Ischemic heart disease (IHD) has several components that conditionate its impact in terms of physical disability, likelihood of future events, and, ultimately, death of the sufferers. These factors can be summarized as: (1) extent and severity of coronary artery atherosclerotic lesions; (2) presence and extent of infarcted myocardium; (3) degree of contractile dysfunction; and (4) extent and significance of inducible ischemia. Therapies for the disease are well established but eventually tailored to every patient in clinical practice, depending on the relative weight of each one of these elements. Essential for the efficiency of this process of management of IHD is thus an accurate and reliable clinical assessment of their particular role in every stage of the disease.

Diagnostic imaging techniques have been in use in cardiology for the last 50 years, but the process of clinical acceptance of those most recently introduced, as Cardiovascular Magnetic Resonance (CMR) and Computed Tomography (CCT), has been particularly fast, after an extensive demonstration of their value since the early 2000s. Nowadays, thus, it is accepted that a rational use of CMR and CCT may provide a comprehensive body of information on the aforementioned components of IHD, which is supported, not only by recommendations issued by committees of experts on each technique [1, 2, 3], but also by current official guidelines endorsed by scientific societies of cardiology [4].

There is little doubt that CCT has largely accomplished the expectations it aroused when it was introduced [5]. Its ability for providing the clinician with non-invasive, reliable information on the anatomy of the coronary vessels, which is the cornerstone of the diagnosis of coronary artery disease (CAD), has led cardiologists to prescribe increasingly the technique, where available, as a first line test in patients with symptoms consistent with the disease [6]. Once CAD has been diagnosed, or when chronic IHD is already known to be present, clinicians have a good deal of diagnostic methods at hand to complement the study of patients, but CMR stands out as an excellent alternative [1]. Thus, on one hand, the accurate information it provides on myocardial function and tissue composition (i.e., necrotic scar) is essential for establishing a baseline profile of every patient with IHD, which is helpful for risk stratification. On the other hand, CMR methods testing the presence of inducible ischemia constitute an important element in decision making for the most appropriate treatment in each case. The present article will review the state-of-the-art of CMR myocardial perfusion under stress agents for the assessment of this important subject in current clinical practice.

In a strict sense, myocardial ischemia occurs at any time when oxygen supply does not meet tissue metabolic demands, the most prevalent underlying cause being impaired perfusion due to CAD [7]. If an ischemic episode is long enough, a series of events are elicited which develop consecutively within minutes, constituting the so-called “ischemic cascade” [8]: thus, once regional hypoperfusion is established, myocardial dysfunction appears, first diastolic and, then, systolic, followed by ECG changes and, eventually, by the perception of anginal pain. Obviously, the maintenance of a severely reduced or abolished flow over time leads to acute myocardial infarction (MI), while a chronic state of reduced myocardial perfusion severe enough may derive into hibernating myocardium.

As the diagnosis of ischemia by direct measurement of myocardial tissue oxygenation is not feasible in clinical practice, investigators focused historically on studying aspects potentially related with, either, its cause (i.e., severity of coronary artery stenoses), or its consequences (i.e., observation of stress-induced changes from the ischemic cascade). This led to the acquisition of several concepts that are still widely in use, as is the significance of a reduction in $>$70% of the coronary lumen at angiography as a source of ischemia. This threshold was chosen as the one where the curve of increased regional myocardial flow under induced hyperemia showed a particularly steep drop, in animal experimentation [9]. The concept of coronary flow reserve (CFR) was thus coined as a referential physiological indicator of impeded flow. Alternatively, attention was put on the potential as surrogates of the ischemic process of the components of the ischemic cascade when induced by some form of stress. On these grounds, non-invasive procedures able to show the appearance of ECG-changes [10] or contractile dysfunction [11] found their place and are also routine in current practice nowadays.

