CMR of microvascular obstruction and hemorrhage in myocardial infarction

Microvascular obstruction (MO) occurs in the setting of reperfusion following prolonged myocardial ischemia and provides incremental prognostic information beyond infarct size, to which it is related. MO is characterized by a number of ultrastructural and functional changes at the microvascular level. Understanding these histopathophysiologic changes can inform the approach by cardiovascular magnetic resonance (CMR) to detecting MO and our interpretation of the results and their subsequent clinical implications. It can also help potentially improve how MO is assessed by CMR which has implications with regard to our understanding of infarct evolution and healing as well as the evaluation of the efficacy and mechanisms of action of adjunctive reperfusion agents. This review will discuss the various pathophysiologic derangements, both anatomic and functional, seen at the microvascular level after myocardial reperfusion; current and evolving CMR techniques to assess MO and the related phenomenon of myocardial hemorrhage; the temporal evolution of MO and its relationship to infarct healing, adverse LV remodeling and clinical prognosis; clinical trial applications; and future directions.

The contribution of microvascular injury in causing anatomic myocardial “no-reflow” was first described in the 1970’s[1‐3]. At that time, it had already been recognized in other organs that despite recanalization or restoration of arterial flow to an ischemic tissue bed, perfusion at the tissue level may not uniformly occur. In a canine model of ischemia-reperfusion, Kloner et al. studied the evolution of myocardial injury using a combination of electron microscopy and the vital dye, thioflavin-S, which stains intact endothelium and thus causes myocardial tissue receiving adequate capillary flow to fluoresce under UV light[1]. Epicardial coronary reperfusion after prolonged ischemia resulted in subendocardial regions of nonfluoresence (Figure1) representing no-flow or low-flow which were confirmed to have substantially reduced regional blood flow measured by quantitative radioactive microsphere analysis[4, 5]. Because of the wavefront of necrosis, the endocardium is the most vulnerable area for ischemic damage.

Under electron microscopy, these areas of nonfluorescence were characterized by striking ultrastructural changes to the microvasculature in addition to evidence of irreversible myocardial cell injury (Figure2). Under normal circumstances, the capillary lumen is widely patent and lined by intact endothelial cells. In contrast, within the nonfluorescent, thioflavin-negative regions following reperfusion, microvessel integrity was compromised with intraluminal obstruction of the lumen caused by regional swelling of the endothelial wall as well as intraluminal protrusions and blebs of the endothelial cytoplasm. Plugging of the microvessels by erythrocytes, platelets, and neutrophils was also seen. Also contributory was external compression of the capillaries by edematous myocytes and subsarcolemmal blebs. Based on these experimental observations, regions of microvascular injury characterized by the aforementioned ultrastructural anatomic abnormalities and reduced regional blood flow following reperfusion were termed anatomic “no-reflow”. The quantitative extent of the anatomic no-reflow zone is determined by how well the vital staining dye penetrates into the ischemic tissue bed.

Although the exact mechanisms and time course of events leading to anatomic no-reflow are unknown, it is hypothesized that a number of interrelated factors contribute[3, 6]. Direct endothelial damage caused by ischemia which leads to the observed protrusions of the endothelial surface that encroach into and compromise the capillary lumen likely play a direct role in the perfusion deficits. Further plugging of the lumen by erythrocytes, fibrin, and platelet thrombi suggest that reperfusion leads to some degree of initial flow into the microvessels which is later compromised. Leukocyte plugging is also observed and this may contribute not only to mechanical obstruction but may also lead to reduced endothelial dependent and independent vasodilation through oxidative stress pathways. Neutrophils elaborate reactive oxygen species that can impair endothelial and platelet function. Oxygen free radicals can also promote the synthesis of tissue factor which may accelerate the accumulation of fibrin/fibrinogen.

In humans, clinical no-reflow is characterized by the additional important contributing factors of microembolization and the resultant inflammatory response[6, 7]. Rupture or erosion of epicardial coronary artery plaques, either spontaneously or induced mechanically by percutaneous coronary intervention, can lead to distal showering of embolic debris, consisting of atherosclerotic plaque components and/or thrombotic material. Distal embolization may not only contribute to mechanical obstruction of the microvessels but also causes an inflammatory response with the elaboration of vasogenic and thrombogenic factors that further exacerbate existing microvascular dysfunction. Hence, pathophysiologically, no-reflow is the result of the complex interplay of a number of related processes.

