerably. Also, convincing evidence has been obtained regarding the robustness of the technique when performed at different sites, the importance of the reader experience for data analysis, and the excellent prognostic power of perfusion CMR. Finally, an increasing number of studies are available that now also address the economic consequences of introducing this new technique for diagnosing and managing coronary artery disease (CAD).

The CMR technique offers an almost unbeatable versatility for exploring any aspect of cardiac disease. Therefore this chapter starts by illustrating how perfusion CMR is embedded in the concepts of atherosclerosis development and the vulnerable plaque theory. It then addresses established and not yet established aspects of the perfusion CMR protocol, including the different approaches proposed for data interpretation and analysis. In this chapter, the current performance of clinical perfusion CMR is presented to diagnose CAD and predict outcome. A perspective for the future of perfusion CMR is provided at the end of the chapter.

The Rationale for Perfusion Imaging

Endothelial dysfunction is generally recognized as a crucial initial step in the development of atherosclerosis. Among many alterations in dysfunctional endothelium, the production of nitric oxide is reduced, which promotes the recruitment of leukocytes to the vessel intima through enhanced expression of endothelial adhesion molecules. In the vessel wall, mononuclear leukocytes become foam cells by lipid accumulation and form reversible fatty streaks. In addition, recruitment of smooth muscle cells and their production of extracellular matrix induce the formation of fibrous lesions. During the course of atherogenesis, inflammatory triggers and other factors can affect the balance between the production of matrix-degrading enzymes and collagen production by macrophages, which can cause fibrous cap destabilization. Subsequent repetitive plaque ruptures (and healing), even when clinically silent, may cause a stepwise progression of atherosclerotic lesions. A final rupture then of larger plaques can cause the ultimate thrombus formation and acute vessel occlusion. In line with this concept, invasive studies conducted mainly in the preinterventional era demonstrated a positive correlation between stenosis degree in coronary angiography and risk for plaque rupture and acute myocardial infarction. Fig. 18.1 demonstrates in approximately 17,000 patients an exponential increase in the likelihood of a coronary occlusion associated with a plaque rupture with an increasing degree of coronary artery stenosis. Accordingly, vulnerable plaques were defined by four major criteria : (1) thin cap with large lipid core, (2) stenosis degree >90%, (3) endothelial denudation with superficial platelet aggregation, and (4) fissured plaque.

FIG. 18.1, The relationship between the degree of diameter stenosis in invasive coronary angiography and occlusion rate per year in 4554 patients (light 15 17 and dark 12 brown bars) as well as the relationship between pathologic/normal single-photon emission computed tomography (SPECT) scans (brown bars) and complication rate per year (deaths or nonfatal myocardial infarctions) in 12,360 patients. 156 For this bar graph, the finding of ischemia in SPECT scans was assigned to the stenosis class of ≥50% diameter reduction, the normal scans to <50% diameter reduction. For the coronary angiography studies, occlusions at sites of stenoses were considered. For the SPECT studies, the occlusions cannot be related to the location of stenoses; therefore the SPECT data are indirect support for the idea that low-grade and high-grade stenoses are associated with low and high risk for plaque rupture, respectively. Overall, these compiled data from almost 17,000 patients 27 31 33 156 demonstrate a considerable (i.e., exponential) increase in risk of complications with increasing stenosis severity. This graph also illustrates that the risk of a plaque rupture is not zero in mild stenoses (in line with smaller retrospective studies). Nevertheless, diagnostics in this low-risk subpopulation would require a very high “number to test” to identify patients with significant stenosis: that is, patients at increased risk.

In a large autopsy study evaluating the entire coronary tree of sudden cardiac death patients, active (vulnerable) plaques were found in ~60%, and in all patients severe stenoses (>75% diameter reduction) were observed. In another 20% healed postinfarct scares were found (most likely causing malignant arrhythmias) and in another 20% severe nonactive stenoses were present. Given these events in the development of CAD described earlier, a major goal of noninvasive imaging is to identify vulnerable plaques in the coronary circulation that are at risk for occlusion in case of rupture: that is, severely stenosing plaques. Ideally, identification of these plaques would include determination of both plaque mass (to some degree related to stenosis severity) and plaque composition. Although direct visualization of plaques in the coronary circulation and assessment of plaque components is still in an early clinical phase, assessment of stenosis degree by means of perfusion CMR has been evaluated in large clinical single-center and even multicenter trials, which convincingly demonstrated its high prognostic power to predict cardiac death and nonfatal myocardial infarction.