Advances in both hemodynamic and non-invasive media have contributed to introduce new concepts and helped to refine the diagnosis of myocardial ischemia. On the invasive side, after warning calls upon the risk of considering exclusively angiographic measures of vessel stenoses as an indicator of significant obstruction [12], physiological invasive measurement procedures have been developed. The most relevant in practice have been the method of Fractional Flow Reserve (FFR), that estimates the pressure drop across an obstructive coronary lesion under maximal vasodilatation, and the Instantaneous free-wave Ratio (iFR), a simplified technique that does not require induced hyperemia [13]. These techniques are specific, and an invasive standard of reference, for the study of discrete epicardial stenoses, in contrast with the Index of Microcirculatory Resistance (IMR), a new thermodilution-based method for the assessment of coronary microvasculature [14], still not widely available in current practice.

The approach of radionuclide studies to the field is based on the direct relationship between myocardial blood flow (MBF) and radiotracer uptake by myocardial cells. The resulting ECG-gated Single-Photon-Emission-Computed-Tomography (SPECT) presents with 3D maps of relative myocardial perfusion allowing for the detection of potential defects when comparing stress and rest acquisitions [15], this being probably the most commonly used method of perfusion imaging nowadays. Particularly relevant in the study of myocardial perfusion is the technique that uses positron-emitting radiotracers and dedicated PET scanners: the metabolic shift in the ischemic myocardium from free-fatty acids to glucose [16] can be detected by the incorporation into this pathway of glucose-analog positron-emitting tracers, this allowing for the technique to be considered a truly metabolic one [17], which is theoretically the closest we may get to the very concept of ischemia stated above. In addition, and most important, the use of PET perfusion tracers with appropriate rates of blood extraction and myocardial retention and, thus, a nearly lineal relationship between MBF and the measured tracer activity, permits an absolute quantification of regional MBF, in terms of volume of flow per time and left ventricular mass units (mL/min/g) [18], which is the formal expression of myocardial perfusion. Although not widely used in current clinical practice for logistical reasons, Rest and Stress MBF and the derived Myocardial Perfusion Reserve (MPR) by PET are the standard of reference for the non-invasive quantitative assessment of coronary artery function [19].

The study of myocardial perfusion by CMR was introduced during the early years of the short history of the technique itself [20], and the approach then proposed has not changed essentially. Here, as in nuclear scans, it is the kinetics of a tracer, an MR contrast agent (CA) which is monitored during its first pass through the heart chambers and, finally, the left ventricular myocardium, that enhances its signal as the agent reaches the microcirculatory network of the heart muscle (Fig. 1).

The most immediate and commonest method of analysis is the visual detection of a distinctive deficient increase in signal intensity of a region of the left ventricular myocardium during the first passage of a CA bolus under vasodilatory conditions lasting for at least 3 consecutive frames of the sequence. This perfusion defect should not be present at rest and should not correspond to an area of previous MI. The detection of such a defect is assumed to be due to a reduction in CFR of this territory which, when attributed to a particular coronary artery [39], allows for the conclusion of the presence of a significant epicardial stenosis in the vessel (Fig. 6).

Criteria for an appropriate interpretation of dark, unenhanced regional myocardial signals as due to perfusion defects have been issued by scientific societies [40]. According to these recommendations, it is accepted that a true defect (1) appears when contrast arrives at the left ventricular myocardium, (2) persists for several cardiac cycles, (3) is not restricted to a thin linear contour, (4) it is more prominent at the subendocardial level, extending variably through the whole thickness of the myocardium, (5) it is not present at rest, and (6) corresponds to a distribution territory of a coronary artery.

Perfusion defects in different territories can be detected in case of multivessel disease (Fig. 7) and, in patients with severe 3-vessel CAD, the perfusion study may show a global, intense, persistent defect which is readily recognizable (Fig. 8).