The advent of fast CMR techniques in the 1990’s facilitated the study of the temporal perfusion patterns within acute reperfused infarcts following bolus administration of gadolinium by allowing a temporal resolution of seconds rather than minutes[8‐10]. In seminal publications involving an experimental animal model and a parallel study of patients with acute reperfused infarcts[5, 11], regions of myocardial hypoenhancement within the first 2 min of contrast administration were observed within the infarct region that were characterized by significantly reduced regional blood flow and correlated in size with anatomic no-reflow zones measured by thioflavin-S (Figures3 and4). Subsequently, these hypoenhanced areas seen on CMR were termed microvascular obstruction (MO) and since they were measured within 1–2 min of contrast administration, represent “early” MO. The extent of early MO by CMR was found to correlate closely with the size of the anatomic no-reflow region by thioflavin (r = 0.91) but was significantly smaller, measuring 61% of the pathologic region[12]. Regions of CMR MO most closely approximated regions within the infarct corresponding to <40% of remote myocardial blood flow at 2–9 days after reperfusion, compared to thioflavin-S negative regions which corresponded to regions with <50% of remote flow[12] . In these original studies of CMR MO, a magnetization-driven spoiled gradient-recalled echo pulse sequence was used that did not require adjustment of the inversion time[13].

There have since been numerous additional approaches to assessing CMR MO in the setting of improved pulse sequences. First-pass perfusion sequences tracking the first minute of gadolinium arrival and distribution within the myocardium have been used to detect MO[15‐17] (Figure5). However, these original sequences suffered from relatively low spatial resolution (~2.7 × 3.4 mm) and reduced LV coverage (maximum of 3–4 slices) which reduce diagnostic sensitivity. Subsequently, with the routine use of ultrafast inversion-recovery gradient echo sequences for infarct delineation, MO is increasingly assessed on late gadolinium enhanced (LGE)(Figure5) as well as early gadolinium enhanced (EGE) images[18‐20] with the advantage of improved spatial resolution as well as complete LV coverage. Assessing MO on EGE requires a fixed, high inversion-time (TI) and is generally performed at ~2-4 min following contrast administration (Figure6). With the recent advent of accelerated high-resolution perfusion sequences (1.5 mm × 1.5 mm spatial resolution, 8 slices acquired over 2 R-R intervals)[21], it is now possible to detect true first pass MO and quantify it for the entire LV (Figure6).

Because gadolinium is not a pure intravascular agent and extravasates into the interstitium within minutes following administration, it gradually diffuses into the initially hypoenhanced zone. Hence, the size of MO decreases over time from contrast administration, the largest being during first pass and the smallest during LGE (the latter also termed “persistent” MO in the literature). Although there is a high concordance and correlation amongst the 3 measures of MO (r = 0.91 between FP and EGE; r = 0.55-0.78 between EGE and LGE MO)[16, 21], MO will be detected more frequently on FP than by EGE and LGE (incidence of 22% by FP vs. 14% by LGE in one study[21]). Amongst the 3 measures, there was the least variability in MO quantification using the LGE method[21]. First-pass MO may be overestimated by the concomitant presence of significant epicardial coronary stenoses which impede contrast delivery. Late or persistent MO presence and extent will depend upon the timing of image acquisition following gadolinium administration, particularly in the presence of less severe microvascular dysfunction.

MO extent varies as a function of time from the acute ischemic event. Experimentally, it has been demonstrated that there is an expansion of the anatomic no-reflow area by thioflavin-S in the hours following reperfusion, with a tripling in size between 2 min and 2–4 h and a further smaller increase up to 8 h following reperfusion[4, 22, 23]. Following an initial hyperemic phase within the first 2 min of reperfusion, there is a marked, progressive decline in myocardial blood flow which plateaus at around 50% of normal flow, which supports these findings. Experimental studies using CMR, in addition to thioflavin staining and radioactive microspheres to measure blood flow, have also suggested that MO extent increases from 1–2 h to 24 h post-reperfusion and may further increase up to 48 h post-reperfusion[14, 24, 25]. However, there have not been any confirmatory studies in humans to demonstrate expansion of MO at any timepoint post-reperfusion.