The Perfusion Cardiovascular Magnetic Resonance Protocol

Stress-Only Versus Stress-Rest Examination for the Detection of Hemodynamically Significant Lesions

In fundamental canine experiments, a reduction of coronary artery cross-sectional area was related to resting and hyperemic flow in the vessel. To assess the hemodynamic consequences of a coronary artery stenosis, the maximum hyperemic coronary blood flow during intracoronary adenosine infusion was measured. To compare maximum hyperemic flow in different coronary vessels, hyperemic flow was normalized with resting flow to correct for differences in mass of the vessel-dependent myocardium. Thus the assessment of hemodynamic significance of a coronary artery stenosis was achieved by dividing hyperemic blood flow by resting blood flow: that is, by calculating the coronary flow reserve (CFR). However, the influence of resting hemodynamics on resting flow, and thus on CFR, was recognized. By altering resting heart rate and blood pressure, CFR ranged from 2 to 12 in their experiments. To solve this problem, the relative CFR was proposed as hyperemic blood flow with stenosis divided by hyperemic blood flow without stenosis. This relative CFR eliminates the influence of resting hemodynamics on CFR because hyperemic stenosis flow is normalized by maximum adenosine-induced hyperemic flow that is uncoupled from myocardial oxygen demand. Consequently, this relative CFR remained stable during varying resting hemodynamics (standard deviation [SD] of 17% for relative CFR vs. 45% for the absolute CFR). These findings are in line with a recent CMR study in humans, demonstrating the important effect of resting hemodynamics such as heart rate, contractility, and loading conditions on resting myocardial blood flow and, hence, on absolute CFR. Further evidence that absolute CFR is affected by resting hemodynamics is provided by positron emission tomography (PET) studies on myocardial perfusion performed in patients with CAD. With this technique, a correlation of CFR with percent area stenosis was demonstrated; however, there was a tendency toward even better correlations for hyperemic myocardial flow alone. These data suggest that relating hyperemic flow supplied by stenosed coronary arteries to hyperemic flow supplied by nonstenosed coronary arteries (e.g., represented in a normal database) is advantageous because it eliminates the influence of resting hemodynamics as encountered with the CFR approach. This stress-only approach, which compares hyperemic stenosis flow with hyperemic flow of a normal database, has been successfully applied in several clinical CMR studies. Although the CFR measurements are affected by resting hemodynamics, two additional problems should be kept in mind: (1) matching myocardial regions, for example, the subendocardial layer, for both rest and hyperemic conditions may be difficult because geometry of the heart is changing with changing heart rate and loading, and (2) to obtain accurate results for the CFR calculation, the technique must guarantee a linear relationship between CMR-derived perfusion parameters and true flow over a wide range of flow conditions covering resting and hyperemic flow. A synopsis on advantages and disadvantages on the stress-only and stress-rest protocols is shown in Table 18.1 .

TABLE 18.1
Assessment of Coronary Artery Disease: Characteristics of Stress-Only Protocol and Stress-Rest Protocol
Stress-Only Perfusion Protocol Stress-Rest Perfusion Protocol
Acquisitions Hyperemic data only Hyperemic and rest data waiting time between acquisitions
Oxygen demand–perfusion relation
  • For resting perfusion

Not applicable Determined by “confounding factors”
  • Heart rate

  • Contractility

  • Loading condition

  • For hyperemic perfusion

Oxygen demand–supply uncoupled Oxygen demand–supply uncoupled
Analysis Only hyperemic data Two analyses (stress and rest) matching anatomy for rest/stress condition “late enhancement effect” for second acquisition
“Ideal” MR parameter–perfusion relation Linear for low perfusion range (no perfusion increase during hyperemia in myocardium supplied by significantly stenosed vessel) Linear for both resting and hyperemic perfusion, because CFR is calculated as hyperemic data/rest data
Normal/abnormal discrimination With normal database for regional hyperemic perfusion With normal database for regional CFR
Viability assessment With late enhancement acquisition advantage: acquisition window: minutes Low resting perfusion indicates scar disadvantage: acquisition window: seconds (first pass)
Assessment of perfusion and viability
  • CMR:

  • Perfusion

Hyperemic perfusion With resting perfusion yielding CFR
  • Viability

Late enhancement Resting perfusion
  • SPECT:

  • Perfusion

Hyperemic perfusion (tracer injection at stress)
  • Viability

Rest-tracer redistribution (tracer injection at rest)
  • PET:

  • Perfusion

Perfusion (flow tracer injection at stress) With resting perfusion yielding CFR
  • Viability

Rest FDG uptake (metabolic tracer injection at rest)
“Stress” indicates hyperemic condition induced by adenosine or dipyridamole.
CFR, Coronary flow reserve; CMR , cardiovascular magnetic resonance; FDG, fluorodeoxyglucose; MR, magnetic resonance; PET, positron emission tomography; SPECT, single-photon emission computed tomography.