A not infrequent finding is a diffuse, annular, transient, subendocardial defect not present at rest (Fig. 9), which, in the absence of any recognizable obstructive lesion in an epicardial coronary artery, has been deemed to correspond to diffuse microvascular dysfunction (MVD) [41, 42]. CMR perfusion is sensitive to both the downstream effects of discrete epicardial stenoses and those due to disturbances of coronary microcirculation, and subendocardial hypoperfusion is on the pathophysiological basis of each one. For this reason, theoretically, a global, inducible subendocardial defect, could also be due to multivessel epicardial disease with balanced ischemia. In practice, however, this latter instance presents with extensive but heterogenous defects in terms of intensity, persistence and transmurality and, not rarely, also with perfusion defects at rest (Fig. 10A). Importantly, an accompanying induced contractile disfunction may be found in particularly impaired regions (Fig. 10B), a finding which is never seen in MVD.

Important for a reliable identification of perfusion defects is the recognition of artifacts that may mimic a true defect which, unfortunately, are frequent, even with advanced MR systems. The most relevant is the so-called “dark-rim artifact”, which consists on a subendocardial, sharp linear signal of very low intensity that is due to magnetic susceptibility effect, to spatial resolution, and/or to motion effects [43, 44]. This dark band may appear as soon as the contrast arrives to the left ventricular chamber, is usually very thin (1–2 pixel wide), transient, and can be present at both the rest and stress studies [40] (Figs. 11,12). Despite these hints, cases where an artifact occasionally challenges the judgement of the observer still persist (Fig. 13).

Particularly important, in practice, is the issue of perfusion studies in patients with previous MI. In theory, a basal rest perfusion study should exhibit a reduced signal intensity in the infarcted area, as is the case in the acute phase of a large transmural MI (Fig. 14A). However, not infrequently, the presence and extent of a rest perfusion defect does no correlate with the actual area of necrosis in either, acute (Fig. 14B) or chronic MI (Fig. 15). Signal intensity in areas of previous MI is thus, not dependent merely on the presence of scar tissue, but some other factors lead to differences in CA concentration between MI and remote regions and, in this sense, rest perfusion CMR cannot be equated to SPECT studies, where rest defects are directly related to the presence of infarcted myocardial tissue not amenable to radiotracer uptake [45].

The presence of inducible residual peri-infarction ischemia in a territory with a previous MI has been addressed by the consideration of the extension of a stress perfusion defect in comparison with that of the LGE. Those defects with equal size on both sequences would be labelled as “fixed” (Fig. 16), while those perfusion defects exceeding the area of MI would be considered as “partially reversible” (Fig. 17), independently of rest perfusion studies. The diagnostic performance of the latter finding for the diagnosis of a residual significant stenosis in the infarct-related artery has been found to be adequate, though a good deal of cases with apparently fixed defects had also significantly obstructed arteries [46].

When considering both, the variability of rest perfusion in the presence of previous MI, and the rather low sensitivity of partially reversible defects, the interest in knowing the actual extension of fixed defects emerges. An option to address the issue is to invert the order of perfusion studies in patients with previous MI, performing the rest study first. This allows for a reliable delineation of the fixed defect that can be easily subtracted from the inducible one, when present (Fig. 18). Moreover, in the case of patients with advanced, severe CAD, in whom a combination of infarcted regions and underperfused territories can be expected, a basal rest perfusion study followed by the stress one is even more helpful (Fig. 19). Certainly, the presence of CA in an area of MI after the first rest perfusion study may contaminate the SI of the second one, but this does not preclude the detection, or exclusion, of newly induced defects (Fig. 20).