It is increasingly apparent that the temporal course of MO may vary in the subsequent days to weeks as infarct healing ensues and may in fact influence LV remodeling. Both animal models and humans studies, using both CMR and myocardial contrast echocardiography to assess MO, demonstrate that microvascular dysfunction may persist up to 1 month post-reperfusion[26‐28]. Persistence of MO at 1 month was associated with worse regional wall motion, scar thinning, and infarct expansion[26, 27]. In a rat model, at 4 weeks post-reperfusion, regions of anatomic no-reflow remain visible and are associated with a markedly reduced capillary density and decreased myocardial blood flow[27]. A recent study using a porcine infarct model lends further insight by showing that histopathologic evidence of obstructed capillaries filled with necrotic debris can be seen at 2 weeks but the infarct is replaced by extensive collagen deposition and fibrosis at 6 weeks[29]. However, particularly in patients, microvascular dysfunction may in fact resolve much earlier than 1 month (Figure7), as several CMR and myocardial contrast echocardiography studies have shown, and earlier resolution appears to correlate with improved functional recovery post-MI and outcome[28, 30‐32]. The severity of MO likely affects its degree of persistence and may account for conflicting data regarding temporal changes in the size of MO in experimental studies with some studies showing no significant change in MO extent between 2 and 9–10 days[12, 25] and others in which MO appeared smaller at 1 week[29, 33]. Also contributory may be the complex pathophysiologic mechanisms of MO that can be differentiated as “structural” versus “functional”[34]. “Structural” or irreversible alterations to the microvascular bed can be characterized by the luminal obstruction by cellular debris and endothelial wall damage. In humans in particular, there may be a larger contribution of “functional” or potentially reversible changes such as microvascular spasm and microembolization which promotes vasoconstriction via the elaboration of vasoactive and proinflammatory mediators[35] and may result in reduced myocardial perfusion reserve which can be potentially assessed by adenosine myocardial perfusion[36, 37]. The time courses of structural and functional no-reflow may thus differ and explain in part, variability in the time course data. Functional no-reflow may perhaps resolve more quickly as vasoactive/proinflammatory mediator levels return to normal post-infarction compared to structural no-reflow that requires prolonged infarct tissue healing.

With the caveat that many of the studies were relatively small, which limits the robustness of multivariate modeling, CMR MO has been found to be predictive of clinical outcome (Table1), independently of or when adjusted for other indices such as infarct size and LVEF. This is consistent with the results of studies using other imaging technique such as myocardial contrast echocardiography and coronary angiography to assess no-reflow[23, 38‐42]. Many of these patient outcome studies also showed a relationship between presence of MO and adverse LV remodeling (Table2) with reduced global systolic function and larger LV volumes at follow-up (Figures8 and9), suggesting a possible mechanism for the poor prognosis. Experimental studies have supported these observations. In an animal model, a larger region of anatomic no-reflow was associated with thinner infarct walls and worse infarct expansion[27]. Theoretically, obstruction of blood flow to the infarct core could limit the delivery and transit of cellular components such as macrophages required for phagocytosis of cellular debris and nutrients needed for optimal infarct healing[23]. Additionally, obstructed microvessels may limit the future potential for collateral blood flow development[23].

An experimental model using myocardial tissue tagging suggests a possible mechanism linking MO to poor infarct remodeling: MO may cause heterogeneity in local myocardial tissue properties across the infarct wall that could adversely affect wall stress and subsequent infarct healing[25]. In a canine model of acute reperfused MI, myocardial deformation was measured at 4–6 h, 48 h, and 10 days post-reperfusion. MO extent, measured both early (4–6 h post-reperfusion) and at 48 h, was a stronger predictor of increases in LV volumes than infarct size. There were demonstrable changes in myocardial deformation up to 48 h post-reperfusion in the infarcted and adjacent myocardium based upon the amount of MO, expressed as a percent of infarct size. Compared to animals with MO comprising <35% of the infarct volume, in those with MO comprising >35% of the infarct volume, infarcted myocardial segments exhibited significantly less stretching in the longitudinal direction; noninfarcted, adjacent regions exhibited reduced radial thickening; and there was a reduction in the first principal strain. This suggests that infarcts with greater amounts of MO are characterized by increased stiffness and reduced myocardial deformation in both the necrotic region and also the adjacent, noninfarcted region. Changes within the infarct region were seen at 4–6 h and preceded the alterations in the adjacent segments, which were not observed until 48 h post-reperfusion. Thus, increased myocardial stiffness and reduced elasticity in the infarcted regions with large amounts of MO could result in increased local wall stress that subsequently limits systolic shortening in adjacent segments, all of which may lead to adverse LV remodeling.