Options for Inducing Stress in Cardiac Perfusion Studies

From the considerations mentioned earlier, it appears mandatory to apply some kind of stress to the myocardium to assess stenoses that reduce maximum blood flow, but not resting flow. If oxygen demand of the myocardium is increased either by positive inotropic agents such as dobutamine or by direct physical stress, the supply-demand imbalance induces a vasodilation and a hyperemic reaction in the myocardium to meet demand ( Fig. 18.2 ). However, in myocardial regions supplied by a stenosed coronary artery, increased oxygen demand is not met by supply, causing ischemia and consequently regional hypokinesia or akinesia. In this latter situation, assessments of both myocardial (contractile) dysfunction and reduced hyperemic reaction allow detection of a hemodynamically significant stenosis. Alternatively, hyperemia can be induced directly by administration of vasodilators such as dipyridamole or adenosine. This strategy detects coronary artery stenoses by assessment of a compromised hyperemic flow, although ischemia (O 2 supply-demand imbalance) is very rarely induced by vasodilators. To cause a steal effect by vasodilators, the stenosis has to be severe (i.e., >90%) as demonstrated in experimental studies. Consequently, when contractile function was assessed during vasodilator-induced stress, the sensitivity for detection of CAD in clinical studies was very low. Moreover, in comparison with positive inotropic approaches, the vasodilator approach as used for perfusion tests appears advantageous by reducing ischemia-associated side effects such as angina pectoris and arrhythmias. Adenosine causes hyperemia via A1 receptors, but it also induces negative chronotropic, dromotropic, and inotropic effects and, in addition, it can cause bronchospasm, atrioventricular (AV) block, and mast cell degranulation. Despite this profile, it is a very safe stressor as was shown in the EuroCMR registry with only two severe reactions in 18,840 patients. In clinical routine, an insufficient hemodynamic response to adenosine can occur. Karamitsos and colleagues proposed a high-dose regimen for adenosine (by increasing the standard dose of 140 µg/kg/min for 3 minutes to 210 µg/kg/min for 7 minutes), which was safely applied in patients with suspected CAD, increasing the proportion of adequate hemodynamic response (heart rate increase >10 beats per minute or systolic blood pressure decrease >10 mm Hg) from 82% to 98%. Conversely, dipyridamole induces hyperemia by blocking adenosine re-uptake. Finally, regadenoson represents a more selective adenosine 2A receptor, which is administered as a single intravenous (IV) bolus. This compound was shown to induce hyperemia at a similar degree to adenosine and it was also successfully used in studies on prognosis (see later).

FIG. 18.2, The two major options to provoke perfusion deficits in the presence of a stenosis. Ischemia can be detected by wall motion analysis as well as by perfusion analysis. A compromised hyperemic response is detected by perfusion analysis but not by wall motion analysis unless the vasodilator induces ischemia by a steal effect. O 2 , Oxygen.

Endogenous Versus Exogenous Contrast Media

Endogenous Contrast Media

Arterial spin labeling exploits the fact that unsaturated protons entering saturated tissue shorten the tissue T1. Using appropriate electrocardiogram (ECG)-triggered pulse sequences, T1 measurements are performed after global and slice-selective spin preparation. Absolute tissue perfusion is then calculated by assuming a two-compartment model. It should be kept in mind that this approach assumes that the direction of flowing blood in intramural vessels is orthogonal to the slice orientation, which is not the case for all myocardial layers. Furthermore, fiber orientations change during contraction and relaxation, making this approach problematic in the beating heart.

In blood-oxygen-level-dependent (BOLD) imaging, deoxygenated and thus paramagnetic hemoglobin shortens T2 relaxation time and therefore can be used as an endogenous contrast media (CM) (although oxygenated hemoglobin is slightly diamagnetic, causing less T2 shortening). A T2- or T2*-weighted pulse sequence allows for an estimation of an increased content of oxygenated blood. During pharmacologically induced hyperemia, oxygen content will increase in well-perfused myocardium but not in myocardium supplied by a stenosed vessel (which is associated with a higher O 2 extraction). However, signal differences in normally perfused regions versus hyperemic regions (with a 4-fold increase in flow) were reported to be as low as 32% in an experimental study. These BOLD data correlated closely with microsphere data, but the slope of the correlation was as low as 0.08. This limitation in signal difference might be problematic, because CM first-pass studies suggested several hundred percent of signal change being required for a reliable stenosis detection. In another study applying BOLD, ΔR2 measurements yielded an adequate slope of 0.94. But again, the sensitivity to absolute flow changes was low, because a 100% increase in flow yielded a signal increase of only 5% (i.e., a change in myocardial R2 of 0.94/s) in that study. Accordingly, in a recent study in patients, R2 increased by only 16% during adenosine stress in nonstenotic areas (although adenosine is typically increasing flow 300%–500%). The BOLD approach is also sensitive to magnetic field inhomogeneities, predominantly occurring in the posterior wall close the coronary sinus draining deoxygenated blood. Considering these aspects, the robustness of the method is not yet fully explored.