Studies of validation of visually estimated myocardial perfusion CMR for the detection of significant angiographic coronary stenoses started in the early 2000s with encouraging results [63, 64], despite the limited strengths of the images in comparison with current sequences (Fig. 31). Interest in the technique prompted a good deal of studies with a more reliable standard of reference as is invasive FFR, which have been repeatedly submitted to metanalysis [65, 66], showing pooled values of sensitivity and specificity on the range of 0.90. Qualitative perfusion CMR has shown to compare favorably with SPECT for the diagnosis of CAD in the multicenter study MR-IMPACT [67], a finding confirmed in the landmark CE-MARC study [68], from which a strong predictive value of perfusion for events was also shown [69]. The fairly higher spatial resolution of CMR compared with SPECT (2–3 mm vs 10 mm, respectively) may account for these results [68]. The usefulness of Perfusion CMR for risk stratification and prognosis has been also proven in other multicentric studies [70, 71, 72], and even shown to be noninferior to invasive FFR in decision-making upon revascularization in patients with CAD randomly assigned to either CMR or FFR diagnostic strategies [73]. A derived benefit is the favorable cost-effectiveness profile of those protocols where CMR is introduced in the diagnostic workup [74], a finding supported by data from large, “real world” registry data showing advantages in this respect for CMR in comparison with other techniques [75].

Semi-quantitative methods of analysis of CMR perfusion have also a well-founded basis on early studies of comparison with PET [76], and have shown similar results to qualitative methods in meta-analyses of pooled data [77]. Also, a proven prognostic value has been found in follow-up studies [78] but, as stated above, semi-quantitative analysis techniques have not entered the field of clinical practice due to the complexity of measurements they require.

Although introduced more recently, absolute QP using the dual-sequence approach [56] has been the subject of intense clinical investigation, starting with a comparison study with PET [79], still the recognized reference in myocardial blood flow quantitation, with which CMR showed an excellent correlation. Feasibility and good reproducibility of studies were established [80], the automatic in-line processing of data emerging as a unique tool removing the need for complex and time-consuming analysis of previous methods. A study comparing QP with quantitative coronary angiography found excellent diagnostic performance of the technique using thresholds of regional stress MBF of 1.29 mL/min/g and MPR of 1.47 for the detection of significant obstructive lesions ($\ge$70%) [81]. With the reference of 3D reconstructed quantitative coronary angiography, an optimal diagnostic threshold of 1.5 mL/min/g of regional stress MBF was found, again for the detection of $\ge$70% coronary stenoses [82]. A key investigation where QP was performed in patients in whom invasive measures of coronary physiology (FFR and IMR) were available [83] established an optimal cut-off point for the detection of obstructive CAD (FFR 0.80) of 1.94 mL/min/g for regional stress MBF. Additionally, global stress MBF $>$2.25 mL/min/g was shown to be able to distinguish normal from abnormal coronary flow (either due to obstructive epicardial lesions or to MVD) and, to distinguish three-vessel disease from diffuse MVD, a cut-off point $\le$1.82 mL/min/g was established. Advanced CAD frequently involves one or more epicardial obstructive lesions and MVD, either regional or global, and although cut-off values of stress MBF are progressively reduced, as observed, from normal perfusion to MVD and, then, to severe multi-vessel disease, overlapping is common in practice. Useful in this sense is to combine perfusion mapping data with visual assessment of defects into a diagnostic algorithm [83]: in brief, when a visual regional defect accompanies a matching reduced stress MBF ($\le$1.94 mL/min/g), obstructive epicardial coronary artery is the cause (Fig. 30); in the absence of a visible regional defect, a normal global stress MBF ($>$2.25 mL/min/g) would indicate normal coronary function (Fig. 32), while a global MBF below this value would suggest MVD (Fig. 33); finally, significant three-vessel disease should exhibit global stress MBF $\le$1.82 mL/min/g and, importantly, a visual defect globally distributed (Fig. 34).

Multivessel disease is a scenario where the diagnostic performance of perfusion studies visually assessed is particularly challenging. Defects from different territories are not necessarily homogenous (Figs. 7,8,10), and CMR, although superior to SPECT, fails to identify visual defects in all territories in up to a third of patients with three-vessel disease [84]. This figure was shown to be as low as one-half in a subsequent study where QP was fairly more accurate (87%) than visual assessment for correctly identifying three-vessel disease when any coronary territory with $\ge$2 adjacent myocardial segments presenting with regional stress MBF 1.94 mL/min/g was considered as ischemic [85]. Based on this cut-off value [83, 85], an additional information derived from the dual sequence QP application is the calculation of a global ischemic burden with values of MBF assigned to every coronary territory (Fig. 35), which can be useful in the estimation of extension of the process.