Most of the human studies showing a positive independent association between MO and clinical outcome tended to measure MO within 3–7 days post-reperfusion and utilized the LGE technique for assessment of MO. There have been a few studies that have suggested that MO is not independently predictive of outcome, after controlling for infarct size[47, 48] (Table1). In one of the studies, MO was measured relatively early post-infarct (median of 4.5 h) which may have underestimated the true incidence of MO since MO likely increases in size over the first 24–48 h post-reperfusion. In the other study, the study results may have been influenced by heterogeneity in the times to actual epicardial revascularization which will affect the evolution of MO[51].

As part of the disruption to the microvasculature observed upon reperfusion, large gaps can be seen in the endothelial wall that cause extravasation of red blood cells, i.e. hemorrhage. In experimental models performed as early as the 1980’s, it was observed that any hemorrhage was limited to the region of severe microvascular injury,. lagged behind the no-reflow process and its extent was directly proportional to the duration and severity of the preceding ischemia (with the severity of ischemia determined predominantly by collateral flow)[22, 52‐58]. Specifically, myocardial hemorrhages occurred when myocardial blood flow during the time of coronary occlusion prior to reperfusion was <21% of control[54]. Macroscopic hemorrhage does not occur in the absence of reperfusion, i.e. in the presence of a persistent coronary occlusion[55] (Figure10). The extent of hemorrhage has been found to be highly correlated with pathologic infarct size (r = 0.90) and increases in proportion to the occlusion time prior to reperfusion but is independent of thrombolytic therapy (i.e. a lytic state does not increase the amount of hemorrhage seen)[59‐61].

Hemorrhage can be assessed using T2-weighted and T2* imaging. The appearance of hemorrhage on MRI is based upon the paramagnetic effects of hemoglobin degradation products[63, 64]. Initially, hemorrhage may consist of oxyhemoglobin which lacks paramagnetic effects. Subsequently, probably within the first few days after acute MI, oxyhemoglobin denatures into deoxyhemoglobin which does exert paramagnetic effects and will significantly reduce the T2 time. Deoxyhemoglobin is then gradually converted over the following few days into methemoglobin, which is strongly paramagnetic with respect to both T1- and T2-weighted sequences. After ~2 weeks, methemoglobin is converted into hemosiderin which is contained within macrophages and also results in low T2 values. Hence, acutely post-MI, hemorrhage can be visualized on CMR as hypoenhanced regions surrounded by elevated signal intensity representing myocardial edema on T2-weighted imaging and most studies show that hemorrhage is limited to those patients with evidence of MO[26, 65‐67]. Evaluation of human autopsy specimens in 2 cases showed a close correlation between pathologic hemorrhage and hypoenhanced signal intensity on T2 sequences[68]. However, MVO without hemorrhage can also result in hypoenhanced regions seen on T2-weighted sequences[69] and both can appear as rims of persistent hypoenhancement on LGE[70, 71] (Figure11). A potentially promising approach in differentiating MVO with and without hemorrhage could be the method of direct T2 quantification, which has yielded more robust findings for the detection of myocardial edema compared to conventional T2-weighted short tau inversion recovery[72, 73]. Within the infarct core, T2 mapping demonstrates significant reductions in T2 values corresponding to LGE MO (Figure11). Whether or not such a technique can also differentiate between MO with and without hemorrhage by detecting gradations in reduced T2 values requires further investigation.

As opposed to T2 relaxation which depends primarily on spin-spin interactions, T2* decay is caused by a combination of spin-spin relaxation and magnetic field inhomogeneity[74]. Because iron deposits and blood products cause magnetic field inhomogeneities, T2* imaging has been used to measure myocardial iron deposition in the setting of iron overload cardiomyopathies and recently applied to detecting hemorrhage in the setting of acute reperfused MI[75‐78]. Since the paramagnetic effects of hemoglobin byproducts on T2* is even stronger than on T2, T2*-weighted sequences should generally exhibit higher sensitivity for hemorrhage[78‐80]. In a recent experimental canine model, T2* imaging was compared to pathologic analysis of the infarct region at day 3 post-reperfusion[76]. Gross hemorrhage was limited to those animals with the largest infarct sizes and extents of MVO. Pathologically, all animals with hemorrhage had evidence of MVO. There was high concordance between hemorrhage detected pathologically and that detected by T2* imaging (kappa = 0.96, P < 0.01). The size of the hemorrhagic area was highly correlated with that measured pathologically (r = 0.91, p < 0.01), though CMR overestimated the extent (mean difference of 2.4 ± 1.4%). The amount of hemorrhage was always smaller than the amount of MVO and was contained within it. Use of T2* sequences has the potential to improve the differentiation of MVO alone from MVO with hemorrhage: hemorrhage will be accompanied by reduced, below normal T2* levels whereas MVO alone will not[71] (Figure12).