Exogenous Contrast Media for Perfusion Cardiovascular Magnetic Resonance

These CM are typically injected into a peripheral vein, and the signal change in the myocardium occurring during the first-pass of the CM is measured by a fast CMR acquisition. All CM first-pass techniques either under development or in clinical routine are designed to meet the following requirements: (1) provide high spatial resolution to permit detection of small subendocardial perfusion deficits, (2) provide adequate cardiac coverage to allow for assessment of the extent of perfusion deficits, (3) feature high CM sensitivity to generate optimum contrast between normally and abnormally perfused myocardium during CM first pass, and (4) allow acquisition of perfusion data every one to two heartbeats to yield signal intensity–time curves of adequate temporal resolution that allow for extraction of various perfusion parameters (see later). To reach these goals, high-speed data acquisition and time-efficient magnetization schemes are most important. Because the first pass of CM during hyperemia lasts only 5 to 10 seconds, breathing motion is minimized by a breath-hold maneuver, although cardiac motion is eliminated by ECG triggering. This control of cardiac and breathing motion in perfusion CMR preserves its high spatial resolution of data acquisition: on the order of 1 to 2 mm × 1 to 2 mm. This is not the case for scintigraphic techniques with acquisition windows of several minutes, which preclude breath-holding for elimination of respiratory motion, and ECG triggering during single-photon emission computed tomography (SPECT) studies requires higher tracer amounts to improve counts statistics.

Extravascular Contrast Media

For perfusion CMR techniques, the relationship between myocardial CM concentration and myocardial signal depends on a variety of factors. Normal perfusion can cause a signal increase during first pass of a gadolinium (Gd) chelate when combined with a T1-weighted pulse sequence, although a T2-weighted sequence with a higher dose of a Gd chelate can even cause a signal drop during first pass. This is fundamentally different from ischemia detection based on the assessment of wall motion, where new onset of dysfunction unambiguously indicates the presence of ischemia.

Today, extravascular Gd chelates are most commonly used for first-pass perfusion CMR in combination with heavily T1-weighted pulse sequences. These CM are excluded from the intracellular compartment (i.e., from viable cells with intact cell membranes); therefore a perfusion deficit during the first pass reflects either hypoperfused viable myocardium (which would become ischemic during inotropic stress) and/or scar tissue (with severe reduction of perfusion even at rest). To differentiate hypoperfused tissue further, it is recommended to inject another dose of CM and to wait for the establishment of a dynamic equilibrium of CM concentrations in the various compartments (blood, viable myocardium, scar tissue), to obtain conditions where tissue CM concentrations are governed by distribution volumes (and no longer by perfusion). During this condition, which typically occurs within 10 to 20 minutes after CM injection in humans, late enhancement imaging (with the inversion time set to null normal myocardium) is ideal to discriminate hypoperfused but viable myocardium as dark tissue from scar, which appears bright. For viability assessment, scintigraphic techniques also exploit the equilibrium distribution of tracers, which is observed after rest injection or rest reinjections. However, the radioactive tracers are not taken up by scar tissue, which consequently appears as a cold spot, whereas viable tissue appears as a hot spot (see Table 18.1 ).

Several multicenter studies were performed for the assessment of the optimal CM dose for perfusion imaging. This is important because higher CM doses can cause susceptibility artifacts at the subendocardium, where differences in CM concentrations between blood and myocardium are high during first pass. Strongly T1-weighted sequences showed an absence of a susceptibility-induced signal drop in the subendocardial layer up to doses of 0.15 mmol/kg of an extravascular Gd chelate, which resulted in superior diagnostic performance of doses of 0.1 and 0.15 mmol/kg versus 0.05 mmol/kg for semiautomatic analysis of the upslope parameter ( Fig. 18.3 ).

FIG. 18.3, (A) Increasing doses of the extravascular contrast medium gadolinium diethylenetriamine penta-acetic acid from 0.05 to 0.10 and 0.15 mmol/kg administered intravenously induce an increasing myocardial signal in the subendocardium (inner half of myocardial wall) during first pass. 4 (B) Even a dose as high as 0.15 mmol/kg did not cause measurable susceptibility-induced signal loss in the subendocardial layer using a hybrid echo planar pulse sequence (echo time: 1.3–2.2 ms). Accordingly, the highest dose (D3) detected stenoses with ≥50% diameter reduction significantly better than the 0.05 mmol/kg dose (D1) . ROC, Receiver operator characteristics; SI, signal intensity.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here