MVD is another entity where the contribution of perfusion mapping has been essential, as the disorder is frequently overlooked due to the lack of appropriate, non-invasive, reliable tests for its assessment. While patients with MVD present with subendocardial hypoperfusion [41], a circumferential defective enhancement at first pass is detected at visual analysis of perfusion CMR only in a proportion of them (Fig. 9). A first study using QP CMR on patients with angina and no obstructive CAD showed that stress MBF and MPR were significantly reduced with respect to a control group of normal individuals [86]. A further study comparing QP CMR with invasive measurements of CFR confirmed these findings [87] establishing as optimal thresholds for the diagnosis a global MPR of 2.19 and, particularly, a subendocardial global MPR of 2.41. Interestingly, only 58% of patients with proven MVD presented with visually detectable defects in this study.

The prognostic value of perfusion mapping with automatic in-line flow measurement has been studied on a large population of 1049 patients with known or suspected CAD followed-up for a median of 605 days [88], showing that stress MBF and MPR are strong, independent predictors of adverse cardiovascular outcomes. Interestingly, MPR was superior to stress MBF in predicting death (although not a composite of events), an observation in accordance with a previous PET study [89] on a population of 4029 patients with a median follow-up of 5.6 years, where CFR (the equivalent of MPR) was a stronger predictor of cardiovascular mortality than maximal MBF. These findings seem to focus on the impaired vasodilator capacity, expressed by CFR or MPR, as a more relevant mechanism than peak MBF in estimating CAD risk.

In the subset of patients with previous CABG, a study of 341 patients with a mean follow-up of 1.7 years proved that both stress MBF and MPR independently predict adverse outcomes, with a cut-off point of 1.48 mL/min/g of mean stress MBF [90].

Although unrelated with MBF, an additional parameter of interest that is automatically estimated in-line with the dual-sequence QP is the time interval for the contrast-bolus to pass from the right to the left circulation, or pulmonary transit time (PTT) (Fig. 36). As a measure of global cardiopulmonary function, PTT has shown to be independently associated with cardiovascular events [91], with longer values ($>$8 sec) observed in patients with higher incidence of events. PTT can thus be considered as a biomarker with prognostic information that is available in every study of QP without user interaction.

Attempting the relief of ischemia by revascularization, either PCI or coronary artery by-pass grafting (CABG), seems to be a matter of straightforward common sense and, in fact, it turned into a paradigm in clinical cardiology in the past. Nevertheless, the role of inducible ischemia as a target for a prompt intervention has been challenged thereafter by results from several large, randomized trials failing to show an improved long-term outcome of patients with CAD and objective evidence of myocardial ischemia when treated by PCI compared with medical management [92], even when the presence of more than mild ischemia had been proven on stress testing [93]. In contrast, several other studies of equally high level have shown that choosing to intervene only on vessels with definite flow obstruction by invasive FFR leads to a lower rate of events [94], and that this benefit persists over time [95].

Far from underestimate the value of ischemia, these disparities can also be interpreted as indicating a need for a refinement in its grading, an aspect where CMR perfusion methods discussed here could play a potential role. Of note, some of them, as QP, were not even available at the time these important studies were carried out.

In current practice, however, we have to refer to the present position of CMR perfusion in scientific guidelines for the management of patients with CAD. To be considered, first of all, is the fact that CMR is not treated particularly in most of them, but is included into the group of “non-invasive functional imaging”, together with SPECT, PET, and stress ECHO. Certainly, the usefulness of each of these techniques has been extensively proven, although the study of their relative value has shown a somewhat higher diagnostic accuracy of PET and CMR in comparison with SPECT and ECHO [96]. Logically enough, local expertise and availability of tests also should play a role in the selection.