The time course over 6 weeks of MO and hemorrhage was further examined recently in a porcine model of acute reperfused myocardial infarction[29]. MO was assessed using EGE and hemorrhage by T2* mapping at 2 days and 1, 2, 4, and 6 weeks post-reperfusion. MO extent was the largest at 2 days with gradual reduction over the ensuing 2 weeks and complete resolution by week 4. T2* values reflecting hemorrhage followed the same temporal course with the lowest T2* levels seen acutely in the hemorrhagic core which gradually normalized to control levels at 4 weeks. Detailed histopathologic analysis was also performed and compared to the CMR findings. At day 2 post-infarct, T2* signal voids corresponded to histopathologic areas of extensive hemorrhage within the infarct. At week 2, histologic sections within the MVO region continued to show obstructed microvessels with necrotic debris and macrophage infiltration. Iron deposition within the MVO region corresponded to T2* signal voids at this time point. By 6 weeks post-MI, the infarct region was characterized by extensive collagen deposition and myocardial fibrosis with no evidence of iron deposition consistent with resolution of the hemorrhage and MVO process. In this study, only infarcts with hemorrhage were studied.

Subsequent clinical studies have supported these experimental findings. In patients, the extent of late MO has been found to be highly correlated with the size of the hemorrhage area (r² = 0.87, p < 0.001) and interestingly, the correlation with early MO was less robust (r² = 0.30, p < 0.003)[78]. Several other studies have shown that the presence of hemorrhage is a strong predictor of adverse LV remodeling and MACE (Table3). Thus, the data suggest that hemorrhage reflects more severe, and possibly irreversible, persistent, “structural” rather than potentially reversible, “functional” microvascular injury and its resolution parallels that of anatomic MVO as infarct healing ensues. It may thus be a more specific marker of adverse events because it reflects irreversible, structural injury to the microvasculature, but this requires further investigation and confirmation.

MO is increasingly incorporated into clinical trial methodology as a surrogate clinical outcome in studies investigating adjunctive reperfusion agents and strategies, particularly when the treatment has the potential to directly impact microvascular function[82‐92]. On the web-based clinical trial registry,http://​www.​clinicaltrials.​gov, there are currently 25 open studies in which microvascular obstruction is an endpoint and there are an additional 22 closed studies (http://​clinicaltrials.​gov/​ct2/​results?​term=​microvascular+ob​struction). In doing so, early signals of detrimental effects may be detected before larger scale trials are embarked upon. An example of this was observed in a 51 patient randomized study of erythropoietin (EPO), in which there was a doubling of MO incidence in the treated group, associated with increased LV volumes acutely post-MI[83]. Although the mechanisms for this are not definitively known, the contribution of increased platelet reactivity and the prothrombotic effects of EPO provide a plausible explanation for the observed increase in MO. Notably, there was a nonsignificant trend toward an increased infarct size in the treated group in this study. In a subsequent trial in which a single high dose of EPO was administered to 110 patients, there was no change in infarct size at 3 months but MO incidence was significantly decreased acutely[93]. The authors postulated that the higher dose and earlier delivery of EPO may have accounted for the differences in MO rates. Nonetheless, both of these studies highlight the potential for MO evaluation to provide better study power as an endpoint compared to infarct size.

A potentially valuable application is in the evaluation of stem cell therapy for acute MI, particularly when such cells are delivered via the intracoronary route. Many of such published human studies have used CMR to predominantly assess LVEF and LV volumes with an increasing number also assessing infarct size[94‐99] and MO[96, 100‐103]. As larger efficacy trials are being conducted and planned, incorporating systematic MO assessment may help in identifying which patients may derive the most benefit (i.e. larger amounts of MO could theoretically impair stem cell engraftment); optimizing the timing of cell delivery in relation to MO temporal evolution; and understanding possible pathophysiologic benefits of stem cells (i.e. is there neovascularization or restoration of microvascular function or are there detrimental effects on microvascular perfusion due to cellular plugging of microvessels[104]).