This considered, non-invasive imaging merits a Class IB indication for initial diagnosis of CAD in symptomatic patients in whom obstructive CAD cannot be excluded by clinical assessment alone. Either, functional or anatomic (CCT) tests are valid options, their selection depending on the pre-test likelihood of the disease [4].

The precise term “myocardial ischemia” is referred to in the guidelines where a consideration of non-invasive tests is recognized (Class IIB) for its assessment, together with myocardial viability, in patients with CAD who are considered suitable for coronary revascularization [97] and, more straightforward, where a large area of ischemia by functional testing is considered as Class IB for prognostic purposes in revascularization planning [98]. From this last source, a list of “ten commandments” is issued [99] where two of them are particularly relevant in regard to myocardial ischemia: (1) “Objective evidence of myocardial ischemia by non-invasive stress imaging and/or intravascular assessment of the functional relevance of coronary artery stenoses are needed to indicate myocardial revascularization through PCI or CABG and to select the appropriate targets for PCI”; and (2) “With large areas of inducible myocardial ischemia or relevant LV systolic dysfunction myocardial revascularization through CABG or PCI is indicated to improve long-term survival”.

Importantly, a first direct reference to QP CMR appears in the most recent version of the American guidelines for management of chest pain [100], where stress CMR with the addition of MPR measurement is recognized as reasonable (Class IIA) for patients with persistent stable chest pain and nonobstructive CAD to improve diagnosis of coronary myocardial dysfunction and for estimating risk of events.

Conventional CMR is, thus, well positioned, among the diagnostic methods of ischemia. Advances in this area mainly relate to whole heart acquisition using 3T MR systems [101], with the derived increase in spatial coverage of perfusion studies.

Moreover, the availability of a reliable method for QP offers a wide range of new perspectives that will arguably lead to conceptual changes on the issue and, eventually, also into the adoption of new paradigms. Automatic workflow with pixelwise myocardial perfusion maps has immediate advantages in terms of reproducibility and objectivity avoiding the recognized drawback of visual perfusion analysis based on the heterogeneity of SI among different myocardial regions within the same study plane, where those taken as a reference are not invariably perfused normally. This will probably help to the recognition of “new” or insufficiently known aspects of CAD: subclinical, MVD, or relative regional ischemic burden in multivessel disease, all of them with a potential impact on the whole spectrum of IHD that deserves to be studied, and that may help to explain why the disease frequently ignites and escalates in an unpredictable way. Important, in this sense, is the potential of hybrid PET-MRI systems in combining molecular and structural imaging [102], where the information from PET on myocardial and, particularly, coronary artery inflammation [103], can be added to the rest of data from CMR imaging.

Also, and most important, a great change is coming for the CMR practitioner with the routine use of automatic QP applications, where a reliable set of automatically obtained numerical data must be integrated with the conventional reading of perfusion imaging [104]. Similar processes of automation have been also extended to other measurements of CMR [105], and a near future of studies fully based on outputs obtained from machine learning can be anticipated. The integration of this information requires both, a recognition of the limitations of our subjective analysis and a critical, well-informed evaluation of quantitative data. When considered into the clinical context of every patient, this new mode of information will prove beneficial in terms of diagnosis, prognosis, and therapy planning in IHD.

In conclusion, the assessment of myocardial perfusion is an essential part of a CMR study in patients with proven or suspected coronary artery disease. It has shown fairly good accuracy for the detection of the disease, compares favorably with other methods, and is useful for the prognostic stratification of patients. The recent introduction of a reliable technique of quantitative analysis with automatic, user-independent, processes of calculation of myocardial perfusion (in mL/min/g), myocardial segmentation, and allocation of values, constitutes a major step forward toward the consideration of CMR Perfusion as a unique diagnostic tool in the study of patients with ischemic heart disease.

GPL and PK conceived the manuscript, performed the writing, and approved the editorial changes.

Not applicable.

Not applicable.

This research received no external funding.

The authors declare no conflict of interest.